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 of various fio actions. May be ``all`` for all types
58 or individual types separated by a comma (e.g. ``--debug=file,mem`` will
59 enable file and memory debugging). Currently, additional logging is
63 Dump info related to processes.
65 Dump info related to file actions.
67 Dump info related to I/O queuing.
69 Dump info related to memory allocations.
71 Dump info related to blktrace setup.
73 Dump info related to I/O verification.
75 Enable all debug options.
77 Dump info related to random offset generation.
79 Dump info related to option matching and parsing.
81 Dump info related to disk utilization updates.
83 Dump info only related to job number x.
85 Dump info only related to mutex up/down ops.
87 Dump info related to profile extensions.
89 Dump info related to internal time keeping.
91 Dump info related to networking connections.
93 Dump info related to I/O rate switching.
95 Dump info related to log compress/decompress.
97 Show available debug options.
99 .. option:: --parse-only
101 Parse options only, don't start any I/O.
103 .. option:: --output=filename
105 Write output to file `filename`.
107 .. option:: --output-format=type
109 Set the reporting format to `normal`, `terse`, `json`, or `json+`. Multiple
110 formats can be selected, separated by a comma. `terse` is a CSV based
111 format. `json+` is like `json`, except it adds a full dump of the latency
114 .. option:: --runtime
115 Limit run time to runtime seconds.
117 .. option:: --bandwidth-log
119 Generate aggregate bandwidth logs.
121 .. option:: --minimal
123 Print statistics in a terse, semicolon-delimited format.
125 .. option:: --append-terse
127 Print statistics in selected mode AND terse, semicolon-delimited format.
128 **Deprecated**, use :option:`--output-format` instead to select multiple
131 .. option:: --terse-version=type
133 Set terse version output format (default 3, or 2 or 4 or 5).
135 .. option:: --version
137 Print version information and exit.
141 Print a summary of the command line options and exit.
143 .. option:: --cpuclock-test
145 Perform test and validation of internal CPU clock.
147 .. option:: --crctest=[test]
149 Test the speed of the built-in checksumming functions. If no argument is
150 given, all of them are tested. Alternatively, a comma separated list can
151 be passed, in which case the given ones are tested.
153 .. option:: --cmdhelp=command
155 Print help information for `command`. May be ``all`` for all commands.
157 .. option:: --enghelp=[ioengine[,command]]
159 List all commands defined by :option:`ioengine`, or print help for `command`
160 defined by :option:`ioengine`. If no :option:`ioengine` is given, list all
163 .. option:: --showcmd=jobfile
165 Convert `jobfile` to a set of command-line options.
167 .. option:: --readonly
169 Turn on safety read-only checks, preventing writes. The ``--readonly``
170 option is an extra safety guard to prevent users from accidentally starting
171 a write workload when that is not desired. Fio will only write if
172 `rw=write/randwrite/rw/randrw` is given. This extra safety net can be used
173 as an extra precaution as ``--readonly`` will also enable a write check in
174 the I/O engine core to prevent writes due to unknown user space bug(s).
176 .. option:: --eta=when
178 Specifies when real-time ETA estimate should be printed. `when` may be
179 `always`, `never` or `auto`.
181 .. option:: --eta-newline=time
183 Force a new line for every `time` period passed. When the unit is omitted,
184 the value is interpreted in seconds.
186 .. option:: --status-interval=time
188 Force full status dump every `time` period passed. When the unit is
189 omitted, the value is interpreted in seconds.
191 .. option:: --section=name
193 Only run specified section `name` in job file. Multiple sections can be specified.
194 The ``--section`` option allows one to combine related jobs into one file.
195 E.g. one job file could define light, moderate, and heavy sections. Tell
196 fio to run only the "heavy" section by giving ``--section=heavy``
197 command line option. One can also specify the "write" operations in one
198 section and "verify" operation in another section. The ``--section`` option
199 only applies to job sections. The reserved *global* section is always
202 .. option:: --alloc-size=kb
204 Set the internal smalloc pool size to `kb` in KiB. The
205 ``--alloc-size`` switch allows one to use a larger pool size for smalloc.
206 If running large jobs with randommap enabled, fio can run out of memory.
207 Smalloc is an internal allocator for shared structures from a fixed size
208 memory pool and can grow to 16 pools. The pool size defaults to 16MiB.
210 NOTE: While running :file:`.fio_smalloc.*` backing store files are visible
213 .. option:: --warnings-fatal
215 All fio parser warnings are fatal, causing fio to exit with an
218 .. option:: --max-jobs=nr
220 Set the maximum number of threads/processes to support.
222 .. option:: --server=args
224 Start a backend server, with `args` specifying what to listen to.
225 See `Client/Server`_ section.
227 .. option:: --daemonize=pidfile
229 Background a fio server, writing the pid to the given `pidfile` file.
231 .. option:: --client=hostname
233 Instead of running the jobs locally, send and run them on the given host or
234 set of hosts. See `Client/Server`_ section.
236 .. option:: --remote-config=file
238 Tell fio server to load this local file.
240 .. option:: --idle-prof=option
242 Report CPU idleness. `option` is one of the following:
245 Run unit work calibration only and exit.
248 Show aggregate system idleness and unit work.
251 As **system** but also show per CPU idleness.
253 .. option:: --inflate-log=log
255 Inflate and output compressed log.
257 .. option:: --trigger-file=file
259 Execute trigger cmd when file exists.
261 .. option:: --trigger-timeout=t
263 Execute trigger at this time.
265 .. option:: --trigger=cmd
267 Set this command as local trigger.
269 .. option:: --trigger-remote=cmd
271 Set this command as remote trigger.
273 .. option:: --aux-path=path
275 Use this path for fio state generated files.
277 Any parameters following the options will be assumed to be job files, unless
278 they match a job file parameter. Multiple job files can be listed and each job
279 file will be regarded as a separate group. Fio will :option:`stonewall`
280 execution between each group.
286 As previously described, fio accepts one or more job files describing what it is
287 supposed to do. The job file format is the classic ini file, where the names
288 enclosed in [] brackets define the job name. You are free to use any ASCII name
289 you want, except *global* which has special meaning. Following the job name is
290 a sequence of zero or more parameters, one per line, that define the behavior of
291 the job. If the first character in a line is a ';' or a '#', the entire line is
292 discarded as a comment.
294 A *global* section sets defaults for the jobs described in that file. A job may
295 override a *global* section parameter, and a job file may even have several
296 *global* sections if so desired. A job is only affected by a *global* section
299 The :option:`--cmdhelp` option also lists all options. If used with an `option`
300 argument, :option:`--cmdhelp` will detail the given `option`.
302 See the `examples/` directory for inspiration on how to write job files. Note
303 the copyright and license requirements currently apply to `examples/` files.
305 So let's look at a really simple job file that defines two processes, each
306 randomly reading from a 128MiB file:
310 ; -- start job file --
321 As you can see, the job file sections themselves are empty as all the described
322 parameters are shared. As no :option:`filename` option is given, fio makes up a
323 `filename` for each of the jobs as it sees fit. On the command line, this job
324 would look as follows::
326 $ fio --name=global --rw=randread --size=128m --name=job1 --name=job2
329 Let's look at an example that has a number of processes writing randomly to
334 ; -- start job file --
345 Here we have no *global* section, as we only have one job defined anyway. We
346 want to use async I/O here, with a depth of 4 for each file. We also increased
347 the buffer size used to 32KiB and define numjobs to 4 to fork 4 identical
348 jobs. The result is 4 processes each randomly writing to their own 64MiB
349 file. Instead of using the above job file, you could have given the parameters
350 on the command line. For this case, you would specify::
352 $ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4
354 When fio is utilized as a basis of any reasonably large test suite, it might be
355 desirable to share a set of standardized settings across multiple job files.
356 Instead of copy/pasting such settings, any section may pull in an external
357 :file:`filename.fio` file with *include filename* directive, as in the following
360 ; -- start job file including.fio --
364 include glob-include.fio
371 include test-include.fio
372 ; -- end job file including.fio --
376 ; -- start job file glob-include.fio --
379 ; -- end job file glob-include.fio --
383 ; -- start job file test-include.fio --
386 ; -- end job file test-include.fio --
388 Settings pulled into a section apply to that section only (except *global*
389 section). Include directives may be nested in that any included file may contain
390 further include directive(s). Include files may not contain [] sections.
393 Environment variables
394 ~~~~~~~~~~~~~~~~~~~~~
396 Fio also supports environment variable expansion in job files. Any sub-string of
397 the form ``${VARNAME}`` as part of an option value (in other words, on the right
398 of the '='), will be expanded to the value of the environment variable called
399 `VARNAME`. If no such environment variable is defined, or `VARNAME` is the
400 empty string, the empty string will be substituted.
402 As an example, let's look at a sample fio invocation and job file::
404 $ SIZE=64m NUMJOBS=4 fio jobfile.fio
408 ; -- start job file --
415 This will expand to the following equivalent job file at runtime:
419 ; -- start job file --
426 Fio ships with a few example job files, you can also look there for inspiration.
431 Additionally, fio has a set of reserved keywords that will be replaced
432 internally with the appropriate value. Those keywords are:
436 The architecture page size of the running system.
440 Megabytes of total memory in the system.
444 Number of online available CPUs.
446 These can be used on the command line or in the job file, and will be
447 automatically substituted with the current system values when the job is
448 run. Simple math is also supported on these keywords, so you can perform actions
453 and get that properly expanded to 8 times the size of memory in the machine.
459 This section describes in details each parameter associated with a job. Some
460 parameters take an option of a given type, such as an integer or a
461 string. Anywhere a numeric value is required, an arithmetic expression may be
462 used, provided it is surrounded by parentheses. Supported operators are:
471 For time values in expressions, units are microseconds by default. This is
472 different than for time values not in expressions (not enclosed in
473 parentheses). The following types are used:
480 String: A sequence of alphanumeric characters.
483 Integer with possible time suffix. Without a unit value is interpreted as
484 seconds unless otherwise specified. Accepts a suffix of 'd' for days, 'h' for
485 hours, 'm' for minutes, 's' for seconds, 'ms' (or 'msec') for milliseconds and
486 'us' (or 'usec') for microseconds. For example, use 10m for 10 minutes.
491 Integer. A whole number value, which may contain an integer prefix
492 and an integer suffix:
494 [*integer prefix*] **number** [*integer suffix*]
496 The optional *integer prefix* specifies the number's base. The default
497 is decimal. *0x* specifies hexadecimal.
499 The optional *integer suffix* specifies the number's units, and includes an
500 optional unit prefix and an optional unit. For quantities of data, the
501 default unit is bytes. For quantities of time, the default unit is seconds
502 unless otherwise specified.
504 With :option:`kb_base`\=1000, fio follows international standards for unit
505 prefixes. To specify power-of-10 decimal values defined in the
506 International System of Units (SI):
508 * *Ki* -- means kilo (K) or 1000
509 * *Mi* -- means mega (M) or 1000**2
510 * *Gi* -- means giga (G) or 1000**3
511 * *Ti* -- means tera (T) or 1000**4
512 * *Pi* -- means peta (P) or 1000**5
514 To specify power-of-2 binary values defined in IEC 80000-13:
516 * *K* -- means kibi (Ki) or 1024
517 * *M* -- means mebi (Mi) or 1024**2
518 * *G* -- means gibi (Gi) or 1024**3
519 * *T* -- means tebi (Ti) or 1024**4
520 * *P* -- means pebi (Pi) or 1024**5
522 With :option:`kb_base`\=1024 (the default), the unit prefixes are opposite
523 from those specified in the SI and IEC 80000-13 standards to provide
524 compatibility with old scripts. For example, 4k means 4096.
526 For quantities of data, an optional unit of 'B' may be included
527 (e.g., 'kB' is the same as 'k').
529 The *integer suffix* is not case sensitive (e.g., m/mi mean mebi/mega,
530 not milli). 'b' and 'B' both mean byte, not bit.
532 Examples with :option:`kb_base`\=1000:
534 * *4 KiB*: 4096, 4096b, 4096B, 4ki, 4kib, 4kiB, 4Ki, 4KiB
535 * *1 MiB*: 1048576, 1mi, 1024ki
536 * *1 MB*: 1000000, 1m, 1000k
537 * *1 TiB*: 1099511627776, 1ti, 1024gi, 1048576mi
538 * *1 TB*: 1000000000, 1t, 1000m, 1000000k
540 Examples with :option:`kb_base`\=1024 (default):
542 * *4 KiB*: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB
543 * *1 MiB*: 1048576, 1m, 1024k
544 * *1 MB*: 1000000, 1mi, 1000ki
545 * *1 TiB*: 1099511627776, 1t, 1024g, 1048576m
546 * *1 TB*: 1000000000, 1ti, 1000mi, 1000000ki
548 To specify times (units are not case sensitive):
552 * *M* -- means minutes
553 * *s* -- or sec means seconds (default)
554 * *ms* -- or *msec* means milliseconds
555 * *us* -- or *usec* means microseconds
557 If the option accepts an upper and lower range, use a colon ':' or
558 minus '-' to separate such values. See :ref:`irange <irange>`.
559 If the lower value specified happens to be larger than the upper value
560 the two values are swapped.
565 Boolean. Usually parsed as an integer, however only defined for
566 true and false (1 and 0).
571 Integer range with suffix. Allows value range to be given, such as
572 1024-4096. A colon may also be used as the separator, e.g. 1k:4k. If the
573 option allows two sets of ranges, they can be specified with a ',' or '/'
574 delimiter: 1k-4k/8k-32k. Also see :ref:`int <int>`.
577 A list of floating point numbers, separated by a ':' character.
579 With the above in mind, here follows the complete list of fio job parameters.
585 .. option:: kb_base=int
587 Select the interpretation of unit prefixes in input parameters.
590 Inputs comply with IEC 80000-13 and the International
591 System of Units (SI). Use:
593 - power-of-2 values with IEC prefixes (e.g., KiB)
594 - power-of-10 values with SI prefixes (e.g., kB)
597 Compatibility mode (default). To avoid breaking old scripts:
599 - power-of-2 values with SI prefixes
600 - power-of-10 values with IEC prefixes
602 See :option:`bs` for more details on input parameters.
604 Outputs always use correct prefixes. Most outputs include both
607 bw=2383.3kB/s (2327.4KiB/s)
609 If only one value is reported, then kb_base selects the one to use:
611 **1000** -- SI prefixes
613 **1024** -- IEC prefixes
615 .. option:: unit_base=int
617 Base unit for reporting. Allowed values are:
620 Use auto-detection (default).
632 ASCII name of the job. This may be used to override the name printed by fio
633 for this job. Otherwise the job name is used. On the command line this
634 parameter has the special purpose of also signaling the start of a new job.
636 .. option:: description=str
638 Text description of the job. Doesn't do anything except dump this text
639 description when this job is run. It's not parsed.
641 .. option:: loops=int
643 Run the specified number of iterations of this job. Used to repeat the same
644 workload a given number of times. Defaults to 1.
646 .. option:: numjobs=int
648 Create the specified number of clones of this job. Each clone of job
649 is spawned as an independent thread or process. May be used to setup a
650 larger number of threads/processes doing the same thing. Each thread is
651 reported separately; to see statistics for all clones as a whole, use
652 :option:`group_reporting` in conjunction with :option:`new_group`.
653 See :option:`--max-jobs`. Default: 1.
656 Time related parameters
657 ~~~~~~~~~~~~~~~~~~~~~~~
659 .. option:: runtime=time
661 Tell fio to terminate processing after the specified period of time. It
662 can be quite hard to determine for how long a specified job will run, so
663 this parameter is handy to cap the total runtime to a given time. When
664 the unit is omitted, the value is intepreted in seconds.
