4 The first step in getting fio to simulate a desired I/O workload, is writing a
5 job file describing that specific setup. A job file may contain any number of
6 threads and/or files -- the typical contents of the job file is a *global*
7 section defining shared parameters, and one or more job sections describing the
8 jobs involved. When run, fio parses this file and sets everything up as
9 described. If we break down a job from top to bottom, it contains the following
14 Defines the I/O pattern issued to the file(s). We may only be reading
15 sequentially from this file(s), or we may be writing randomly. Or even
16 mixing reads and writes, sequentially or randomly.
17 Should we be doing buffered I/O, or direct/raw I/O?
21 In how large chunks are we issuing I/O? This may be a single value,
22 or it may describe a range of block sizes.
26 How much data are we going to be reading/writing.
30 How do we issue I/O? We could be memory mapping the file, we could be
31 using regular read/write, we could be using splice, async I/O, or even
36 If the I/O engine is async, how large a queuing depth do we want to
42 How many files are we spreading the workload over.
44 `Threads, processes and job synchronization`_
46 How many threads or processes should we spread this workload over.
48 The above are the basic parameters defined for a workload, in addition there's a
49 multitude of parameters that modify other aspects of how this job behaves.
55 .. option:: --debug=type
57 Enable verbose tracing `type` of various fio actions. May be ``all`` for all types
58 or individual types separated by a comma (e.g. ``--debug=file,mem`` will
59 enable file and memory debugging). Currently, additional logging is
63 Dump info related to processes.
65 Dump info related to file actions.
67 Dump info related to I/O queuing.
69 Dump info related to memory allocations.
71 Dump info related to blktrace setup.
73 Dump info related to I/O verification.
75 Enable all debug options.
77 Dump info related to random offset generation.
79 Dump info related to option matching and parsing.
81 Dump info related to disk utilization updates.
83 Dump info only related to job number x.
85 Dump info only related to mutex up/down ops.
87 Dump info related to profile extensions.
89 Dump info related to internal time keeping.
91 Dump info related to networking connections.
93 Dump info related to I/O rate switching.
95 Dump info related to log compress/decompress.
97 Show available debug options.
99 .. option:: --parse-only
101 Parse options only, don't start any I/O.
103 .. option:: --output=filename
105 Write output to file `filename`.
107 .. option:: --output-format=format
109 Set the reporting `format` to `normal`, `terse`, `json`, or `json+`. Multiple
110 formats can be selected, separated by a comma. `terse` is a CSV based
111 format. `json+` is like `json`, except it adds a full dump of the latency
114 .. option:: --bandwidth-log
116 Generate aggregate bandwidth logs.
118 .. option:: --minimal
120 Print statistics in a terse, semicolon-delimited format.
122 .. option:: --append-terse
124 Print statistics in selected mode AND terse, semicolon-delimited format.
125 **Deprecated**, use :option:`--output-format` instead to select multiple
128 .. option:: --terse-version=version
130 Set terse `version` output format (default 3, or 2 or 4 or 5).
132 .. option:: --version
134 Print version information and exit.
138 Print a summary of the command line options and exit.
140 .. option:: --cpuclock-test
142 Perform test and validation of internal CPU clock.
144 .. option:: --crctest=[test]
146 Test the speed of the built-in checksumming functions. If no argument is
147 given, all of them are tested. Alternatively, a comma separated list can
148 be passed, in which case the given ones are tested.
150 .. option:: --cmdhelp=command
152 Print help information for `command`. May be ``all`` for all commands.
154 .. option:: --enghelp=[ioengine[,command]]
156 List all commands defined by `ioengine`, or print help for `command`
157 defined by `ioengine`. If no `ioengine` is given, list all
160 .. option:: --showcmd=jobfile
162 Convert `jobfile` to a set of command-line options.
164 .. option:: --readonly
166 Turn on safety read-only checks, preventing writes. The ``--readonly``
167 option is an extra safety guard to prevent users from accidentally starting
168 a write workload when that is not desired. Fio will only write if
169 `rw=write/randwrite/rw/randrw` is given. This extra safety net can be used
170 as an extra precaution as ``--readonly`` will also enable a write check in
171 the I/O engine core to prevent writes due to unknown user space bug(s).
173 .. option:: --eta=when
175 Specifies when real-time ETA estimate should be printed. `when` may be
176 `always`, `never` or `auto`.
178 .. option:: --eta-newline=time
180 Force a new line for every `time` period passed. When the unit is omitted,
181 the value is interpreted in seconds.