666 .. option:: time_based
668 If set, fio will run for the duration of the :option:`runtime` specified
669 even if the file(s) are completely read or written. It will simply loop over
670 the same workload as many times as the :option:`runtime` allows.
672 .. option:: startdelay=irange(time)
674 Delay the start of job for the specified amount of time. Can be a single
675 value or a range. When given as a range, each thread will choose a value
676 randomly from within the range. Value is in seconds if a unit is omitted.
678 .. option:: ramp_time=time
680 If set, fio will run the specified workload for this amount of time before
681 logging any performance numbers. Useful for letting performance settle
682 before logging results, thus minimizing the runtime required for stable
683 results. Note that the ``ramp_time`` is considered lead in time for a job,
684 thus it will increase the total runtime if a special timeout or
685 :option:`runtime` is specified. When the unit is omitted, the value is
688 .. option:: clocksource=str
690 Use the given clocksource as the base of timing. The supported options are:
693 :manpage:`gettimeofday(2)`
696 :manpage:`clock_gettime(2)`
699 Internal CPU clock source
701 cpu is the preferred clocksource if it is reliable, as it is very fast (and
702 fio is heavy on time calls). Fio will automatically use this clocksource if
703 it's supported and considered reliable on the system it is running on,
704 unless another clocksource is specifically set. For x86/x86-64 CPUs, this
705 means supporting TSC Invariant.
707 .. option:: gtod_reduce=bool
709 Enable all of the :manpage:`gettimeofday(2)` reducing options
710 (:option:`disable_clat`, :option:`disable_slat`, :option:`disable_bw_measurement`) plus
711 reduce precision of the timeout somewhat to really shrink the
712 :manpage:`gettimeofday(2)` call count. With this option enabled, we only do
713 about 0.4% of the :manpage:`gettimeofday(2)` calls we would have done if all
714 time keeping was enabled.
716 .. option:: gtod_cpu=int
718 Sometimes it's cheaper to dedicate a single thread of execution to just
719 getting the current time. Fio (and databases, for instance) are very
720 intensive on :manpage:`gettimeofday(2)` calls. With this option, you can set
721 one CPU aside for doing nothing but logging current time to a shared memory
722 location. Then the other threads/processes that run I/O workloads need only
723 copy that segment, instead of entering the kernel with a
724 :manpage:`gettimeofday(2)` call. The CPU set aside for doing these time
725 calls will be excluded from other uses. Fio will manually clear it from the
726 CPU mask of other jobs.
732 .. option:: directory=str
734 Prefix filenames with this directory. Used to place files in a different
735 location than :file:`./`. You can specify a number of directories by
736 separating the names with a ':' character. These directories will be
737 assigned equally distributed to job clones created by :option:`numjobs` as
738 long as they are using generated filenames. If specific `filename(s)` are
739 set fio will use the first listed directory, and thereby matching the
740 `filename` semantic which generates a file each clone if not specified, but
741 let all clones use the same if set.
743 See the :option:`filename` option for information on how to escape "``:``" and
744 "``\``" characters within the directory path itself.
746 .. option:: filename=str
748 Fio normally makes up a `filename` based on the job name, thread number, and
749 file number (see :option:`filename_format`). If you want to share files
750 between threads in a job or several
751 jobs with fixed file paths, specify a `filename` for each of them to override
752 the default. If the ioengine is file based, you can specify a number of files
753 by separating the names with a ':' colon. So if you wanted a job to open
754 :file:`/dev/sda` and :file:`/dev/sdb` as the two working files, you would use
755 ``filename=/dev/sda:/dev/sdb``. This also means that whenever this option is
756 specified, :option:`nrfiles` is ignored. The size of regular files specified
757 by this option will be :option:`size` divided by number of files unless an
758 explicit size is specified by :option:`filesize`.
760 Each colon and backslash in the wanted path must be escaped with a ``\``
761 character. For instance, if the path is :file:`/dev/dsk/foo@3,0:c` then you
762 would use ``filename=/dev/dsk/foo@3,0\:c`` and if the path is
763 :file:`F:\\filename` then you would use ``filename=F\:\\filename``.
765 On Windows, disk devices are accessed as :file:`\\\\.\\PhysicalDrive0` for
766 the first device, :file:`\\\\.\\PhysicalDrive1` for the second etc.
767 Note: Windows and FreeBSD prevent write access to areas
768 of the disk containing in-use data (e.g. filesystems).
770 The filename "`-`" is a reserved name, meaning *stdin* or *stdout*. Which
771 of the two depends on the read/write direction set.
773 .. option:: filename_format=str
775 If sharing multiple files between jobs, it is usually necessary to have fio
776 generate the exact names that you want. By default, fio will name a file
777 based on the default file format specification of
778 :file:`jobname.jobnumber.filenumber`. With this option, that can be
779 customized. Fio will recognize and replace the following keywords in this
783 The name of the worker thread or process.
785 The incremental number of the worker thread or process.
787 The incremental number of the file for that worker thread or
790 To have dependent jobs share a set of files, this option can be set to have
791 fio generate filenames that are shared between the two. For instance, if
792 :file:`testfiles.$filenum` is specified, file number 4 for any job will be
793 named :file:`testfiles.4`. The default of :file:`$jobname.$jobnum.$filenum`
794 will be used if no other format specifier is given.
796 .. option:: unique_filename=bool
798 To avoid collisions between networked clients, fio defaults to prefixing any
799 generated filenames (with a directory specified) with the source of the
800 client connecting. To disable this behavior, set this option to 0.
802 .. option:: opendir=str
804 Recursively open any files below directory `str`.
806 .. option:: lockfile=str
808 Fio defaults to not locking any files before it does I/O to them. If a file
809 or file descriptor is shared, fio can serialize I/O to that file to make the
810 end result consistent. This is usual for emulating real workloads that share
811 files. The lock modes are:
814 No locking. The default.
816 Only one thread or process may do I/O at a time, excluding all
819 Read-write locking on the file. Many readers may
820 access the file at the same time, but writes get exclusive access.
822 .. option:: nrfiles=int
824 Number of files to use for this job. Defaults to 1. The size of files
825 will be :option:`size` divided by this unless explicit size is specified by
826 :option:`filesize`. Files are created for each thread separately, and each
827 file will have a file number within its name by default, as explained in
828 :option:`filename` section.
831 .. option:: openfiles=int
833 Number of files to keep open at the same time. Defaults to the same as
834 :option:`nrfiles`, can be set smaller to limit the number simultaneous
837 .. option:: file_service_type=str
839 Defines how fio decides which file from a job to service next. The following
843 Choose a file at random.
846 Round robin over opened files. This is the default.
849 Finish one file before moving on to the next. Multiple files can
850 still be open depending on 'openfiles'.
853 Use a *Zipf* distribution to decide what file to access.
856 Use a *Pareto* distribution to decide what file to access.
859 Use a *Gaussian* (normal) distribution to decide what file to
865 For *random*, *roundrobin*, and *sequential*, a postfix can be appended to
866 tell fio how many I/Os to issue before switching to a new file. For example,
867 specifying ``file_service_type=random:8`` would cause fio to issue
868 8 I/Os before selecting a new file at random. For the non-uniform
869 distributions, a floating point postfix can be given to influence how the
870 distribution is skewed. See :option:`random_distribution` for a description
871 of how that would work.
873 .. option:: ioscheduler=str
875 Attempt to switch the device hosting the file to the specified I/O scheduler
878 .. option:: create_serialize=bool
880 If true, serialize the file creation for the jobs. This may be handy to
881 avoid interleaving of data files, which may greatly depend on the filesystem
882 used and even the number of processors in the system. Default: true.
884 .. option:: create_fsync=bool
886 :manpage:`fsync(2)` the data file after creation. This is the default.
888 .. option:: create_on_open=bool
890 If true, don't pre-create files but allow the job's open() to create a file
891 when it's time to do I/O. Default: false -- pre-create all necessary files
894 .. option:: create_only=bool
896 If true, fio will only run the setup phase of the job. If files need to be
897 laid out or updated on disk, only that will be done -- the actual job contents
898 are not executed. Default: false.
900 .. option:: allow_file_create=bool
902 If true, fio is permitted to create files as part of its workload. If this
903 option is false, then fio will error out if
904 the files it needs to use don't already exist. Default: true.
906 .. option:: allow_mounted_write=bool
908 If this isn't set, fio will abort jobs that are destructive (e.g. that write)
909 to what appears to be a mounted device or partition. This should help catch
910 creating inadvertently destructive tests, not realizing that the test will
911 destroy data on the mounted file system. Note that some platforms don't allow
912 writing against a mounted device regardless of this option. Default: false.
914 .. option:: pre_read=bool
916 If this is given, files will be pre-read into memory before starting the
917 given I/O operation. This will also clear the :option:`invalidate` flag,
918 since it is pointless to pre-read and then drop the cache. This will only
919 work for I/O engines that are seek-able, since they allow you to read the
920 same data multiple times. Thus it will not work on non-seekable I/O engines
921 (e.g. network, splice). Default: false.
923 .. option:: unlink=bool
925 Unlink the job files when done. Not the default, as repeated runs of that
926 job would then waste time recreating the file set again and again. Default:
929 .. option:: unlink_each_loop=bool
931 Unlink job files after each iteration or loop. Default: false.
933 .. option:: zonesize=int
935 Divide a file into zones of the specified size. See :option:`zoneskip`.
937 .. option:: zonerange=int
939 Give size of an I/O zone. See :option:`zoneskip`.
941 .. option:: zoneskip=int
943 Skip the specified number of bytes when :option:`zonesize` data has been
944 read. The two zone options can be used to only do I/O on zones of a file.
950 .. option:: direct=bool
952 If value is true, use non-buffered I/O. This is usually O_DIRECT. Note that
953 ZFS on Solaris doesn't support direct I/O. On Windows the synchronous
954 ioengines don't support direct I/O. Default: false.
956 .. option:: atomic=bool
958 If value is true, attempt to use atomic direct I/O. Atomic writes are
959 guaranteed to be stable once acknowledged by the operating system. Only
960 Linux supports O_ATOMIC right now.
962 .. option:: buffered=bool
964 If value is true, use buffered I/O. This is the opposite of the
965 :option:`direct` option. Defaults to true.
967 .. option:: readwrite=str, rw=str
969 Type of I/O pattern. Accepted values are:
976 Sequential trims (Linux block devices only).
982 Random trims (Linux block devices only).
984 Sequential mixed reads and writes.
986 Random mixed reads and writes.
988 Sequential trim+write sequences. Blocks will be trimmed first,
989 then the same blocks will be written to.
991 Fio defaults to read if the option is not specified. For the mixed I/O
992 types, the default is to split them 50/50. For certain types of I/O the
993 result may still be skewed a bit, since the speed may be different.
995 It is possible to specify the number of I/Os to do before getting a new
996 offset by appending ``:<nr>`` to the end of the string given. For a
997 random read, it would look like ``rw=randread:8`` for passing in an offset
998 modifier with a value of 8. If the suffix is used with a sequential I/O
999 pattern, then the *<nr>* value specified will be **added** to the generated
1000 offset for each I/O turning sequential I/O into sequential I/O with holes.
1001 For instance, using ``rw=write:4k`` will skip 4k for every write. Also see
1002 the :option:`rw_sequencer` option.
1004 .. option:: rw_sequencer=str
1006 If an offset modifier is given by appending a number to the ``rw=<str>``
1007 line, then this option controls how that number modifies the I/O offset
1008 being generated. Accepted values are:
1011 Generate sequential offset.
1013 Generate the same offset.
1015 ``sequential`` is only useful for random I/O, where fio would normally
1016 generate a new random offset for every I/O. If you append e.g. 8 to randread,
1017 you would get a new random offset for every 8 I/Os. The result would be a
1018 seek for only every 8 I/Os, instead of for every I/O. Use ``rw=randread:8``
1019 to specify that. As sequential I/O is already sequential, setting
1020 ``sequential`` for that would not result in any differences. ``identical``
1021 behaves in a similar fashion, except it sends the same offset 8 number of
1022 times before generating a new offset.
1024 .. option:: unified_rw_reporting=bool
1026 Fio normally reports statistics on a per data direction basis, meaning that
1027 reads, writes, and trims are accounted and reported separately. If this
1028 option is set fio sums the results and report them as "mixed" instead.
1030 .. option:: randrepeat=bool
1032 Seed the random number generator used for random I/O patterns in a
1033 predictable way so the pattern is repeatable across runs. Default: true.
1035 .. option:: allrandrepeat=bool
1037 Seed all random number generators in a predictable way so results are
1038 repeatable across runs. Default: false.
1040 .. option:: randseed=int
1042 Seed the random number generators based on this seed value, to be able to
1043 control what sequence of output is being generated. If not set, the random
1044 sequence depends on the :option:`randrepeat` setting.
1046 .. option:: fallocate=str
1048 Whether pre-allocation is performed when laying down files.
1049 Accepted values are:
1052 Do not pre-allocate space.
1055 Use a platform's native pre-allocation call but fall back to
1056 **none** behavior if it fails/is not implemented.
1059 Pre-allocate via :manpage:`posix_fallocate(3)`.
1062 Pre-allocate via :manpage:`fallocate(2)` with
1063 FALLOC_FL_KEEP_SIZE set.
1066 Backward-compatible alias for **none**.
1069 Backward-compatible alias for **posix**.
1071 May not be available on all supported platforms. **keep** is only available
1072 on Linux. If using ZFS on Solaris this cannot be set to **posix**
1073 because ZFS doesn't support pre-allocation. Default: **native** if any
1074 pre-allocation methods are available, **none** if not.
1076 .. option:: fadvise_hint=str
1078 Use :manpage:`posix_fadvise(2)` to advise the kernel on what I/O patterns
1079 are likely to be issued. Accepted values are:
1082 Backwards-compatible hint for "no hint".
1085 Backwards compatible hint for "advise with fio workload type". This
1086 uses **FADV_RANDOM** for a random workload, and **FADV_SEQUENTIAL**
1087 for a sequential workload.
1090 Advise using **FADV_SEQUENTIAL**.
1093 Advise using **FADV_RANDOM**.
1095 .. option:: write_hint=str
1097 Use :manpage:`fcntl(2)` to advise the kernel what life time to expect
1098 from a write. Only supported on Linux, as of version 4.13. Accepted
1102 No particular life time associated with this file.
1105 Data written to this file has a short life time.
1108 Data written to this file has a medium life time.
1111 Data written to this file has a long life time.
1114 Data written to this file has a very long life time.
1116 The values are all relative to each other, and no absolute meaning
1117 should be associated with them.
1119 .. option:: offset=int
1121 Start I/O at the provided offset in the file, given as either a fixed size in
1122 bytes or a percentage. If a percentage is given, the next ``blockalign``-ed
1123 offset will be used. Data before the given offset will not be touched. This
1124 effectively caps the file size at `real_size - offset`. Can be combined with
1125 :option:`size` to constrain the start and end range of the I/O workload.
1126 A percentage can be specified by a number between 1 and 100 followed by '%',
1127 for example, ``offset=20%`` to specify 20%.
1129 .. option:: offset_increment=int
1131 If this is provided, then the real offset becomes `offset + offset_increment
1132 * thread_number`, where the thread number is a counter that starts at 0 and
1133 is incremented for each sub-job (i.e. when :option:`numjobs` option is
1134 specified). This option is useful if there are several jobs which are
1135 intended to operate on a file in parallel disjoint segments, with even
1136 spacing between the starting points.
1138 .. option:: number_ios=int
1140 Fio will normally perform I/Os until it has exhausted the size of the region
1141 set by :option:`size`, or if it exhaust the allocated time (or hits an error
1142 condition). With this setting, the range/size can be set independently of
1143 the number of I/Os to perform. When fio reaches this number, it will exit
1144 normally and report status. Note that this does not extend the amount of I/O
1145 that will be done, it will only stop fio if this condition is met before
1146 other end-of-job criteria.
1148 .. option:: fsync=int
1150 If writing to a file, issue an :manpage:`fsync(2)` (or its equivalent) of
1151 the dirty data for every number of blocks given. For example, if you give 32
1152 as a parameter, fio will sync the file after every 32 writes issued. If fio is
1153 using non-buffered I/O, we may not sync the file. The exception is the sg
1154 I/O engine, which synchronizes the disk cache anyway. Defaults to 0, which
1155 means fio does not periodically issue and wait for a sync to complete. Also
1156 see :option:`end_fsync` and :option:`fsync_on_close`.
1158 .. option:: fdatasync=int
1160 Like :option:`fsync` but uses :manpage:`fdatasync(2)` to only sync data and
1161 not metadata blocks. In Windows, FreeBSD, and DragonFlyBSD there is no
1162 :manpage:`fdatasync(2)` so this falls back to using :manpage:`fsync(2)`.
1163 Defaults to 0, which means fio does not periodically issue and wait for a
1164 data-only sync to complete.
1166 .. option:: write_barrier=int
1168 Make every `N-th` write a barrier write.
1170 .. option:: sync_file_range=str:val
1172 Use :manpage:`sync_file_range(2)` for every `val` number of write
1173 operations. Fio will track range of writes that have happened since the last
1174 :manpage:`sync_file_range(2)` call. `str` can currently be one or more of:
1177 SYNC_FILE_RANGE_WAIT_BEFORE
1179 SYNC_FILE_RANGE_WRITE
1181 SYNC_FILE_RANGE_WAIT_AFTER
1183 So if you do ``sync_file_range=wait_before,write:8``, fio would use
1184 ``SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE`` for every 8
1185 writes. Also see the :manpage:`sync_file_range(2)` man page. This option is
1188 .. option:: overwrite=bool
1190 If true, writes to a file will always overwrite existing data. If the file
1191 doesn't already exist, it will be created before the write phase begins. If
1192 the file exists and is large enough for the specified write phase, nothing
1193 will be done. Default: false.
1195 .. option:: end_fsync=bool
1197 If true, :manpage:`fsync(2)` file contents when a write stage has completed.
1200 .. option:: fsync_on_close=bool
1202 If true, fio will :manpage:`fsync(2)` a dirty file on close. This differs
1203 from :option:`end_fsync` in that it will happen on every file close, not
1204 just at the end of the job. Default: false.
1206 .. option:: rwmixread=int
1208 Percentage of a mixed workload that should be reads. Default: 50.
1210 .. option:: rwmixwrite=int
1212 Percentage of a mixed workload that should be writes. If both
1213 :option:`rwmixread` and :option:`rwmixwrite` is given and the values do not
1214 add up to 100%, the latter of the two will be used to override the
1215 first. This may interfere with a given rate setting, if fio is asked to
1216 limit reads or writes to a certain rate. If that is the case, then the
1217 distribution may be skewed. Default: 50.
1219 .. option:: random_distribution=str:float[,str:float][,str:float]
1221 By default, fio will use a completely uniform random distribution when asked
1222 to perform random I/O. Sometimes it is useful to skew the distribution in
1223 specific ways, ensuring that some parts of the data is more hot than others.
1224 fio includes the following distribution models:
1227 Uniform random distribution
1236 Normal (Gaussian) distribution
1239 Zoned random distribution
1241 When using a **zipf** or **pareto** distribution, an input value is also
1242 needed to define the access pattern. For **zipf**, this is the `zipf
1243 theta`. For **pareto**, it's the `Pareto power`. Fio includes a test
1244 program, :command:`genzipf`, that can be used visualize what the given input
1245 values will yield in terms of hit rates. If you wanted to use **zipf** with
1246 a `theta` of 1.2, you would use ``random_distribution=zipf:1.2`` as the
1247 option. If a non-uniform model is used, fio will disable use of the random
1248 map. For the **normal** distribution, a normal (Gaussian) deviation is
1249 supplied as a value between 0 and 100.
1251 For a **zoned** distribution, fio supports specifying percentages of I/O
1252 access that should fall within what range of the file or device. For
1253 example, given a criteria of:
1255 * 60% of accesses should be to the first 10%
1256 * 30% of accesses should be to the next 20%
1257 * 8% of accesses should be to the next 30%
1258 * 2% of accesses should be to the next 40%
1260 we can define that through zoning of the random accesses. For the above
1261 example, the user would do::
1263 random_distribution=zoned:60/10:30/20:8/30:2/40
1265 similarly to how :option:`bssplit` works for setting ranges and percentages
1266 of block sizes. Like :option:`bssplit`, it's possible to specify separate
1267 zones for reads, writes, and trims. If just one set is given, it'll apply to
1270 .. option:: percentage_random=int[,int][,int]
1272 For a random workload, set how big a percentage should be random. This
1273 defaults to 100%, in which case the workload is fully random. It can be set
1274 from anywhere from 0 to 100. Setting it to 0 would make the workload fully
1275 sequential. Any setting in between will result in a random mix of sequential
1276 and random I/O, at the given percentages. Comma-separated values may be
1277 specified for reads, writes, and trims as described in :option:`blocksize`.
1279 .. option:: norandommap
1281 Normally fio will cover every block of the file when doing random I/O. If
1282 this option is given, fio will just get a new random offset without looking
1283 at past I/O history. This means that some blocks may not be read or written,
1284 and that some blocks may be read/written more than once. If this option is
1285 used with :option:`verify` and multiple blocksizes (via :option:`bsrange`),
1286 only intact blocks are verified, i.e., partially-overwritten blocks are
1289 .. option:: softrandommap=bool
1291 See :option:`norandommap`. If fio runs with the random block map enabled and
1292 it fails to allocate the map, if this option is set it will continue without
1293 a random block map. As coverage will not be as complete as with random maps,
1294 this option is disabled by default.
1296 .. option:: random_generator=str
1298 Fio supports the following engines for generating
1299 I/O offsets for random I/O:
1302 Strong 2^88 cycle random number generator
1304 Linear feedback shift register generator
1306 Strong 64-bit 2^258 cycle random number generator
1308 **tausworthe** is a strong random number generator, but it requires tracking
1309 on the side if we want to ensure that blocks are only read or written
1310 once. **LFSR** guarantees that we never generate the same offset twice, and
1311 it's also less computationally expensive. It's not a true random generator,
1312 however, though for I/O purposes it's typically good enough. **LFSR** only
1313 works with single block sizes, not with workloads that use multiple block
1314 sizes. If used with such a workload, fio may read or write some blocks
1315 multiple times. The default value is **tausworthe**, unless the required
1316 space exceeds 2^32 blocks. If it does, then **tausworthe64** is
1317 selected automatically.
1323 .. option:: blocksize=int[,int][,int], bs=int[,int][,int]
1325 The block size in bytes used for I/O units. Default: 4096. A single value
1326 applies to reads, writes, and trims. Comma-separated values may be
1327 specified for reads, writes, and trims. A value not terminated in a comma
1328 applies to subsequent types.
1333 means 256k for reads, writes and trims.
1336 means 8k for reads, 32k for writes and trims.
1339 means 8k for reads, 32k for writes, and default for trims.
1342 means default for reads, 8k for writes and trims.
1345 means default for reads, 8k for writes, and default for trims.
1347 .. option:: blocksize_range=irange[,irange][,irange], bsrange=irange[,irange][,irange]
1349 A range of block sizes in bytes for I/O units. The issued I/O unit will
1350 always be a multiple of the minimum size, unless
1351 :option:`blocksize_unaligned` is set.
1353 Comma-separated ranges may be specified for reads, writes, and trims as
1354 described in :option:`blocksize`.
1356 Example: ``bsrange=1k-4k,2k-8k``.
1358 .. option:: bssplit=str[,str][,str]
1360 Sometimes you want even finer grained control of the block sizes issued, not
1361 just an even split between them. This option allows you to weight various
1362 block sizes, so that you are able to define a specific amount of block sizes
1363 issued. The format for this option is::
1365 bssplit=blocksize/percentage:blocksize/percentage
1367 for as many block sizes as needed. So if you want to define a workload that
1368 has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would write::
1370 bssplit=4k/10:64k/50:32k/40
1372 Ordering does not matter. If the percentage is left blank, fio will fill in
1373 the remaining values evenly. So a bssplit option like this one::
1375 bssplit=4k/50:1k/:32k/
1377 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always add up
1378 to 100, if bssplit is given a range that adds up to more, it will error out.
1380 Comma-separated values may be specified for reads, writes, and trims as
1381 described in :option:`blocksize`.
1383 If you want a workload that has 50% 2k reads and 50% 4k reads, while having
1384 90% 4k writes and 10% 8k writes, you would specify::
1386 bssplit=2k/50:4k/50,4k/90,8k/10
1388 .. option:: blocksize_unaligned, bs_unaligned
1390 If set, fio will issue I/O units with any size within
1391 :option:`blocksize_range`, not just multiples of the minimum size. This
1392 typically won't work with direct I/O, as that normally requires sector
1395 .. option:: bs_is_seq_rand=bool
1397 If this option is set, fio will use the normal read,write blocksize settings
1398 as sequential,random blocksize settings instead. Any random read or write
1399 will use the WRITE blocksize settings, and any sequential read or write will
1400 use the READ blocksize settings.
1402 .. option:: blockalign=int[,int][,int], ba=int[,int][,int]
1404 Boundary to which fio will align random I/O units. Default:
1405 :option:`blocksize`. Minimum alignment is typically 512b for using direct
1406 I/O, though it usually depends on the hardware block size. This option is
1407 mutually exclusive with using a random map for files, so it will turn off
1408 that option. Comma-separated values may be specified for reads, writes, and
1409 trims as described in :option:`blocksize`.
1415 .. option:: zero_buffers
1417 Initialize buffers with all zeros. Default: fill buffers with random data.
1419 .. option:: refill_buffers
1421 If this option is given, fio will refill the I/O buffers on every
1422 submit. The default is to only fill it at init time and reuse that
1423 data. Only makes sense if zero_buffers isn't specified, naturally. If data
1424 verification is enabled, `refill_buffers` is also automatically enabled.
1426 .. option:: scramble_buffers=bool
1428 If :option:`refill_buffers` is too costly and the target is using data
1429 deduplication, then setting this option will slightly modify the I/O buffer
1430 contents to defeat normal de-dupe attempts. This is not enough to defeat
1431 more clever block compression attempts, but it will stop naive dedupe of
1432 blocks. Default: true.
1434 .. option:: buffer_compress_percentage=int
1436 If this is set, then fio will attempt to provide I/O buffer content (on
1437 WRITEs) that compresses to the specified level. Fio does this by providing a
1438 mix of random data and a fixed pattern. The fixed pattern is either zeros,
1439 or the pattern specified by :option:`buffer_pattern`. If the pattern option
1440 is used, it might skew the compression ratio slightly. Note that this is per
1441 block size unit, for file/disk wide compression level that matches this
1442 setting, you'll also want to set :option:`refill_buffers`.
1444 .. option:: buffer_compress_chunk=int
1446 See :option:`buffer_compress_percentage`. This setting allows fio to manage
1447 how big the ranges of random data and zeroed data is. Without this set, fio
1448 will provide :option:`buffer_compress_percentage` of blocksize random data,
1449 followed by the remaining zeroed. With this set to some chunk size smaller
1450 than the block size, fio can alternate random and zeroed data throughout the
1453 .. option:: buffer_pattern=str
1455 If set, fio will fill the I/O buffers with this pattern or with the contents
1456 of a file. If not set, the contents of I/O buffers are defined by the other
1457 options related to buffer contents. The setting can be any pattern of bytes,
1458 and can be prefixed with 0x for hex values. It may also be a string, where
1459 the string must then be wrapped with ``""``. Or it may also be a filename,
1460 where the filename must be wrapped with ``''`` in which case the file is
1461 opened and read. Note that not all the file contents will be read if that
1462 would cause the buffers to overflow. So, for example::
1464 buffer_pattern='filename'
1468 buffer_pattern="abcd"
1476 buffer_pattern=0xdeadface
1478 Also you can combine everything together in any order::
1480 buffer_pattern=0xdeadface"abcd"-12'filename'
1482 .. option:: dedupe_percentage=int
1484 If set, fio will generate this percentage of identical buffers when
1485 writing. These buffers will be naturally dedupable. The contents of the
1486 buffers depend on what other buffer compression settings have been set. It's
1487 possible to have the individual buffers either fully compressible, or not at
1488 all. This option only controls the distribution of unique buffers.
1490 .. option:: invalidate=bool
1492 Invalidate the buffer/page cache parts of the files to be used prior to
1493 starting I/O if the platform and file type support it. Defaults to true.
1494 This will be ignored if :option:`pre_read` is also specified for the
1497 .. option:: sync=bool
1499 Use synchronous I/O for buffered writes. For the majority of I/O engines,
1500 this means using O_SYNC. Default: false.
1502 .. option:: iomem=str, mem=str
1504 Fio can use various types of memory as the I/O unit buffer. The allowed
1508 Use memory from :manpage:`malloc(3)` as the buffers. Default memory
1512 Use shared memory as the buffers. Allocated through
1513 :manpage:`shmget(2)`.
1516 Same as shm, but use huge pages as backing.
1519 Use :manpage:`mmap(2)` to allocate buffers. May either be anonymous memory, or can
1520 be file backed if a filename is given after the option. The format
1521 is `mem=mmap:/path/to/file`.
1524 Use a memory mapped huge file as the buffer backing. Append filename
1525 after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file`.
1528 Same as mmap, but use a MMAP_SHARED mapping.
1531 Use GPU memory as the buffers for GPUDirect RDMA benchmark.
1532 The ioengine must be rdma.
1534 The area allocated is a function of the maximum allowed bs size for the job,
1535 multiplied by the I/O depth given. Note that for **shmhuge** and
1536 **mmaphuge** to work, the system must have free huge pages allocated. This
1537 can normally be checked and set by reading/writing
1538 :file:`/proc/sys/vm/nr_hugepages` on a Linux system. Fio assumes a huge page
1539 is 4MiB in size. So to calculate the number of huge pages you need for a
1540 given job file, add up the I/O depth of all jobs (normally one unless
1541 :option:`iodepth` is used) and multiply by the maximum bs set. Then divide
1542 that number by the huge page size. You can see the size of the huge pages in
1543 :file:`/proc/meminfo`. If no huge pages are allocated by having a non-zero
1544 number in `nr_hugepages`, using **mmaphuge** or **shmhuge** will fail. Also
1545 see :option:`hugepage-size`.
1547 **mmaphuge** also needs to have hugetlbfs mounted and the file location
1548 should point there. So if it's mounted in :file:`/huge`, you would use
1549 `mem=mmaphuge:/huge/somefile`.