183 .. option:: --status-interval=time
185 Force a full status dump of cumulative (from job start) values at `time`
186 intervals. This option does *not* provide per-period measurements. So
187 values such as bandwidth are running averages. When the time unit is omitted,
188 `time` is interpreted in seconds.
190 .. option:: --section=name
192 Only run specified section `name` in job file. Multiple sections can be specified.
193 The ``--section`` option allows one to combine related jobs into one file.
194 E.g. one job file could define light, moderate, and heavy sections. Tell
195 fio to run only the "heavy" section by giving ``--section=heavy``
196 command line option. One can also specify the "write" operations in one
197 section and "verify" operation in another section. The ``--section`` option
198 only applies to job sections. The reserved *global* section is always
201 .. option:: --alloc-size=kb
203 Set the internal smalloc pool size to `kb` in KiB. The
204 ``--alloc-size`` switch allows one to use a larger pool size for smalloc.
205 If running large jobs with randommap enabled, fio can run out of memory.
206 Smalloc is an internal allocator for shared structures from a fixed size
207 memory pool and can grow to 16 pools. The pool size defaults to 16MiB.
209 NOTE: While running :file:`.fio_smalloc.*` backing store files are visible
212 .. option:: --warnings-fatal
214 All fio parser warnings are fatal, causing fio to exit with an
217 .. option:: --max-jobs=nr
219 Set the maximum number of threads/processes to support to `nr`.
220 NOTE: On Linux, it may be necessary to increase the shared-memory
221 limit (:file:`/proc/sys/kernel/shmmax`) if fio runs into errors while
224 .. option:: --server=args
226 Start a backend server, with `args` specifying what to listen to.
227 See `Client/Server`_ section.
229 .. option:: --daemonize=pidfile
231 Background a fio server, writing the pid to the given `pidfile` file.
233 .. option:: --client=hostname
235 Instead of running the jobs locally, send and run them on the given `hostname`
236 or set of `hostname`\s. See `Client/Server`_ section.
238 .. option:: --remote-config=file
240 Tell fio server to load this local `file`.
242 .. option:: --idle-prof=option
244 Report CPU idleness. `option` is one of the following:
247 Run unit work calibration only and exit.
250 Show aggregate system idleness and unit work.
253 As **system** but also show per CPU idleness.
255 .. option:: --inflate-log=log
257 Inflate and output compressed `log`.
259 .. option:: --trigger-file=file
261 Execute trigger command when `file` exists.
263 .. option:: --trigger-timeout=time
265 Execute trigger at this `time`.
267 .. option:: --trigger=command
269 Set this `command` as local trigger.
271 .. option:: --trigger-remote=command
273 Set this `command` as remote trigger.
275 .. option:: --aux-path=path
277 Use this `path` for fio state generated files.
279 Any parameters following the options will be assumed to be job files, unless
280 they match a job file parameter. Multiple job files can be listed and each job
281 file will be regarded as a separate group. Fio will :option:`stonewall`
282 execution between each group.
288 As previously described, fio accepts one or more job files describing what it is
289 supposed to do. The job file format is the classic ini file, where the names
290 enclosed in [] brackets define the job name. You are free to use any ASCII name
291 you want, except *global* which has special meaning. Following the job name is
292 a sequence of zero or more parameters, one per line, that define the behavior of
293 the job. If the first character in a line is a ';' or a '#', the entire line is
294 discarded as a comment.
296 A *global* section sets defaults for the jobs described in that file. A job may
297 override a *global* section parameter, and a job file may even have several
298 *global* sections if so desired. A job is only affected by a *global* section
301 The :option:`--cmdhelp` option also lists all options. If used with a `command`
302 argument, :option:`--cmdhelp` will detail the given `command`.
304 See the `examples/` directory for inspiration on how to write job files. Note
305 the copyright and license requirements currently apply to `examples/` files.
307 So let's look at a really simple job file that defines two processes, each
308 randomly reading from a 128MiB file:
312 ; -- start job file --
323 As you can see, the job file sections themselves are empty as all the described
324 parameters are shared. As no :option:`filename` option is given, fio makes up a
325 `filename` for each of the jobs as it sees fit. On the command line, this job
326 would look as follows::
328 $ fio --name=global --rw=randread --size=128m --name=job1 --name=job2
331 Let's look at an example that has a number of processes writing randomly to
336 ; -- start job file --
347 Here we have no *global* section, as we only have one job defined anyway. We
348 want to use async I/O here, with a depth of 4 for each file. We also increased
349 the buffer size used to 32KiB and define numjobs to 4 to fork 4 identical
350 jobs. The result is 4 processes each randomly writing to their own 64MiB
351 file. Instead of using the above job file, you could have given the parameters
352 on the command line. For this case, you would specify::
354 $ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4
356 When fio is utilized as a basis of any reasonably large test suite, it might be
357 desirable to share a set of standardized settings across multiple job files.