1551 .. option:: iomem_align=int
1553 This indicates the memory alignment of the I/O memory buffers. Note that
1554 the given alignment is applied to the first I/O unit buffer, if using
1555 :option:`iodepth` the alignment of the following buffers are given by the
1556 :option:`bs` used. In other words, if using a :option:`bs` that is a
1557 multiple of the page sized in the system, all buffers will be aligned to
1558 this value. If using a :option:`bs` that is not page aligned, the alignment
1559 of subsequent I/O memory buffers is the sum of the :option:`iomem_align` and
1562 .. option:: hugepage-size=int
1564 Defines the size of a huge page. Must at least be equal to the system
1565 setting, see :file:`/proc/meminfo`. Defaults to 4MiB. Should probably
1566 always be a multiple of megabytes, so using ``hugepage-size=Xm`` is the
1567 preferred way to set this to avoid setting a non-pow-2 bad value.
1569 .. option:: lockmem=int
1571 Pin the specified amount of memory with :manpage:`mlock(2)`. Can be used to
1572 simulate a smaller amount of memory. The amount specified is per worker.
1578 .. option:: size=int
1580 The total size of file I/O for each thread of this job. Fio will run until
1581 this many bytes has been transferred, unless runtime is limited by other options
1582 (such as :option:`runtime`, for instance, or increased/decreased by :option:`io_size`).
1583 Fio will divide this size between the available files determined by options
1584 such as :option:`nrfiles`, :option:`filename`, unless :option:`filesize` is
1585 specified by the job. If the result of division happens to be 0, the size is
1586 set to the physical size of the given files or devices if they exist.
1587 If this option is not specified, fio will use the full size of the given
1588 files or devices. If the files do not exist, size must be given. It is also
1589 possible to give size as a percentage between 1 and 100. If ``size=20%`` is
1590 given, fio will use 20% of the full size of the given files or devices.
1591 Can be combined with :option:`offset` to constrain the start and end range
1592 that I/O will be done within.
1594 .. option:: io_size=int, io_limit=int
1596 Normally fio operates within the region set by :option:`size`, which means
1597 that the :option:`size` option sets both the region and size of I/O to be
1598 performed. Sometimes that is not what you want. With this option, it is
1599 possible to define just the amount of I/O that fio should do. For instance,
1600 if :option:`size` is set to 20GiB and :option:`io_size` is set to 5GiB, fio
1601 will perform I/O within the first 20GiB but exit when 5GiB have been
1602 done. The opposite is also possible -- if :option:`size` is set to 20GiB,
1603 and :option:`io_size` is set to 40GiB, then fio will do 40GiB of I/O within
1604 the 0..20GiB region.
1606 .. option:: filesize=irange(int)
1608 Individual file sizes. May be a range, in which case fio will select sizes
1609 for files at random within the given range and limited to :option:`size` in
1610 total (if that is given). If not given, each created file is the same size.
1611 This option overrides :option:`size` in terms of file size, which means
1612 this value is used as a fixed size or possible range of each file.
1614 .. option:: file_append=bool
1616 Perform I/O after the end of the file. Normally fio will operate within the
1617 size of a file. If this option is set, then fio will append to the file
1618 instead. This has identical behavior to setting :option:`offset` to the size
1619 of a file. This option is ignored on non-regular files.
1621 .. option:: fill_device=bool, fill_fs=bool
1623 Sets size to something really large and waits for ENOSPC (no space left on
1624 device) as the terminating condition. Only makes sense with sequential
1625 write. For a read workload, the mount point will be filled first then I/O
1626 started on the result. This option doesn't make sense if operating on a raw
1627 device node, since the size of that is already known by the file system.
1628 Additionally, writing beyond end-of-device will not return ENOSPC there.
1634 .. option:: ioengine=str
1636 Defines how the job issues I/O to the file. The following types are defined:
1639 Basic :manpage:`read(2)` or :manpage:`write(2)`
1640 I/O. :manpage:`lseek(2)` is used to position the I/O location.
1641 See :option:`fsync` and :option:`fdatasync` for syncing write I/Os.
1644 Basic :manpage:`pread(2)` or :manpage:`pwrite(2)` I/O. Default on
1645 all supported operating systems except for Windows.
1648 Basic :manpage:`readv(2)` or :manpage:`writev(2)` I/O. Will emulate
1649 queuing by coalescing adjacent I/Os into a single submission.
1652 Basic :manpage:`preadv(2)` or :manpage:`pwritev(2)` I/O.
1655 Basic :manpage:`preadv2(2)` or :manpage:`pwritev2(2)` I/O.
1658 Linux native asynchronous I/O. Note that Linux may only support
1659 queued behavior with non-buffered I/O (set ``direct=1`` or
1661 This engine defines engine specific options.
1664 POSIX asynchronous I/O using :manpage:`aio_read(3)` and
1665 :manpage:`aio_write(3)`.
1668 Solaris native asynchronous I/O.
1671 Windows native asynchronous I/O. Default on Windows.
1674 File is memory mapped with :manpage:`mmap(2)` and data copied
1675 to/from using :manpage:`memcpy(3)`.
1678 :manpage:`splice(2)` is used to transfer the data and
1679 :manpage:`vmsplice(2)` to transfer data from user space to the
1683 SCSI generic sg v3 I/O. May either be synchronous using the SG_IO
1684 ioctl, or if the target is an sg character device we use
1685 :manpage:`read(2)` and :manpage:`write(2)` for asynchronous
1686 I/O. Requires filename option to specify either block or character
1690 Doesn't transfer any data, just pretends to. This is mainly used to
1691 exercise fio itself and for debugging/testing purposes.
1694 Transfer over the network to given ``host:port``. Depending on the
1695 :option:`protocol` used, the :option:`hostname`, :option:`port`,
1696 :option:`listen` and :option:`filename` options are used to specify
1697 what sort of connection to make, while the :option:`protocol` option
1698 determines which protocol will be used. This engine defines engine
1702 Like **net**, but uses :manpage:`splice(2)` and
1703 :manpage:`vmsplice(2)` to map data and send/receive.
1704 This engine defines engine specific options.
1707 Doesn't transfer any data, but burns CPU cycles according to the
1708 :option:`cpuload` and :option:`cpuchunks` options. Setting
1709 :option:`cpuload`\=85 will cause that job to do nothing but burn 85%
1710 of the CPU. In case of SMP machines, use :option:`numjobs`
1711 =<no_of_cpu> to get desired CPU usage, as the cpuload only loads a
1712 single CPU at the desired rate. A job never finishes unless there is
1713 at least one non-cpuio job.
1716 The GUASI I/O engine is the Generic Userspace Asyncronous Syscall
1717 Interface approach to async I/O. See
1719 http://www.xmailserver.org/guasi-lib.html
1721 for more info on GUASI.
1724 The RDMA I/O engine supports both RDMA memory semantics
1725 (RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the
1726 InfiniBand, RoCE and iWARP protocols.
1729 I/O engine that does regular fallocate to simulate data transfer as
1733 does fallocate(,mode = FALLOC_FL_KEEP_SIZE,).
1736 does fallocate(,mode = 0).
1739 does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE).
1742 I/O engine that sends :manpage:`ftruncate(2)` operations in response
1743 to write (DDIR_WRITE) events. Each ftruncate issued sets the file's
1744 size to the current block offset. Block size is ignored.
1747 I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate
1748 defragment activity in request to DDIR_WRITE event.
1751 I/O engine supporting direct access to Ceph Rados Block Devices
1752 (RBD) via librbd without the need to use the kernel rbd driver. This
1753 ioengine defines engine specific options.
1756 Using GlusterFS libgfapi sync interface to direct access to
1757 GlusterFS volumes without having to go through FUSE. This ioengine
1758 defines engine specific options.
1761 Using GlusterFS libgfapi async interface to direct access to
1762 GlusterFS volumes without having to go through FUSE. This ioengine
1763 defines engine specific options.
1766 Read and write through Hadoop (HDFS). The :file:`filename` option
1767 is used to specify host,port of the hdfs name-node to connect. This
1768 engine interprets offsets a little differently. In HDFS, files once
1769 created cannot be modified so random writes are not possible. To
1770 imitate this the libhdfs engine expects a bunch of small files to be
1771 created over HDFS and will randomly pick a file from them
1772 based on the offset generated by fio backend (see the example
1773 job file to create such files, use ``rw=write`` option). Please
1774 note, it may be necessary to set environment variables to work
1775 with HDFS/libhdfs properly. Each job uses its own connection to
1779 Read, write and erase an MTD character device (e.g.,
1780 :file:`/dev/mtd0`). Discards are treated as erases. Depending on the
1781 underlying device type, the I/O may have to go in a certain pattern,
1782 e.g., on NAND, writing sequentially to erase blocks and discarding
1783 before overwriting. The `trimwrite` mode works well for this
1787 Read and write using filesystem DAX to a file on a filesystem
1788 mounted with DAX on a persistent memory device through the NVML
1792 Read and write using device DAX to a persistent memory device (e.g.,
1793 /dev/dax0.0) through the NVML libpmem library.
1796 Prefix to specify loading an external I/O engine object file. Append
1797 the engine filename, e.g. ``ioengine=external:/tmp/foo.o`` to load
1798 ioengine :file:`foo.o` in :file:`/tmp`.
1801 I/O engine specific parameters
1802 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1804 In addition, there are some parameters which are only valid when a specific
1805 ioengine is in use. These are used identically to normal parameters, with the
1806 caveat that when used on the command line, they must come after the
1807 :option:`ioengine` that defines them is selected.
1809 .. option:: userspace_reap : [libaio]
1811 Normally, with the libaio engine in use, fio will use the
1812 :manpage:`io_getevents(2)` system call to reap newly returned events. With
1813 this flag turned on, the AIO ring will be read directly from user-space to
1814 reap events. The reaping mode is only enabled when polling for a minimum of
1815 0 events (e.g. when :option:`iodepth_batch_complete` `=0`).
1817 .. option:: hipri : [pvsync2]
1819 Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority
1822 .. option:: hipri_percentage : [pvsync2]
1824 When hipri is set this determines the probability of a pvsync2 IO being high
1825 priority. The default is 100%.
1827 .. option:: cpuload=int : [cpuio]
1829 Attempt to use the specified percentage of CPU cycles. This is a mandatory
1830 option when using cpuio I/O engine.
1832 .. option:: cpuchunks=int : [cpuio]
1834 Split the load into cycles of the given time. In microseconds.
1836 .. option:: exit_on_io_done=bool : [cpuio]
1838 Detect when I/O threads are done, then exit.
1840 .. option:: hostname=str : [netsplice] [net]
1842 The hostname or IP address to use for TCP or UDP based I/O. If the job is
1843 a TCP listener or UDP reader, the hostname is not used and must be omitted
1844 unless it is a valid UDP multicast address.
1846 .. option:: namenode=str : [libhdfs]
1848 The hostname or IP address of a HDFS cluster namenode to contact.
1850 .. option:: port=int
1854 The TCP or UDP port to bind to or connect to. If this is used with
1855 :option:`numjobs` to spawn multiple instances of the same job type, then
1856 this will be the starting port number since fio will use a range of
1861 The listening port of the HFDS cluster namenode.
1863 .. option:: interface=str : [netsplice] [net]
1865 The IP address of the network interface used to send or receive UDP
1868 .. option:: ttl=int : [netsplice] [net]
1870 Time-to-live value for outgoing UDP multicast packets. Default: 1.
1872 .. option:: nodelay=bool : [netsplice] [net]
1874 Set TCP_NODELAY on TCP connections.
1876 .. option:: protocol=str : [netsplice] [net]
1878 .. option:: proto=str : [netsplice] [net]
1880 The network protocol to use. Accepted values are:
1883 Transmission control protocol.
1885 Transmission control protocol V6.
1887 User datagram protocol.
1889 User datagram protocol V6.
1893 When the protocol is TCP or UDP, the port must also be given, as well as the
1894 hostname if the job is a TCP listener or UDP reader. For unix sockets, the
1895 normal filename option should be used and the port is invalid.
1897 .. option:: listen : [netsplice] [net]
1899 For TCP network connections, tell fio to listen for incoming connections
1900 rather than initiating an outgoing connection. The :option:`hostname` must
1901 be omitted if this option is used.
1903 .. option:: pingpong : [netsplice] [net]
1905 Normally a network writer will just continue writing data, and a network
1906 reader will just consume packages. If ``pingpong=1`` is set, a writer will
1907 send its normal payload to the reader, then wait for the reader to send the
1908 same payload back. This allows fio to measure network latencies. The
1909 submission and completion latencies then measure local time spent sending or
1910 receiving, and the completion latency measures how long it took for the
1911 other end to receive and send back. For UDP multicast traffic
1912 ``pingpong=1`` should only be set for a single reader when multiple readers
1913 are listening to the same address.
1915 .. option:: window_size : [netsplice] [net]
1917 Set the desired socket buffer size for the connection.
1919 .. option:: mss : [netsplice] [net]
1921 Set the TCP maximum segment size (TCP_MAXSEG).
1923 .. option:: donorname=str : [e4defrag]
1925 File will be used as a block donor (swap extents between files).
1927 .. option:: inplace=int : [e4defrag]
1929 Configure donor file blocks allocation strategy:
1932 Default. Preallocate donor's file on init.
1934 Allocate space immediately inside defragment event, and free right
1937 .. option:: clustername=str : [rbd]
1939 Specifies the name of the Ceph cluster.
1941 .. option:: rbdname=str : [rbd]
1943 Specifies the name of the RBD.
1945 .. option:: pool=str : [rbd]
1947 Specifies the name of the Ceph pool containing RBD.
1949 .. option:: clientname=str : [rbd]
1951 Specifies the username (without the 'client.' prefix) used to access the
1952 Ceph cluster. If the *clustername* is specified, the *clientname* shall be
1953 the full *type.id* string. If no type. prefix is given, fio will add
1954 'client.' by default.
1956 .. option:: skip_bad=bool : [mtd]
1958 Skip operations against known bad blocks.
1960 .. option:: hdfsdirectory : [libhdfs]
1962 libhdfs will create chunk in this HDFS directory.
1964 .. option:: chunk_size : [libhdfs]
1966 The size of the chunk to use for each file.
1972 .. option:: iodepth=int
1974 Number of I/O units to keep in flight against the file. Note that
1975 increasing *iodepth* beyond 1 will not affect synchronous ioengines (except
1976 for small degrees when :option:`verify_async` is in use). Even async
1977 engines may impose OS restrictions causing the desired depth not to be
1978 achieved. This may happen on Linux when using libaio and not setting
1979 :option:`direct`\=1, since buffered I/O is not async on that OS. Keep an
1980 eye on the I/O depth distribution in the fio output to verify that the
1981 achieved depth is as expected. Default: 1.
1983 .. option:: iodepth_batch_submit=int, iodepth_batch=int
1985 This defines how many pieces of I/O to submit at once. It defaults to 1
1986 which means that we submit each I/O as soon as it is available, but can be
1987 raised to submit bigger batches of I/O at the time. If it is set to 0 the
1988 :option:`iodepth` value will be used.
1990 .. option:: iodepth_batch_complete_min=int, iodepth_batch_complete=int
1992 This defines how many pieces of I/O to retrieve at once. It defaults to 1
1993 which means that we'll ask for a minimum of 1 I/O in the retrieval process
1994 from the kernel. The I/O retrieval will go on until we hit the limit set by
1995 :option:`iodepth_low`. If this variable is set to 0, then fio will always
1996 check for completed events before queuing more I/O. This helps reduce I/O
1997 latency, at the cost of more retrieval system calls.
1999 .. option:: iodepth_batch_complete_max=int
2001 This defines maximum pieces of I/O to retrieve at once. This variable should
2002 be used along with :option:`iodepth_batch_complete_min`\=int variable,
2003 specifying the range of min and max amount of I/O which should be
2004 retrieved. By default it is equal to the :option:`iodepth_batch_complete_min`
2009 iodepth_batch_complete_min=1
2010 iodepth_batch_complete_max=<iodepth>
2012 which means that we will retrieve at least 1 I/O and up to the whole
2013 submitted queue depth. If none of I/O has been completed yet, we will wait.