358 Instead of copy/pasting such settings, any section may pull in an external
359 :file:`filename.fio` file with *include filename* directive, as in the following
362 ; -- start job file including.fio --
366 include glob-include.fio
373 include test-include.fio
374 ; -- end job file including.fio --
378 ; -- start job file glob-include.fio --
381 ; -- end job file glob-include.fio --
385 ; -- start job file test-include.fio --
388 ; -- end job file test-include.fio --
390 Settings pulled into a section apply to that section only (except *global*
391 section). Include directives may be nested in that any included file may contain
392 further include directive(s). Include files may not contain [] sections.
395 Environment variables
396 ~~~~~~~~~~~~~~~~~~~~~
398 Fio also supports environment variable expansion in job files. Any sub-string of
399 the form ``${VARNAME}`` as part of an option value (in other words, on the right
400 of the '='), will be expanded to the value of the environment variable called
401 `VARNAME`. If no such environment variable is defined, or `VARNAME` is the
402 empty string, the empty string will be substituted.
404 As an example, let's look at a sample fio invocation and job file::
406 $ SIZE=64m NUMJOBS=4 fio jobfile.fio
410 ; -- start job file --
417 This will expand to the following equivalent job file at runtime:
421 ; -- start job file --
428 Fio ships with a few example job files, you can also look there for inspiration.
433 Additionally, fio has a set of reserved keywords that will be replaced
434 internally with the appropriate value. Those keywords are:
438 The architecture page size of the running system.
442 Megabytes of total memory in the system.
446 Number of online available CPUs.
448 These can be used on the command line or in the job file, and will be
449 automatically substituted with the current system values when the job is
450 run. Simple math is also supported on these keywords, so you can perform actions
455 and get that properly expanded to 8 times the size of memory in the machine.
461 This section describes in details each parameter associated with a job. Some
462 parameters take an option of a given type, such as an integer or a
463 string. Anywhere a numeric value is required, an arithmetic expression may be
464 used, provided it is surrounded by parentheses. Supported operators are:
473 For time values in expressions, units are microseconds by default. This is
474 different than for time values not in expressions (not enclosed in
475 parentheses). The following types are used:
482 String: A sequence of alphanumeric characters.
485 Integer with possible time suffix. Without a unit value is interpreted as
486 seconds unless otherwise specified. Accepts a suffix of 'd' for days, 'h' for
487 hours, 'm' for minutes, 's' for seconds, 'ms' (or 'msec') for milliseconds and
488 'us' (or 'usec') for microseconds. For example, use 10m for 10 minutes.
493 Integer. A whole number value, which may contain an integer prefix
494 and an integer suffix:
496 [*integer prefix*] **number** [*integer suffix*]
498 The optional *integer prefix* specifies the number's base. The default
499 is decimal. *0x* specifies hexadecimal.
501 The optional *integer suffix* specifies the number's units, and includes an
502 optional unit prefix and an optional unit. For quantities of data, the
503 default unit is bytes. For quantities of time, the default unit is seconds
504 unless otherwise specified.
506 With :option:`kb_base`\=1000, fio follows international standards for unit
507 prefixes. To specify power-of-10 decimal values defined in the
508 International System of Units (SI):
510 * *K* -- means kilo (K) or 1000
511 * *M* -- means mega (M) or 1000**2
512 * *G* -- means giga (G) or 1000**3
513 * *T* -- means tera (T) or 1000**4
514 * *P* -- means peta (P) or 1000**5
516 To specify power-of-2 binary values defined in IEC 80000-13:
518 * *Ki* -- means kibi (Ki) or 1024
519 * *Mi* -- means mebi (Mi) or 1024**2
520 * *Gi* -- means gibi (Gi) or 1024**3
521 * *Ti* -- means tebi (Ti) or 1024**4
522 * *Pi* -- means pebi (Pi) or 1024**5
524 With :option:`kb_base`\=1024 (the default), the unit prefixes are opposite
525 from those specified in the SI and IEC 80000-13 standards to provide
526 compatibility with old scripts. For example, 4k means 4096.