2017 iodepth_batch_complete_min=0
2018 iodepth_batch_complete_max=<iodepth>
2020 which means that we can retrieve up to the whole submitted queue depth, but
2021 if none of I/O has been completed yet, we will NOT wait and immediately exit
2022 the system call. In this example we simply do polling.
2024 .. option:: iodepth_low=int
2026 The low water mark indicating when to start filling the queue
2027 again. Defaults to the same as :option:`iodepth`, meaning that fio will
2028 attempt to keep the queue full at all times. If :option:`iodepth` is set to
2029 e.g. 16 and *iodepth_low* is set to 4, then after fio has filled the queue of
2030 16 requests, it will let the depth drain down to 4 before starting to fill
2033 .. option:: serialize_overlap=bool
2035 Serialize in-flight I/Os that might otherwise cause or suffer from data races.
2036 When two or more I/Os are submitted simultaneously, there is no guarantee that
2037 the I/Os will be processed or completed in the submitted order. Further, if
2038 two or more of those I/Os are writes, any overlapping region between them can
2039 become indeterminate/undefined on certain storage. These issues can cause
2040 verification to fail erratically when at least one of the racing I/Os is
2041 changing data and the overlapping region has a non-zero size. Setting
2042 ``serialize_overlap`` tells fio to avoid provoking this behavior by explicitly
2043 serializing in-flight I/Os that have a non-zero overlap. Note that setting
2044 this option can reduce both performance and the `:option:iodepth` achieved.
2045 Additionally this option does not work when :option:`io_submit_mode` is set to
2046 offload. Default: false.
2048 .. option:: io_submit_mode=str
2050 This option controls how fio submits the I/O to the I/O engine. The default
2051 is `inline`, which means that the fio job threads submit and reap I/O
2052 directly. If set to `offload`, the job threads will offload I/O submission
2053 to a dedicated pool of I/O threads. This requires some coordination and thus
2054 has a bit of extra overhead, especially for lower queue depth I/O where it
2055 can increase latencies. The benefit is that fio can manage submission rates
2056 independently of the device completion rates. This avoids skewed latency
2057 reporting if I/O gets backed up on the device side (the coordinated omission
2064 .. option:: thinktime=time
2066 Stall the job for the specified period of time after an I/O has completed before issuing the
2067 next. May be used to simulate processing being done by an application.
2068 When the unit is omitted, the value is interpreted in microseconds. See
2069 :option:`thinktime_blocks` and :option:`thinktime_spin`.
2071 .. option:: thinktime_spin=time
2073 Only valid if :option:`thinktime` is set - pretend to spend CPU time doing
2074 something with the data received, before falling back to sleeping for the
2075 rest of the period specified by :option:`thinktime`. When the unit is
2076 omitted, the value is interpreted in microseconds.
2078 .. option:: thinktime_blocks=int
2080 Only valid if :option:`thinktime` is set - control how many blocks to issue,
2081 before waiting `thinktime` usecs. If not set, defaults to 1 which will make
2082 fio wait `thinktime` usecs after every block. This effectively makes any
2083 queue depth setting redundant, since no more than 1 I/O will be queued
2084 before we have to complete it and do our thinktime. In other words, this
2085 setting effectively caps the queue depth if the latter is larger.
2087 .. option:: rate=int[,int][,int]
2089 Cap the bandwidth used by this job. The number is in bytes/sec, the normal
2090 suffix rules apply. Comma-separated values may be specified for reads,
2091 writes, and trims as described in :option:`blocksize`.
2093 For example, using `rate=1m,500k` would limit reads to 1MiB/sec and writes to
2094 500KiB/sec. Capping only reads or writes can be done with `rate=,500k` or
2095 `rate=500k,` where the former will only limit writes (to 500KiB/sec) and the
2096 latter will only limit reads.
2098 .. option:: rate_min=int[,int][,int]
2100 Tell fio to do whatever it can to maintain at least this bandwidth. Failing
2101 to meet this requirement will cause the job to exit. Comma-separated values
2102 may be specified for reads, writes, and trims as described in
2103 :option:`blocksize`.
2105 .. option:: rate_iops=int[,int][,int]
2107 Cap the bandwidth to this number of IOPS. Basically the same as
2108 :option:`rate`, just specified independently of bandwidth. If the job is
2109 given a block size range instead of a fixed value, the smallest block size
2110 is used as the metric. Comma-separated values may be specified for reads,
2111 writes, and trims as described in :option:`blocksize`.
2113 .. option:: rate_iops_min=int[,int][,int]
2115 If fio doesn't meet this rate of I/O, it will cause the job to exit.
2116 Comma-separated values may be specified for reads, writes, and trims as
2117 described in :option:`blocksize`.
2119 .. option:: rate_process=str
2121 This option controls how fio manages rated I/O submissions. The default is
2122 `linear`, which submits I/O in a linear fashion with fixed delays between
2123 I/Os that gets adjusted based on I/O completion rates. If this is set to
2124 `poisson`, fio will submit I/O based on a more real world random request
2125 flow, known as the Poisson process
2126 (https://en.wikipedia.org/wiki/Poisson_point_process). The lambda will be
2127 10^6 / IOPS for the given workload.
2133 .. option:: latency_target=time
2135 If set, fio will attempt to find the max performance point that the given
2136 workload will run at while maintaining a latency below this target. When
2137 the unit is omitted, the value is interpreted in microseconds. See
2138 :option:`latency_window` and :option:`latency_percentile`.
2140 .. option:: latency_window=time
2142 Used with :option:`latency_target` to specify the sample window that the job
2143 is run at varying queue depths to test the performance. When the unit is
2144 omitted, the value is interpreted in microseconds.
2146 .. option:: latency_percentile=float
2148 The percentage of I/Os that must fall within the criteria specified by
2149 :option:`latency_target` and :option:`latency_window`. If not set, this
2150 defaults to 100.0, meaning that all I/Os must be equal or below to the value
2151 set by :option:`latency_target`.
2153 .. option:: max_latency=time
2155 If set, fio will exit the job with an ETIMEDOUT error if it exceeds this
2156 maximum latency. When the unit is omitted, the value is interpreted in
2159 .. option:: rate_cycle=int
2161 Average bandwidth for :option:`rate` and :option:`rate_min` over this number
2162 of milliseconds. Defaults to 1000.
2168 .. option:: write_iolog=str
2170 Write the issued I/O patterns to the specified file. See
2171 :option:`read_iolog`. Specify a separate file for each job, otherwise the
2172 iologs will be interspersed and the file may be corrupt.
2174 .. option:: read_iolog=str
2176 Open an iolog with the specified filename and replay the I/O patterns it
2177 contains. This can be used to store a workload and replay it sometime
2178 later. The iolog given may also be a blktrace binary file, which allows fio
2179 to replay a workload captured by :command:`blktrace`. See
2180 :manpage:`blktrace(8)` for how to capture such logging data. For blktrace
2181 replay, the file needs to be turned into a blkparse binary data file first
2182 (``blkparse <device> -o /dev/null -d file_for_fio.bin``).
2184 .. option:: replay_no_stall=bool
2186 When replaying I/O with :option:`read_iolog` the default behavior is to
2187 attempt to respect the timestamps within the log and replay them with the
2188 appropriate delay between IOPS. By setting this variable fio will not
2189 respect the timestamps and attempt to replay them as fast as possible while
2190 still respecting ordering. The result is the same I/O pattern to a given
2191 device, but different timings.
2193 .. option:: replay_redirect=str
2195 While replaying I/O patterns using :option:`read_iolog` the default behavior
2196 is to replay the IOPS onto the major/minor device that each IOP was recorded
2197 from. This is sometimes undesirable because on a different machine those
2198 major/minor numbers can map to a different device. Changing hardware on the
2199 same system can also result in a different major/minor mapping.
2200 ``replay_redirect`` causes all I/Os to be replayed onto the single specified
2201 device regardless of the device it was recorded
2202 from. i.e. :option:`replay_redirect`\= :file:`/dev/sdc` would cause all I/O
2203 in the blktrace or iolog to be replayed onto :file:`/dev/sdc`. This means
2204 multiple devices will be replayed onto a single device, if the trace
2205 contains multiple devices. If you want multiple devices to be replayed
2206 concurrently to multiple redirected devices you must blkparse your trace
2207 into separate traces and replay them with independent fio invocations.
2208 Unfortunately this also breaks the strict time ordering between multiple
2211 .. option:: replay_align=int
2213 Force alignment of I/O offsets and lengths in a trace to this power of 2
2216 .. option:: replay_scale=int
2218 Scale sector offsets down by this factor when replaying traces.
2221 Threads, processes and job synchronization
2222 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2226 Fio defaults to creating jobs by using fork, however if this option is
2227 given, fio will create jobs by using POSIX Threads' function
2228 :manpage:`pthread_create(3)` to create threads instead.
2230 .. option:: wait_for=str
2232 If set, the current job won't be started until all workers of the specified
2233 waitee job are done.
2235 ``wait_for`` operates on the job name basis, so there are a few
2236 limitations. First, the waitee must be defined prior to the waiter job
2237 (meaning no forward references). Second, if a job is being referenced as a
2238 waitee, it must have a unique name (no duplicate waitees).
2240 .. option:: nice=int
2242 Run the job with the given nice value. See man :manpage:`nice(2)`.
2244 On Windows, values less than -15 set the process class to "High"; -1 through
2245 -15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle"
2248 .. option:: prio=int
2250 Set the I/O priority value of this job. Linux limits us to a positive value
2251 between 0 and 7, with 0 being the highest. See man
2252 :manpage:`ionice(1)`. Refer to an appropriate manpage for other operating
2253 systems since meaning of priority may differ.
2255 .. option:: prioclass=int
2257 Set the I/O priority class. See man :manpage:`ionice(1)`.
2259 .. option:: cpumask=int
2261 Set the CPU affinity of this job. The parameter given is a bit mask of
2262 allowed CPUs the job may run on. So if you want the allowed CPUs to be 1
2263 and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man
2264 :manpage:`sched_setaffinity(2)`. This may not work on all supported
2265 operating systems or kernel versions. This option doesn't work well for a
2266 higher CPU count than what you can store in an integer mask, so it can only
2267 control cpus 1-32. For boxes with larger CPU counts, use
2268 :option:`cpus_allowed`.
2270 .. option:: cpus_allowed=str
2272 Controls the same options as :option:`cpumask`, but accepts a textual
2273 specification of the permitted CPUs instead. So to use CPUs 1 and 5 you
2274 would specify ``cpus_allowed=1,5``. This option also allows a range of CPUs
2275 to be specified -- say you wanted a binding to CPUs 1, 5, and 8 to 15, you
2276 would set ``cpus_allowed=1,5,8-15``.
2278 .. option:: cpus_allowed_policy=str
2280 Set the policy of how fio distributes the CPUs specified by
2281 :option:`cpus_allowed` or :option:`cpumask`. Two policies are supported:
2284 All jobs will share the CPU set specified.
2286 Each job will get a unique CPU from the CPU set.
2288 **shared** is the default behavior, if the option isn't specified. If
2289 **split** is specified, then fio will will assign one cpu per job. If not
2290 enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs
2293 .. option:: numa_cpu_nodes=str
2295 Set this job running on specified NUMA nodes' CPUs. The arguments allow
2296 comma delimited list of cpu numbers, A-B ranges, or `all`. Note, to enable
2297 NUMA options support, fio must be built on a system with libnuma-dev(el)
2300 .. option:: numa_mem_policy=str
2302 Set this job's memory policy and corresponding NUMA nodes. Format of the
2307 ``mode`` is one of the following memory poicies: ``default``, ``prefer``,
2308 ``bind``, ``interleave`` or ``local``. For ``default`` and ``local`` memory
2309 policies, no node needs to be specified. For ``prefer``, only one node is
2310 allowed. For ``bind`` and ``interleave`` the ``nodelist`` may be as
2311 follows: a comma delimited list of numbers, A-B ranges, or `all`.
2313 .. option:: cgroup=str
2315 Add job to this control group. If it doesn't exist, it will be created. The
2316 system must have a mounted cgroup blkio mount point for this to work. If
2317 your system doesn't have it mounted, you can do so with::
2319 # mount -t cgroup -o blkio none /cgroup
2321 .. option:: cgroup_weight=int
2323 Set the weight of the cgroup to this value. See the documentation that comes
2324 with the kernel, allowed values are in the range of 100..1000.
2326 .. option:: cgroup_nodelete=bool
2328 Normally fio will delete the cgroups it has created after the job
2329 completion. To override this behavior and to leave cgroups around after the
2330 job completion, set ``cgroup_nodelete=1``. This can be useful if one wants
2331 to inspect various cgroup files after job completion. Default: false.
2333 .. option:: flow_id=int
2335 The ID of the flow. If not specified, it defaults to being a global
2336 flow. See :option:`flow`.
2338 .. option:: flow=int
2340 Weight in token-based flow control. If this value is used, then there is a
2341 'flow counter' which is used to regulate the proportion of activity between
2342 two or more jobs. Fio attempts to keep this flow counter near zero. The
2343 ``flow`` parameter stands for how much should be added or subtracted to the
2344 flow counter on each iteration of the main I/O loop. That is, if one job has
2345 ``flow=8`` and another job has ``flow=-1``, then there will be a roughly 1:8
2346 ratio in how much one runs vs the other.
2348 .. option:: flow_watermark=int
2350 The maximum value that the absolute value of the flow counter is allowed to
2351 reach before the job must wait for a lower value of the counter.
2353 .. option:: flow_sleep=int
2355 The period of time, in microseconds, to wait after the flow watermark has
2356 been exceeded before retrying operations.
2358 .. option:: stonewall, wait_for_previous
2360 Wait for preceding jobs in the job file to exit, before starting this
2361 one. Can be used to insert serialization points in the job file. A stone
2362 wall also implies starting a new reporting group, see
2363 :option:`group_reporting`.
2367 By default, fio will continue running all other jobs when one job finishes
2368 but sometimes this is not the desired action. Setting ``exitall`` will
2369 instead make fio terminate all other jobs when one job finishes.
2371 .. option:: exec_prerun=str
2373 Before running this job, issue the command specified through
2374 :manpage:`system(3)`. Output is redirected in a file called
2375 :file:`jobname.prerun.txt`.
2377 .. option:: exec_postrun=str
2379 After the job completes, issue the command specified though
2380 :manpage:`system(3)`. Output is redirected in a file called
2381 :file:`jobname.postrun.txt`.
2385 Instead of running as the invoking user, set the user ID to this value
2386 before the thread/process does any work.
2390 Set group ID, see :option:`uid`.
2396 .. option:: verify_only
2398 Do not perform specified workload, only verify data still matches previous
2399 invocation of this workload. This option allows one to check data multiple
2400 times at a later date without overwriting it. This option makes sense only
2401 for workloads that write data, and does not support workloads with the
2402 :option:`time_based` option set.
2404 .. option:: do_verify=bool
2406 Run the verify phase after a write phase. Only valid if :option:`verify` is
2409 .. option:: verify=str
2411 If writing to a file, fio can verify the file contents after each iteration
2412 of the job. Each verification method also implies verification of special
2413 header, which is written to the beginning of each block. This header also
2414 includes meta information, like offset of the block, block number, timestamp
2415 when block was written, etc. :option:`verify` can be combined with
2416 :option:`verify_pattern` option. The allowed values are:
2419 Use an md5 sum of the data area and store it in the header of
2423 Use an experimental crc64 sum of the data area and store it in the
2424 header of each block.