528 For quantities of data, an optional unit of 'B' may be included
529 (e.g., 'kB' is the same as 'k').
531 The *integer suffix* is not case sensitive (e.g., m/mi mean mebi/mega,
532 not milli). 'b' and 'B' both mean byte, not bit.
534 Examples with :option:`kb_base`\=1000:
536 * *4 KiB*: 4096, 4096b, 4096B, 4ki, 4kib, 4kiB, 4Ki, 4KiB
537 * *1 MiB*: 1048576, 1mi, 1024ki
538 * *1 MB*: 1000000, 1m, 1000k
539 * *1 TiB*: 1099511627776, 1ti, 1024gi, 1048576mi
540 * *1 TB*: 1000000000, 1t, 1000m, 1000000k
542 Examples with :option:`kb_base`\=1024 (default):
544 * *4 KiB*: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB
545 * *1 MiB*: 1048576, 1m, 1024k
546 * *1 MB*: 1000000, 1mi, 1000ki
547 * *1 TiB*: 1099511627776, 1t, 1024g, 1048576m
548 * *1 TB*: 1000000000, 1ti, 1000mi, 1000000ki
550 To specify times (units are not case sensitive):
554 * *M* -- means minutes
555 * *s* -- or sec means seconds (default)
556 * *ms* -- or *msec* means milliseconds
557 * *us* -- or *usec* means microseconds
559 If the option accepts an upper and lower range, use a colon ':' or
560 minus '-' to separate such values. See :ref:`irange <irange>`.
561 If the lower value specified happens to be larger than the upper value
562 the two values are swapped.
567 Boolean. Usually parsed as an integer, however only defined for
568 true and false (1 and 0).
573 Integer range with suffix. Allows value range to be given, such as
574 1024-4096. A colon may also be used as the separator, e.g. 1k:4k. If the
575 option allows two sets of ranges, they can be specified with a ',' or '/'
576 delimiter: 1k-4k/8k-32k. Also see :ref:`int <int>`.
579 A list of floating point numbers, separated by a ':' character.
581 With the above in mind, here follows the complete list of fio job parameters.
587 .. option:: kb_base=int
589 Select the interpretation of unit prefixes in input parameters.
592 Inputs comply with IEC 80000-13 and the International
593 System of Units (SI). Use:
595 - power-of-2 values with IEC prefixes (e.g., KiB)
596 - power-of-10 values with SI prefixes (e.g., kB)
599 Compatibility mode (default). To avoid breaking old scripts:
601 - power-of-2 values with SI prefixes
602 - power-of-10 values with IEC prefixes
604 See :option:`bs` for more details on input parameters.
606 Outputs always use correct prefixes. Most outputs include both
609 bw=2383.3kB/s (2327.4KiB/s)
611 If only one value is reported, then kb_base selects the one to use:
613 **1000** -- SI prefixes
615 **1024** -- IEC prefixes
617 .. option:: unit_base=int
619 Base unit for reporting. Allowed values are:
622 Use auto-detection (default).
634 ASCII name of the job. This may be used to override the name printed by fio
635 for this job. Otherwise the job name is used. On the command line this
636 parameter has the special purpose of also signaling the start of a new job.
638 .. option:: description=str
640 Text description of the job. Doesn't do anything except dump this text
641 description when this job is run. It's not parsed.
643 .. option:: loops=int
645 Run the specified number of iterations of this job. Used to repeat the same
646 workload a given number of times. Defaults to 1.
648 .. option:: numjobs=int
650 Create the specified number of clones of this job. Each clone of job
651 is spawned as an independent thread or process. May be used to setup a
652 larger number of threads/processes doing the same thing. Each thread is
653 reported separately; to see statistics for all clones as a whole, use
654 :option:`group_reporting` in conjunction with :option:`new_group`.
655 See :option:`--max-jobs`. Default: 1.
658 Time related parameters
659 ~~~~~~~~~~~~~~~~~~~~~~~
661 .. option:: runtime=time
663 Tell fio to terminate processing after the specified period of time. It
664 can be quite hard to determine for how long a specified job will run, so
665 this parameter is handy to cap the total runtime to a given time. When
666 the unit is omitted, the value is intepreted in seconds.
668 .. option:: time_based
670 If set, fio will run for the duration of the :option:`runtime` specified
671 even if the file(s) are completely read or written. It will simply loop over
672 the same workload as many times as the :option:`runtime` allows.
674 .. option:: startdelay=irange(time)
676 Delay the start of job for the specified amount of time. Can be a single
677 value or a range. When given as a range, each thread will choose a value
678 randomly from within the range. Value is in seconds if a unit is omitted.
680 .. option:: ramp_time=time
682 If set, fio will run the specified workload for this amount of time before
683 logging any performance numbers. Useful for letting performance settle
684 before logging results, thus minimizing the runtime required for stable
685 results. Note that the ``ramp_time`` is considered lead in time for a job,
686 thus it will increase the total runtime if a special timeout or
687 :option:`runtime` is specified. When the unit is omitted, the value is
690 .. option:: clocksource=str
692 Use the given clocksource as the base of timing. The supported options are:
695 :manpage:`gettimeofday(2)`
698 :manpage:`clock_gettime(2)`
701 Internal CPU clock source
703 cpu is the preferred clocksource if it is reliable, as it is very fast (and
704 fio is heavy on time calls). Fio will automatically use this clocksource if
705 it's supported and considered reliable on the system it is running on,
706 unless another clocksource is specifically set. For x86/x86-64 CPUs, this
707 means supporting TSC Invariant.
709 .. option:: gtod_reduce=bool
711 Enable all of the :manpage:`gettimeofday(2)` reducing options
712 (:option:`disable_clat`, :option:`disable_slat`, :option:`disable_bw_measurement`) plus
713 reduce precision of the timeout somewhat to really shrink the
714 :manpage:`gettimeofday(2)` call count. With this option enabled, we only do
715 about 0.4% of the :manpage:`gettimeofday(2)` calls we would have done if all
716 time keeping was enabled.
718 .. option:: gtod_cpu=int
720 Sometimes it's cheaper to dedicate a single thread of execution to just
721 getting the current time. Fio (and databases, for instance) are very
722 intensive on :manpage:`gettimeofday(2)` calls. With this option, you can set
723 one CPU aside for doing nothing but logging current time to a shared memory
724 location. Then the other threads/processes that run I/O workloads need only
725 copy that segment, instead of entering the kernel with a
726 :manpage:`gettimeofday(2)` call. The CPU set aside for doing these time
727 calls will be excluded from other uses. Fio will manually clear it from the
728 CPU mask of other jobs.
734 .. option:: directory=str
736 Prefix filenames with this directory. Used to place files in a different
737 location than :file:`./`. You can specify a number of directories by
738 separating the names with a ':' character. These directories will be
739 assigned equally distributed to job clones created by :option:`numjobs` as
740 long as they are using generated filenames. If specific `filename(s)` are
741 set fio will use the first listed directory, and thereby matching the
742 `filename` semantic which generates a file each clone if not specified, but
743 let all clones use the same if set.
745 See the :option:`filename` option for information on how to escape "``:``" and
746 "``\``" characters within the directory path itself.
748 .. option:: filename=str
750 Fio normally makes up a `filename` based on the job name, thread number, and
751 file number (see :option:`filename_format`). If you want to share files
752 between threads in a job or several
753 jobs with fixed file paths, specify a `filename` for each of them to override
754 the default. If the ioengine is file based, you can specify a number of files
755 by separating the names with a ':' colon. So if you wanted a job to open
756 :file:`/dev/sda` and :file:`/dev/sdb` as the two working files, you would use
757 ``filename=/dev/sda:/dev/sdb``. This also means that whenever this option is
758 specified, :option:`nrfiles` is ignored. The size of regular files specified
759 by this option will be :option:`size` divided by number of files unless an
760 explicit size is specified by :option:`filesize`.
762 Each colon and backslash in the wanted path must be escaped with a ``\``
763 character. For instance, if the path is :file:`/dev/dsk/foo@3,0:c` then you
764 would use ``filename=/dev/dsk/foo@3,0\:c`` and if the path is
765 :file:`F:\\filename` then you would use ``filename=F\:\\filename``.
767 On Windows, disk devices are accessed as :file:`\\\\.\\PhysicalDrive0` for
768 the first device, :file:`\\\\.\\PhysicalDrive1` for the second etc.
769 Note: Windows and FreeBSD prevent write access to areas
770 of the disk containing in-use data (e.g. filesystems).
772 The filename "`-`" is a reserved name, meaning *stdin* or *stdout*. Which
773 of the two depends on the read/write direction set.