2427 Use a crc32c sum of the data area and store it in the header of
2428 each block. This will automatically use hardware acceleration
2429 (e.g. SSE4.2 on an x86 or CRC crypto extensions on ARM64) but will
2430 fall back to software crc32c if none is found. Generally the
2431 fatest checksum fio supports when hardware accelerated.
2437 Use a crc32 sum of the data area and store it in the header of each
2441 Use a crc16 sum of the data area and store it in the header of each
2445 Use a crc7 sum of the data area and store it in the header of each
2449 Use xxhash as the checksum function. Generally the fastest software
2450 checksum that fio supports.
2453 Use sha512 as the checksum function.
2456 Use sha256 as the checksum function.
2459 Use optimized sha1 as the checksum function.
2462 Use optimized sha3-224 as the checksum function.
2465 Use optimized sha3-256 as the checksum function.
2468 Use optimized sha3-384 as the checksum function.
2471 Use optimized sha3-512 as the checksum function.
2474 This option is deprecated, since now meta information is included in
2475 generic verification header and meta verification happens by
2476 default. For detailed information see the description of the
2477 :option:`verify` setting. This option is kept because of
2478 compatibility's sake with old configurations. Do not use it.
2481 Verify a strict pattern. Normally fio includes a header with some
2482 basic information and checksumming, but if this option is set, only
2483 the specific pattern set with :option:`verify_pattern` is verified.
2486 Only pretend to verify. Useful for testing internals with
2487 :option:`ioengine`\=null, not for much else.
2489 This option can be used for repeated burn-in tests of a system to make sure
2490 that the written data is also correctly read back. If the data direction
2491 given is a read or random read, fio will assume that it should verify a
2492 previously written file. If the data direction includes any form of write,
2493 the verify will be of the newly written data.
2495 .. option:: verifysort=bool
2497 If true, fio will sort written verify blocks when it deems it faster to read
2498 them back in a sorted manner. This is often the case when overwriting an
2499 existing file, since the blocks are already laid out in the file system. You
2500 can ignore this option unless doing huge amounts of really fast I/O where
2501 the red-black tree sorting CPU time becomes significant. Default: true.
2503 .. option:: verifysort_nr=int
2505 Pre-load and sort verify blocks for a read workload.
2507 .. option:: verify_offset=int
2509 Swap the verification header with data somewhere else in the block before
2510 writing. It is swapped back before verifying.
2512 .. option:: verify_interval=int
2514 Write the verification header at a finer granularity than the
2515 :option:`blocksize`. It will be written for chunks the size of
2516 ``verify_interval``. :option:`blocksize` should divide this evenly.
2518 .. option:: verify_pattern=str
2520 If set, fio will fill the I/O buffers with this pattern. Fio defaults to
2521 filling with totally random bytes, but sometimes it's interesting to fill
2522 with a known pattern for I/O verification purposes. Depending on the width
2523 of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time (it can
2524 be either a decimal or a hex number). The ``verify_pattern`` if larger than
2525 a 32-bit quantity has to be a hex number that starts with either "0x" or
2526 "0X". Use with :option:`verify`. Also, ``verify_pattern`` supports %o
2527 format, which means that for each block offset will be written and then
2528 verified back, e.g.::
2532 Or use combination of everything::
2534 verify_pattern=0xff%o"abcd"-12
2536 .. option:: verify_fatal=bool
2538 Normally fio will keep checking the entire contents before quitting on a
2539 block verification failure. If this option is set, fio will exit the job on
2540 the first observed failure. Default: false.
2542 .. option:: verify_dump=bool
2544 If set, dump the contents of both the original data block and the data block
2545 we read off disk to files. This allows later analysis to inspect just what
2546 kind of data corruption occurred. Off by default.
2548 .. option:: verify_async=int
2550 Fio will normally verify I/O inline from the submitting thread. This option
2551 takes an integer describing how many async offload threads to create for I/O
2552 verification instead, causing fio to offload the duty of verifying I/O
2553 contents to one or more separate threads. If using this offload option, even
2554 sync I/O engines can benefit from using an :option:`iodepth` setting higher
2555 than 1, as it allows them to have I/O in flight while verifies are running.
2556 Defaults to 0 async threads, i.e. verification is not asynchronous.
2558 .. option:: verify_async_cpus=str
2560 Tell fio to set the given CPU affinity on the async I/O verification
2561 threads. See :option:`cpus_allowed` for the format used.
2563 .. option:: verify_backlog=int
2565 Fio will normally verify the written contents of a job that utilizes verify
2566 once that job has completed. In other words, everything is written then
2567 everything is read back and verified. You may want to verify continually
2568 instead for a variety of reasons. Fio stores the meta data associated with
2569 an I/O block in memory, so for large verify workloads, quite a bit of memory
2570 would be used up holding this meta data. If this option is enabled, fio will
2571 write only N blocks before verifying these blocks.
2573 .. option:: verify_backlog_batch=int
2575 Control how many blocks fio will verify if :option:`verify_backlog` is
2576 set. If not set, will default to the value of :option:`verify_backlog`
2577 (meaning the entire queue is read back and verified). If
2578 ``verify_backlog_batch`` is less than :option:`verify_backlog` then not all
2579 blocks will be verified, if ``verify_backlog_batch`` is larger than
2580 :option:`verify_backlog`, some blocks will be verified more than once.
2582 .. option:: verify_state_save=bool
2584 When a job exits during the write phase of a verify workload, save its
2585 current state. This allows fio to replay up until that point, if the verify
2586 state is loaded for the verify read phase. The format of the filename is,
2589 <type>-<jobname>-<jobindex>-verify.state.
2591 <type> is "local" for a local run, "sock" for a client/server socket
2592 connection, and "ip" (192.168.0.1, for instance) for a networked
2593 client/server connection. Defaults to true.
2595 .. option:: verify_state_load=bool
2597 If a verify termination trigger was used, fio stores the current write state
2598 of each thread. This can be used at verification time so that fio knows how
2599 far it should verify. Without this information, fio will run a full
2600 verification pass, according to the settings in the job file used. Default
2603 .. option:: trim_percentage=int
2605 Number of verify blocks to discard/trim.
2607 .. option:: trim_verify_zero=bool
2609 Verify that trim/discarded blocks are returned as zeros.
2611 .. option:: trim_backlog=int
2613 Trim after this number of blocks are written.
2615 .. option:: trim_backlog_batch=int
2617 Trim this number of I/O blocks.
2619 .. option:: experimental_verify=bool
2621 Enable experimental verification.
2626 .. option:: steadystate=str:float, ss=str:float
2628 Define the criterion and limit for assessing steady state performance. The
2629 first parameter designates the criterion whereas the second parameter sets
2630 the threshold. When the criterion falls below the threshold for the
2631 specified duration, the job will stop. For example, `iops_slope:0.1%` will
2632 direct fio to terminate the job when the least squares regression slope
2633 falls below 0.1% of the mean IOPS. If :option:`group_reporting` is enabled
2634 this will apply to all jobs in the group. Below is the list of available
2635 steady state assessment criteria. All assessments are carried out using only
2636 data from the rolling collection window. Threshold limits can be expressed
2637 as a fixed value or as a percentage of the mean in the collection window.
2640 Collect IOPS data. Stop the job if all individual IOPS measurements
2641 are within the specified limit of the mean IOPS (e.g., ``iops:2``
2642 means that all individual IOPS values must be within 2 of the mean,
2643 whereas ``iops:0.2%`` means that all individual IOPS values must be
2644 within 0.2% of the mean IOPS to terminate the job).
2647 Collect IOPS data and calculate the least squares regression
2648 slope. Stop the job if the slope falls below the specified limit.
2651 Collect bandwidth data. Stop the job if all individual bandwidth
2652 measurements are within the specified limit of the mean bandwidth.
2655 Collect bandwidth data and calculate the least squares regression
2656 slope. Stop the job if the slope falls below the specified limit.
2658 .. option:: steadystate_duration=time, ss_dur=time
2660 A rolling window of this duration will be used to judge whether steady state
2661 has been reached. Data will be collected once per second. The default is 0
2662 which disables steady state detection. When the unit is omitted, the
2663 value is interpreted in seconds.
2665 .. option:: steadystate_ramp_time=time, ss_ramp=time
2667 Allow the job to run for the specified duration before beginning data
2668 collection for checking the steady state job termination criterion. The
2669 default is 0. When the unit is omitted, the value is interpreted in seconds.
2672 Measurements and reporting
2673 ~~~~~~~~~~~~~~~~~~~~~~~~~~
2675 .. option:: per_job_logs=bool
2677 If set, this generates bw/clat/iops log with per file private filenames. If
2678 not set, jobs with identical names will share the log filename. Default:
2681 .. option:: group_reporting
2683 It may sometimes be interesting to display statistics for groups of jobs as
2684 a whole instead of for each individual job. This is especially true if
2685 :option:`numjobs` is used; looking at individual thread/process output
2686 quickly becomes unwieldy. To see the final report per-group instead of
2687 per-job, use :option:`group_reporting`. Jobs in a file will be part of the
2688 same reporting group, unless if separated by a :option:`stonewall`, or by
2689 using :option:`new_group`.
2691 .. option:: new_group
2693 Start a new reporting group. See: :option:`group_reporting`. If not given,
2694 all jobs in a file will be part of the same reporting group, unless
2695 separated by a :option:`stonewall`.
2697 .. option:: stats=bool
2699 By default, fio collects and shows final output results for all jobs
2700 that run. If this option is set to 0, then fio will ignore it in
2701 the final stat output.
2703 .. option:: write_bw_log=str
2705 If given, write a bandwidth log for this job. Can be used to store data of
2706 the bandwidth of the jobs in their lifetime. The included
2707 :command:`fio_generate_plots` script uses :command:`gnuplot` to turn these
2708 text files into nice graphs. See :option:`write_lat_log` for behavior of
2709 given filename. For this option, the postfix is :file:`_bw.x.log`, where `x`
2710 is the index of the job (`1..N`, where `N` is the number of jobs). If
2711 :option:`per_job_logs` is false, then the filename will not include the job
2712 index. See `Log File Formats`_.
2714 .. option:: write_lat_log=str
2716 Same as :option:`write_bw_log`, except that this option stores I/O
2717 submission, completion, and total latencies instead. If no filename is given
2718 with this option, the default filename of :file:`jobname_type.log` is
2719 used. Even if the filename is given, fio will still append the type of
2720 log. So if one specifies::
2724 The actual log names will be :file:`foo_slat.x.log`, :file:`foo_clat.x.log`,
2725 and :file:`foo_lat.x.log`, where `x` is the index of the job (1..N, where N
2726 is the number of jobs). This helps :command:`fio_generate_plot` find the
2727 logs automatically. If :option:`per_job_logs` is false, then the filename
2728 will not include the job index. See `Log File Formats`_.
2730 .. option:: write_hist_log=str
2732 Same as :option:`write_lat_log`, but writes I/O completion latency
2733 histograms. If no filename is given with this option, the default filename
2734 of :file:`jobname_clat_hist.x.log` is used, where `x` is the index of the
2735 job (1..N, where `N` is the number of jobs). Even if the filename is given,
2736 fio will still append the type of log. If :option:`per_job_logs` is false,
2737 then the filename will not include the job index. See `Log File Formats`_.
2739 .. option:: write_iops_log=str
2741 Same as :option:`write_bw_log`, but writes IOPS. If no filename is given
2742 with this option, the default filename of :file:`jobname_type.x.log` is
2743 used,where `x` is the index of the job (1..N, where `N` is the number of
2744 jobs). Even if the filename is given, fio will still append the type of
2745 log. If :option:`per_job_logs` is false, then the filename will not include
2746 the job index. See `Log File Formats`_.
2748 .. option:: log_avg_msec=int
2750 By default, fio will log an entry in the iops, latency, or bw log for every
2751 I/O that completes. When writing to the disk log, that can quickly grow to a
2752 very large size. Setting this option makes fio average the each log entry
2753 over the specified period of time, reducing the resolution of the log. See
2754 :option:`log_max_value` as well. Defaults to 0, logging all entries.
2755 Also see `Log File Formats`_.
2757 .. option:: log_hist_msec=int
2759 Same as :option:`log_avg_msec`, but logs entries for completion latency
2760 histograms. Computing latency percentiles from averages of intervals using
2761 :option:`log_avg_msec` is inaccurate. Setting this option makes fio log
2762 histogram entries over the specified period of time, reducing log sizes for
2763 high IOPS devices while retaining percentile accuracy. See
2764 :option:`log_hist_coarseness` as well. Defaults to 0, meaning histogram
2765 logging is disabled.
2767 .. option:: log_hist_coarseness=int
2769 Integer ranging from 0 to 6, defining the coarseness of the resolution of
2770 the histogram logs enabled with :option:`log_hist_msec`. For each increment
2771 in coarseness, fio outputs half as many bins. Defaults to 0, for which
2772 histogram logs contain 1216 latency bins. See `Log File Formats`_.
2774 .. option:: log_max_value=bool
2776 If :option:`log_avg_msec` is set, fio logs the average over that window. If
2777 you instead want to log the maximum value, set this option to 1. Defaults to
2778 0, meaning that averaged values are logged.
2780 .. option:: log_offset=bool
2782 If this is set, the iolog options will include the byte offset for the I/O
2783 entry as well as the other data values. Defaults to 0 meaning that
2784 offsets are not present in logs. Also see `Log File Formats`_.
2786 .. option:: log_compression=int
2788 If this is set, fio will compress the I/O logs as it goes, to keep the
2789 memory footprint lower. When a log reaches the specified size, that chunk is
2790 removed and compressed in the background. Given that I/O logs are fairly
2791 highly compressible, this yields a nice memory savings for longer runs. The
2792 downside is that the compression will consume some background CPU cycles, so
2793 it may impact the run. This, however, is also true if the logging ends up
2794 consuming most of the system memory. So pick your poison. The I/O logs are
2795 saved normally at the end of a run, by decompressing the chunks and storing
2796 them in the specified log file. This feature depends on the availability of
2799 .. option:: log_compression_cpus=str
2801 Define the set of CPUs that are allowed to handle online log compression for
2802 the I/O jobs. This can provide better isolation between performance
2803 sensitive jobs, and background compression work.
2805 .. option:: log_store_compressed=bool
2807 If set, fio will store the log files in a compressed format. They can be
2808 decompressed with fio, using the :option:`--inflate-log` command line
2809 parameter. The files will be stored with a :file:`.fz` suffix.
2811 .. option:: log_unix_epoch=bool
2813 If set, fio will log Unix timestamps to the log files produced by enabling
2814 write_type_log for each log type, instead of the default zero-based
2817 .. option:: block_error_percentiles=bool
2819 If set, record errors in trim block-sized units from writes and trims and
2820 output a histogram of how many trims it took to get to errors, and what kind
2821 of error was encountered.
2823 .. option:: bwavgtime=int
2825 Average the calculated bandwidth over the given time. Value is specified in
2826 milliseconds. If the job also does bandwidth logging through
2827 :option:`write_bw_log`, then the minimum of this option and
2828 :option:`log_avg_msec` will be used. Default: 500ms.
2830 .. option:: iopsavgtime=int
2832 Average the calculated IOPS over the given time. Value is specified in
2833 milliseconds. If the job also does IOPS logging through
2834 :option:`write_iops_log`, then the minimum of this option and
2835 :option:`log_avg_msec` will be used. Default: 500ms.
2837 .. option:: disk_util=bool
2839 Generate disk utilization statistics, if the platform supports it.
2842 .. option:: disable_lat=bool
2844 Disable measurements of total latency numbers. Useful only for cutting back
2845 the number of calls to :manpage:`gettimeofday(2)`, as that does impact
2846 performance at really high IOPS rates. Note that to really get rid of a
2847 large amount of these calls, this option must be used with
2848 :option:`disable_slat` and :option:`disable_bw_measurement` as well.
2850 .. option:: disable_clat=bool
2852 Disable measurements of completion latency numbers. See
2853 :option:`disable_lat`.
2855 .. option:: disable_slat=bool
2857 Disable measurements of submission latency numbers. See
2858 :option:`disable_slat`.
2860 .. option:: disable_bw_measurement=bool, disable_bw=bool
2862 Disable measurements of throughput/bandwidth numbers. See
2863 :option:`disable_lat`.
2865 .. option:: clat_percentiles=bool
2867 Enable the reporting of percentiles of completion latencies.
2869 .. option:: percentile_list=float_list
2871 Overwrite the default list of percentiles for completion latencies and the
2872 block error histogram. Each number is a floating number in the range
2873 (0,100], and the maximum length of the list is 20. Use ``:`` to separate the
2874 numbers, and list the numbers in ascending order. For example,
2875 ``--percentile_list=99.5:99.9`` will cause fio to report the values of
2876 completion latency below which 99.5% and 99.9% of the observed latencies
2883 .. option:: exitall_on_error
2885 When one job finishes in error, terminate the rest. The default is to wait
2886 for each job to finish.
2888 .. option:: continue_on_error=str
2890 Normally fio will exit the job on the first observed failure. If this option
2891 is set, fio will continue the job when there is a 'non-fatal error' (EIO or
2892 EILSEQ) until the runtime is exceeded or the I/O size specified is
2893 completed. If this option is used, there are two more stats that are
2894 appended, the total error count and the first error. The error field given
2895 in the stats is the first error that was hit during the run.
2897 The allowed values are:
2900 Exit on any I/O or verify errors.
2903 Continue on read errors, exit on all others.
2906 Continue on write errors, exit on all others.
2909 Continue on any I/O error, exit on all others.
2912 Continue on verify errors, exit on all others.
2915 Continue on all errors.
2918 Backward-compatible alias for 'none'.
2921 Backward-compatible alias for 'all'.
2923 .. option:: ignore_error=str
2925 Sometimes you want to ignore some errors during test in that case you can
2926 specify error list for each error type, instead of only being able to
2927 ignore the default 'non-fatal error' using :option:`continue_on_error`.
2928 ``ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST`` errors for
2929 given error type is separated with ':'. Error may be symbol ('ENOSPC',
2930 'ENOMEM') or integer. Example::
2932 ignore_error=EAGAIN,ENOSPC:122
2934 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from
2935 WRITE. This option works by overriding :option:`continue_on_error` with
2936 the list of errors for each error type if any.
2938 .. option:: error_dump=bool
2940 If set dump every error even if it is non fatal, true by default. If
2941 disabled only fatal error will be dumped.
2943 Running predefined workloads
2944 ----------------------------
2946 Fio includes predefined profiles that mimic the I/O workloads generated by
2949 .. option:: profile=str
2951 The predefined workload to run. Current profiles are:
2954 Threaded I/O bench (tiotest/tiobench) like workload.
2957 Aerospike Certification Tool (ACT) like workload.
2959 To view a profile's additional options use :option:`--cmdhelp` after specifying
2960 the profile. For example::
2962 $ fio --profile=act --cmdhelp
2967 .. option:: device-names=str
2972 .. option:: load=int
2975 ACT load multiplier. Default: 1.
2977 .. option:: test-duration=time
2980 How long the entire test takes to run. When the unit is omitted, the value
2981 is given in seconds. Default: 24h.
2983 .. option:: threads-per-queue=int
2986 Number of read IO threads per device. Default: 8.
2988 .. option:: read-req-num-512-blocks=int
2991 Number of 512B blocks to read at the time. Default: 3.
2993 .. option:: large-block-op-kbytes=int
2996 Size of large block ops in KiB (writes). Default: 131072.
3001 Set to run ACT prep phase.
3003 Tiobench profile options
3004 ~~~~~~~~~~~~~~~~~~~~~~~~
3006 .. option:: size=str
3011 .. option:: block=int
3014 Block size in bytes. Default: 4096.
3016 .. option:: numruns=int
3026 .. option:: threads=int
3031 Interpreting the output
3032 -----------------------
3035 Example output was based on the following:
3036 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --time_based \
3037 --rate=1256k --bs=14K --name=quick --runtime=1s --name=mixed \
3038 --runtime=2m --rw=rw
3040 Fio spits out a lot of output. While running, fio will display the status of the
3041 jobs created. An example of that would be::
3043 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]
3045 The characters inside the first set of square brackets denote the current status of
3046 each thread. The first character is the first job defined in the job file, and so
3047 forth. The possible values (in typical life cycle order) are:
3049 +------+-----+-----------------------------------------------------------+
3051 +======+=====+===========================================================+
3052 | P | | Thread setup, but not started. |
3053 +------+-----+-----------------------------------------------------------+
3054 | C | | Thread created. |
3055 +------+-----+-----------------------------------------------------------+
3056 | I | | Thread initialized, waiting or generating necessary data. |
3057 +------+-----+-----------------------------------------------------------+
3058 | | p | Thread running pre-reading file(s). |
3059 +------+-----+-----------------------------------------------------------+
3060 | | / | Thread is in ramp period. |
3061 +------+-----+-----------------------------------------------------------+
3062 | | R | Running, doing sequential reads. |
3063 +------+-----+-----------------------------------------------------------+
3064 | | r | Running, doing random reads. |
3065 +------+-----+-----------------------------------------------------------+
3066 | | W | Running, doing sequential writes. |
3067 +------+-----+-----------------------------------------------------------+
3068 | | w | Running, doing random writes. |
3069 +------+-----+-----------------------------------------------------------+
3070 | | M | Running, doing mixed sequential reads/writes. |
3071 +------+-----+-----------------------------------------------------------+
3072 | | m | Running, doing mixed random reads/writes. |
3073 +------+-----+-----------------------------------------------------------+
3074 | | D | Running, doing sequential trims. |
3075 +------+-----+-----------------------------------------------------------+
3076 | | d | Running, doing random trims. |
3077 +------+-----+-----------------------------------------------------------+
3078 | | F | Running, currently waiting for :manpage:`fsync(2)`. |
3079 +------+-----+-----------------------------------------------------------+
3080 | | V | Running, doing verification of written data. |
3081 +------+-----+-----------------------------------------------------------+
3082 | f | | Thread finishing. |
3083 +------+-----+-----------------------------------------------------------+
3084 | E | | Thread exited, not reaped by main thread yet. |
3085 +------+-----+-----------------------------------------------------------+
3086 | _ | | Thread reaped. |
3087 +------+-----+-----------------------------------------------------------+
3088 | X | | Thread reaped, exited with an error. |
3089 +------+-----+-----------------------------------------------------------+
3090 | K | | Thread reaped, exited due to signal. |
3091 +------+-----+-----------------------------------------------------------+
3094 Example output was based on the following:
3095 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --runtime=58m \
3096 --time_based --rate=2512k --bs=256K --numjobs=10 \
3097 --name=readers --rw=read --name=writers --rw=write
3099 Fio will condense the thread string as not to take up more space on the command
3100 line than needed. For instance, if you have 10 readers and 10 writers running,
3101 the output would look like this::
3103 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]
3105 Note that the status string is displayed in order, so it's possible to tell which of
3106 the jobs are currently doing what. In the example above this means that jobs 1--10
3107 are readers and 11--20 are writers.
3109 The other values are fairly self explanatory -- number of threads currently
3110 running and doing I/O, the number of currently open files (f=), the estimated
3111 completion percentage, the rate of I/O since last check (read speed listed first,
3112 then write speed and optionally trim speed) in terms of bandwidth and IOPS, and time to completion for the current
3113 running group. It's impossible to estimate runtime of the following groups (if
3117 Example output was based on the following:
3118 TZ=UTC fio --iodepth=16 --ioengine=posixaio --filename=/tmp/fiofile \
3119 --direct=1 --size=100M --time_based --runtime=50s --rate_iops=89 \
3120 --bs=7K --name=Client1 --rw=write
3122 When fio is done (or interrupted by :kbd:`Ctrl-C`), it will show the data for
3123 each thread, group of threads, and disks in that order. For each overall thread (or
3124 group) the output looks like::
3126 Client1: (groupid=0, jobs=1): err= 0: pid=16109: Sat Jun 24 12:07:54 2017
3127 write: IOPS=88, BW=623KiB/s (638kB/s)(30.4MiB/50032msec)
3128 slat (nsec): min=500, max=145500, avg=8318.00, stdev=4781.50
3129 clat (usec): min=170, max=78367, avg=4019.02, stdev=8293.31
3130 lat (usec): min=174, max=78375, avg=4027.34, stdev=8291.79
3131 clat percentiles (usec):
3132 | 1.00th=[ 302], 5.00th=[ 326], 10.00th=[ 343], 20.00th=[ 363],
3133 | 30.00th=[ 392], 40.00th=[ 404], 50.00th=[ 416], 60.00th=[ 445],
3134 | 70.00th=[ 816], 80.00th=[ 6718], 90.00th=[12911], 95.00th=[21627],
3135 | 99.00th=[43779], 99.50th=[51643], 99.90th=[68682], 99.95th=[72877],
3137 bw ( KiB/s): min= 532, max= 686, per=0.10%, avg=622.87, stdev=24.82, samples= 100
3138 iops : min= 76, max= 98, avg=88.98, stdev= 3.54, samples= 100
3139 lat (usec) : 250=0.04%, 500=64.11%, 750=4.81%, 1000=2.79%
3140 lat (msec) : 2=4.16%, 4=1.84%, 10=4.90%, 20=11.33%, 50=5.37%
3141 lat (msec) : 100=0.65%
3142 cpu : usr=0.27%, sys=0.18%, ctx=12072, majf=0, minf=21
3143 IO depths : 1=85.0%, 2=13.1%, 4=1.8%, 8=0.1%, 16=0.0%, 32=0.0%, >=64=0.0%
3144 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
3145 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
3146 issued rwt: total=0,4450,0, short=0,0,0, dropped=0,0,0
3147 latency : target=0, window=0, percentile=100.00%, depth=8
3149 The job name (or first job's name when using :option:`group_reporting`) is printed,
3150 along with the group id, count of jobs being aggregated, last error id seen (which
3151 is 0 when there are no errors), pid/tid of that thread and the time the job/group
3152 completed. Below are the I/O statistics for each data direction performed (showing
3153 writes in the example above). In the order listed, they denote:
3156 The string before the colon shows the I/O direction the statistics
3157 are for. **IOPS** is the average I/Os performed per second. **BW**
3158 is the average bandwidth rate shown as: value in power of 2 format
3159 (value in power of 10 format). The last two values show: (**total
3160 I/O performed** in power of 2 format / **runtime** of that thread).
3163 Submission latency (**min** being the minimum, **max** being the
3164 maximum, **avg** being the average, **stdev** being the standard
3165 deviation). This is the time it took to submit the I/O. For
3166 sync I/O this row is not displayed as the slat is really the
3167 completion latency (since queue/complete is one operation there).
3168 This value can be in nanoseconds, microseconds or milliseconds ---
3169 fio will choose the most appropriate base and print that (in the
3170 example above nanoseconds was the best scale). Note: in :option:`--minimal` mode
3171 latencies are always expressed in microseconds.
3174 Completion latency. Same names as slat, this denotes the time from
3175 submission to completion of the I/O pieces. For sync I/O, clat will
3176 usually be equal (or very close) to 0, as the time from submit to
3177 complete is basically just CPU time (I/O has already been done, see slat
3181 Bandwidth statistics based on samples. Same names as the xlat stats,
3182 but also includes the number of samples taken (**samples**) and an
3183 approximate percentage of total aggregate bandwidth this thread
3184 received in its group (**per**). This last value is only really
3185 useful if the threads in this group are on the same disk, since they
3186 are then competing for disk access.
3189 IOPS statistics based on samples. Same names as bw.
3192 CPU usage. User and system time, along with the number of context
3193 switches this thread went through, usage of system and user time, and
3194 finally the number of major and minor page faults. The CPU utilization
3195 numbers are averages for the jobs in that reporting group, while the
3196 context and fault counters are summed.
3199 The distribution of I/O depths over the job lifetime. The numbers are
3200 divided into powers of 2 and each entry covers depths from that value
3201 up to those that are lower than the next entry -- e.g., 16= covers
3202 depths from 16 to 31. Note that the range covered by a depth
3203 distribution entry can be different to the range covered by the
3204 equivalent submit/complete distribution entry.
3207 How many pieces of I/O were submitting in a single submit call. Each
3208 entry denotes that amount and below, until the previous entry -- e.g.,
3209 16=100% means that we submitted anywhere between 9 to 16 I/Os per submit
3210 call. Note that the range covered by a submit distribution entry can
3211 be different to the range covered by the equivalent depth distribution
3215 Like the above submit number, but for completions instead.
3218 The number of read/write/trim requests issued, and how many of them were
3222 The distribution of I/O completion latencies. This is the time from when
3223 I/O leaves fio and when it gets completed. The numbers follow the same
3224 pattern as the I/O depths, meaning that 2=1.6% means that 1.6% of the
3225 I/O completed within 2 msecs, 20=12.8% means that 12.8% of the I/O took
3226 more than 10 msecs, but less than (or equal to) 20 msecs.
3229 Example output was based on the following:
3230 TZ=UTC fio --ioengine=null --iodepth=2 --size=100M --numjobs=2 \
3231 --rate_process=poisson --io_limit=32M --name=read --bs=128k \
3232 --rate=11M --name=write --rw=write --bs=2k --rate=700k
3234 After each client has been listed, the group statistics are printed. They
3235 will look like this::
3237 Run status group 0 (all jobs):
3238 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
3239 WRITE: bw=1231KiB/s (1261kB/s), 616KiB/s-621KiB/s (630kB/s-636kB/s), io=64.0MiB (67.1MB), run=52747-53223msec
3241 For each data direction it prints:
3244 Aggregate bandwidth of threads in this group followed by the
3245 minimum and maximum bandwidth of all the threads in this group.
3246 Values outside of brackets are power-of-2 format and those
3247 within are the equivalent value in a power-of-10 format.
3249 Aggregate I/O performed of all threads in this group. The
3250 format is the same as bw.
3252 The smallest and longest runtimes of the threads in this group.
3254 And finally, the disk statistics are printed. They will look like this::
3256 Disk stats (read/write):
3257 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
3259 Each value is printed for both reads and writes, with reads first. The
3263 Number of I/Os performed by all groups.
3265 Number of merges performed by the I/O scheduler.
3267 Number of ticks we kept the disk busy.
3269 Total time spent in the disk queue.
3271 The disk utilization. A value of 100% means we kept the disk
3272 busy constantly, 50% would be a disk idling half of the time.
3274 It is also possible to get fio to dump the current output while it is running,
3275 without terminating the job. To do that, send fio the **USR1** signal. You can
3276 also get regularly timed dumps by using the :option:`--status-interval`
3277 parameter, or by creating a file in :file:`/tmp` named
3278 :file:`fio-dump-status`. If fio sees this file, it will unlink it and dump the
3279 current output status.
3285 For scripted usage where you typically want to generate tables or graphs of the
3286 results, fio can output the results in a semicolon separated format. The format
3287 is one long line of values, such as::
3289 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%
3290 A description of this job goes here.
3292 The job description (if provided) follows on a second line.
3294 To enable terse output, use the :option:`--minimal` or
3295 :option:`--output-format`\=terse command line options. The
3296 first value is the version of the terse output format. If the output has to be
3297 changed for some reason, this number will be incremented by 1 to signify that
3300 Split up, the format is as follows (comments in brackets denote when a
3301 field was introduced or whether it's specific to some terse version):
3305 terse version, fio version [v3], jobname, groupid, error
3309 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3310 Submission latency: min, max, mean, stdev (usec)
3311 Completion latency: min, max, mean, stdev (usec)
3312 Completion latency percentiles: 20 fields (see below)
3313 Total latency: min, max, mean, stdev (usec)
3314 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
3315 IOPS [v5]: min, max, mean, stdev, number of samples
3321 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3322 Submission latency: min, max, mean, stdev (usec)
3323 Completion latency: min, max, mean, stdev (usec)
3324 Completion latency percentiles: 20 fields (see below)
3325 Total latency: min, max, mean, stdev (usec)
3326 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
3327 IOPS [v5]: min, max, mean, stdev, number of samples
3329 TRIM status [all but version 3]:
3331 Fields are similar to READ/WRITE status.
3335 user, system, context switches, major faults, minor faults
3339 <=1, 2, 4, 8, 16, 32, >=64
3341 I/O latencies microseconds::
3343 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
3345 I/O latencies milliseconds::
3347 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
3349 Disk utilization [v3]::
3351 Disk name, Read ios, write ios,
3352 Read merges, write merges,
3353 Read ticks, write ticks,
3354 Time spent in queue, disk utilization percentage
3356 Additional Info (dependent on continue_on_error, default off)::
3358 total # errors, first error code
3360 Additional Info (dependent on description being set)::
3364 Completion latency percentiles can be a grouping of up to 20 sets, so for the
3365 terse output fio writes all of them. Each field will look like this::
3369 which is the Xth percentile, and the `usec` latency associated with it.
3371 For disk utilization, all disks used by fio are shown. So for each disk there
3372 will be a disk utilization section.
3374 Below is a single line containing short names for each of the fields in the
3375 minimal output v3, separated by semicolons::
3377 terse_version_3;fio_version;jobname;groupid;error;read_kb;read_bandwidth;read_iops;read_runtime_ms;read_slat_min;read_slat_max;read_slat_mean;read_slat_dev;read_clat_min;read_clat_max;read_clat_mean;read_clat_dev;read_clat_pct01;read_clat_pct02;read_clat_pct03;read_clat_pct04;read_clat_pct05;read_clat_pct06;read_clat_pct07;read_clat_pct08;read_clat_pct09;read_clat_pct10;read_clat_pct11;read_clat_pct12;read_clat_pct13;read_clat_pct14;read_clat_pct15;read_clat_pct16;read_clat_pct17;read_clat_pct18;read_clat_pct19;read_clat_pct20;read_tlat_min;read_lat_max;read_lat_mean;read_lat_dev;read_bw_min;read_bw_max;read_bw_agg_pct;read_bw_mean;read_bw_dev;write_kb;write_bandwidth;write_iops;write_runtime_ms;write_slat_min;write_slat_max;write_slat_mean;write_slat_dev;write_clat_min;write_clat_max;write_clat_mean;write_clat_dev;write_clat_pct01;write_clat_pct02;write_clat_pct03;write_clat_pct04;write_clat_pct05;write_clat_pct06;write_clat_pct07;write_clat_pct08;write_clat_pct09;write_clat_pct10;write_clat_pct11;write_clat_pct12;write_clat_pct13;write_clat_pct14;write_clat_pct15;write_clat_pct16;write_clat_pct17;write_clat_pct18;write_clat_pct19;write_clat_pct20;write_tlat_min;write_lat_max;write_lat_mean;write_lat_dev;write_bw_min;write_bw_max;write_bw_agg_pct;write_bw_mean;write_bw_dev;cpu_user;cpu_sys;cpu_csw;cpu_mjf;cpu_minf;iodepth_1;iodepth_2;iodepth_4;iodepth_8;iodepth_16;iodepth_32;iodepth_64;lat_2us;lat_4us;lat_10us;lat_20us;lat_50us;lat_100us;lat_250us;lat_500us;lat_750us;lat_1000us;lat_2ms;lat_4ms;lat_10ms;lat_20ms;lat_50ms;lat_100ms;lat_250ms;lat_500ms;lat_750ms;lat_1000ms;lat_2000ms;lat_over_2000ms;disk_name;disk_read_iops;disk_write_iops;disk_read_merges;disk_write_merges;disk_read_ticks;write_ticks;disk_queue_time;disk_util
3383 The `json+` output format is identical to the `json` output format except that it
3384 adds a full dump of the completion latency bins. Each `bins` object contains a
3385 set of (key, value) pairs where keys are latency durations and values count how
3386 many I/Os had completion latencies of the corresponding duration. For example,
3389 "bins" : { "87552" : 1, "89600" : 1, "94720" : 1, "96768" : 1, "97792" : 1, "99840" : 1, "100864" : 2, "103936" : 6, "104960" : 534, "105984" : 5995, "107008" : 7529, ... }
3391 This data indicates that one I/O required 87,552ns to complete, two I/Os required
3392 100,864ns to complete, and 7529 I/Os required 107,008ns to complete.
3394 Also included with fio is a Python script `fio_jsonplus_clat2csv` that takes
3395 json+ output and generates CSV-formatted latency data suitable for plotting.
3397 The latency durations actually represent the midpoints of latency intervals.
3398 For details refer to stat.h.
3404 There are two trace file format that you can encounter. The older (v1) format is
3405 unsupported since version 1.20-rc3 (March 2008). It will still be described
3406 below in case that you get an old trace and want to understand it.
3408 In any case the trace is a simple text file with a single action per line.
3411 Trace file format v1
3412 ~~~~~~~~~~~~~~~~~~~~
3414 Each line represents a single I/O action in the following format::
3418 where `rw=0/1` for read/write, and the offset and length entries being in bytes.
3420 This format is not supported in fio versions >= 1.20-rc3.
3423 Trace file format v2
3424 ~~~~~~~~~~~~~~~~~~~~
3426 The second version of the trace file format was added in fio version 1.17. It
3427 allows to access more then one file per trace and has a bigger set of possible
3430 The first line of the trace file has to be::
3434 Following this can be lines in two different formats, which are described below.
3436 The file management format::
3440 The filename is given as an absolute path. The action can be one of these:
3443 Add the given filename to the trace.
3445 Open the file with the given filename. The filename has to have
3446 been added with the **add** action before.
3448 Close the file with the given filename. The file has to have been
3452 The file I/O action format::
3454 filename action offset length
3456 The `filename` is given as an absolute path, and has to have been added and
3457 opened before it can be used with this format. The `offset` and `length` are
3458 given in bytes. The `action` can be one of these:
3461 Wait for `offset` microseconds. Everything below 100 is discarded.
3462 The time is relative to the previous `wait` statement.
3464 Read `length` bytes beginning from `offset`.
3466 Write `length` bytes beginning from `offset`.
3468 :manpage:`fsync(2)` the file.
3470 :manpage:`fdatasync(2)` the file.
3472 Trim the given file from the given `offset` for `length` bytes.
3474 CPU idleness profiling
3475 ----------------------
3477 In some cases, we want to understand CPU overhead in a test. For example, we
3478 test patches for the specific goodness of whether they reduce CPU usage.
3479 Fio implements a balloon approach to create a thread per CPU that runs at idle
3480 priority, meaning that it only runs when nobody else needs the cpu.
3481 By measuring the amount of work completed by the thread, idleness of each CPU
3482 can be derived accordingly.
3484 An unit work is defined as touching a full page of unsigned characters. Mean and
3485 standard deviation of time to complete an unit work is reported in "unit work"
3486 section. Options can be chosen to report detailed percpu idleness or overall
3487 system idleness by aggregating percpu stats.
3490 Verification and triggers
3491 -------------------------
3493 Fio is usually run in one of two ways, when data verification is done. The first
3494 is a normal write job of some sort with verify enabled. When the write phase has
3495 completed, fio switches to reads and verifies everything it wrote. The second
3496 model is running just the write phase, and then later on running the same job
3497 (but with reads instead of writes) to repeat the same I/O patterns and verify
3498 the contents. Both of these methods depend on the write phase being completed,
3499 as fio otherwise has no idea how much data was written.
3501 With verification triggers, fio supports dumping the current write state to
3502 local files. Then a subsequent read verify workload can load this state and know
3503 exactly where to stop. This is useful for testing cases where power is cut to a
3504 server in a managed fashion, for instance.
3506 A verification trigger consists of two things:
3508 1) Storing the write state of each job.
3509 2) Executing a trigger command.
3511 The write state is relatively small, on the order of hundreds of bytes to single
3512 kilobytes. It contains information on the number of completions done, the last X
3515 A trigger is invoked either through creation ('touch') of a specified file in
3516 the system, or through a timeout setting. If fio is run with
3517 :option:`--trigger-file`\= :file:`/tmp/trigger-file`, then it will continually
3518 check for the existence of :file:`/tmp/trigger-file`. When it sees this file, it
3519 will fire off the trigger (thus saving state, and executing the trigger
3522 For client/server runs, there's both a local and remote trigger. If fio is
3523 running as a server backend, it will send the job states back to the client for
3524 safe storage, then execute the remote trigger, if specified. If a local trigger
3525 is specified, the server will still send back the write state, but the client
3526 will then execute the trigger.
3528 Verification trigger example
3529 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3531 Let's say we want to run a powercut test on the remote machine 'server'. Our
3532 write workload is in :file:`write-test.fio`. We want to cut power to 'server' at
3533 some point during the run, and we'll run this test from the safety or our local
3534 machine, 'localbox'. On the server, we'll start the fio backend normally::
3536 server# fio --server
3538 and on the client, we'll fire off the workload::
3540 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger-remote="bash -c \"echo b > /proc/sysrq-triger\""
3542 We set :file:`/tmp/my-trigger` as the trigger file, and we tell fio to execute::
3544 echo b > /proc/sysrq-trigger
3546 on the server once it has received the trigger and sent us the write state. This
3547 will work, but it's not **really** cutting power to the server, it's merely
3548 abruptly rebooting it. If we have a remote way of cutting power to the server
3549 through IPMI or similar, we could do that through a local trigger command
3550 instead. Let's assume we have a script that does IPMI reboot of a given hostname,
3551 ipmi-reboot. On localbox, we could then have run fio with a local trigger
3554 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger="ipmi-reboot server"
3556 For this case, fio would wait for the server to send us the write state, then
3557 execute ``ipmi-reboot server`` when that happened.
3559 Loading verify state
3560 ~~~~~~~~~~~~~~~~~~~~
3562 To load stored write state, a read verification job file must contain the
3563 :option:`verify_state_load` option. If that is set, fio will load the previously
3564 stored state. For a local fio run this is done by loading the files directly,
3565 and on a client/server run, the server backend will ask the client to send the
3566 files over and load them from there.
3572 Fio supports a variety of log file formats, for logging latencies, bandwidth,
3573 and IOPS. The logs share a common format, which looks like this:
3575 *time* (`msec`), *value*, *data direction*, *block size* (`bytes`),
3578 *Time* for the log entry is always in milliseconds. The *value* logged depends
3579 on the type of log, it will be one of the following:
3582 Value is latency in usecs
3588 *Data direction* is one of the following:
3597 The entry's *block size* is always in bytes. The *offset* is the offset, in bytes,
3598 from the start of the file, for that particular I/O. The logging of the offset can be
3599 toggled with :option:`log_offset`.
3601 Fio defaults to logging every individual I/O. When IOPS are logged for individual
3602 I/Os the *value* entry will always be 1. If windowed logging is enabled through
3603 :option:`log_avg_msec`, fio logs the average values over the specified period of time.
3604 If windowed logging is enabled and :option:`log_max_value` is set, then fio logs
3605 maximum values in that window instead of averages. Since *data direction*, *block
3606 size* and *offset* are per-I/O values, if windowed logging is enabled they
3607 aren't applicable and will be 0.
3612 Normally fio is invoked as a stand-alone application on the machine where the
3613 I/O workload should be generated. However, the backend and frontend of fio can
3614 be run separately i.e., the fio server can generate an I/O workload on the "Device
3615 Under Test" while being controlled by a client on another machine.
3617 Start the server on the machine which has access to the storage DUT::
3621 where `args` defines what fio listens to. The arguments are of the form
3622 ``type,hostname`` or ``IP,port``. *type* is either ``ip`` (or ip4) for TCP/IP
3623 v4, ``ip6`` for TCP/IP v6, or ``sock`` for a local unix domain socket.
3624 *hostname* is either a hostname or IP address, and *port* is the port to listen
3625 to (only valid for TCP/IP, not a local socket). Some examples:
3629 Start a fio server, listening on all interfaces on the default port (8765).
3631 2) ``fio --server=ip:hostname,4444``
3633 Start a fio server, listening on IP belonging to hostname and on port 4444.
3635 3) ``fio --server=ip6:::1,4444``
3637 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
3639 4) ``fio --server=,4444``
3641 Start a fio server, listening on all interfaces on port 4444.
3643 5) ``fio --server=1.2.3.4``
3645 Start a fio server, listening on IP 1.2.3.4 on the default port.
3647 6) ``fio --server=sock:/tmp/fio.sock``
3649 Start a fio server, listening on the local socket :file:`/tmp/fio.sock`.
3651 Once a server is running, a "client" can connect to the fio server with::
3653 fio <local-args> --client=<server> <remote-args> <job file(s)>
3655 where `local-args` are arguments for the client where it is running, `server`
3656 is the connect string, and `remote-args` and `job file(s)` are sent to the
3657 server. The `server` string follows the same format as it does on the server
3658 side, to allow IP/hostname/socket and port strings.
3660 Fio can connect to multiple servers this way::
3662 fio --client=<server1> <job file(s)> --client=<server2> <job file(s)>
3664 If the job file is located on the fio server, then you can tell the server to
3665 load a local file as well. This is done by using :option:`--remote-config` ::
3667 fio --client=server --remote-config /path/to/file.fio
3669 Then fio will open this local (to the server) job file instead of being passed
3670 one from the client.
3672 If you have many servers (example: 100 VMs/containers), you can input a pathname
3673 of a file containing host IPs/names as the parameter value for the
3674 :option:`--client` option. For example, here is an example :file:`host.list`
3675 file containing 2 hostnames::
3677 host1.your.dns.domain
3678 host2.your.dns.domain
3680 The fio command would then be::
3682 fio --client=host.list <job file(s)>
3684 In this mode, you cannot input server-specific parameters or job files -- all
3685 servers receive the same job file.
3687 In order to let ``fio --client`` runs use a shared filesystem from multiple
3688 hosts, ``fio --client`` now prepends the IP address of the server to the
3689 filename. For example, if fio is using the directory :file:`/mnt/nfs/fio` and is
3690 writing filename :file:`fileio.tmp`, with a :option:`--client` `hostfile`
3691 containing two hostnames ``h1`` and ``h2`` with IP addresses 192.168.10.120 and
3692 192.168.10.121, then fio will create two files::
3694 /mnt/nfs/fio/192.168.10.120.fileio.tmp
3695 /mnt/nfs/fio/192.168.10.121.fileio.tmp