775 .. option:: filename_format=str
777 If sharing multiple files between jobs, it is usually necessary to have fio
778 generate the exact names that you want. By default, fio will name a file
779 based on the default file format specification of
780 :file:`jobname.jobnumber.filenumber`. With this option, that can be
781 customized. Fio will recognize and replace the following keywords in this
785 The name of the worker thread or process.
787 The incremental number of the worker thread or process.
789 The incremental number of the file for that worker thread or
792 To have dependent jobs share a set of files, this option can be set to have
793 fio generate filenames that are shared between the two. For instance, if
794 :file:`testfiles.$filenum` is specified, file number 4 for any job will be
795 named :file:`testfiles.4`. The default of :file:`$jobname.$jobnum.$filenum`
796 will be used if no other format specifier is given.
798 If you specify a path then the directories will be created up to the
799 main directory for the file. So for example if you specify
800 ``filename_format=a/b/c/$jobnum`` then the directories a/b/c will be
801 created before the file setup part of the job. If you specify
802 :option:`directory` then the path will be relative that directory,
803 otherwise it is treated as the absolute path.
805 .. option:: unique_filename=bool
807 To avoid collisions between networked clients, fio defaults to prefixing any
808 generated filenames (with a directory specified) with the source of the
809 client connecting. To disable this behavior, set this option to 0.
811 .. option:: opendir=str
813 Recursively open any files below directory `str`.
815 .. option:: lockfile=str
817 Fio defaults to not locking any files before it does I/O to them. If a file
818 or file descriptor is shared, fio can serialize I/O to that file to make the
819 end result consistent. This is usual for emulating real workloads that share
820 files. The lock modes are:
823 No locking. The default.
825 Only one thread or process may do I/O at a time, excluding all
828 Read-write locking on the file. Many readers may
829 access the file at the same time, but writes get exclusive access.
831 .. option:: nrfiles=int
833 Number of files to use for this job. Defaults to 1. The size of files
834 will be :option:`size` divided by this unless explicit size is specified by
835 :option:`filesize`. Files are created for each thread separately, and each
836 file will have a file number within its name by default, as explained in
837 :option:`filename` section.
840 .. option:: openfiles=int
842 Number of files to keep open at the same time. Defaults to the same as
843 :option:`nrfiles`, can be set smaller to limit the number simultaneous
846 .. option:: file_service_type=str
848 Defines how fio decides which file from a job to service next. The following
852 Choose a file at random.
855 Round robin over opened files. This is the default.
858 Finish one file before moving on to the next. Multiple files can
859 still be open depending on :option:`openfiles`.
862 Use a *Zipf* distribution to decide what file to access.
865 Use a *Pareto* distribution to decide what file to access.
868 Use a *Gaussian* (normal) distribution to decide what file to
874 For *random*, *roundrobin*, and *sequential*, a postfix can be appended to
875 tell fio how many I/Os to issue before switching to a new file. For example,
876 specifying ``file_service_type=random:8`` would cause fio to issue
877 8 I/Os before selecting a new file at random. For the non-uniform
878 distributions, a floating point postfix can be given to influence how the
879 distribution is skewed. See :option:`random_distribution` for a description
880 of how that would work.
882 .. option:: ioscheduler=str
884 Attempt to switch the device hosting the file to the specified I/O scheduler
887 .. option:: create_serialize=bool
889 If true, serialize the file creation for the jobs. This may be handy to
890 avoid interleaving of data files, which may greatly depend on the filesystem
891 used and even the number of processors in the system. Default: true.
893 .. option:: create_fsync=bool
895 :manpage:`fsync(2)` the data file after creation. This is the default.
897 .. option:: create_on_open=bool
899 If true, don't pre-create files but allow the job's open() to create a file
900 when it's time to do I/O. Default: false -- pre-create all necessary files
903 .. option:: create_only=bool
905 If true, fio will only run the setup phase of the job. If files need to be
906 laid out or updated on disk, only that will be done -- the actual job contents
907 are not executed. Default: false.
909 .. option:: allow_file_create=bool
911 If true, fio is permitted to create files as part of its workload. If this
912 option is false, then fio will error out if
913 the files it needs to use don't already exist. Default: true.
915 .. option:: allow_mounted_write=bool
917 If this isn't set, fio will abort jobs that are destructive (e.g. that write)
918 to what appears to be a mounted device or partition. This should help catch
919 creating inadvertently destructive tests, not realizing that the test will
920 destroy data on the mounted file system. Note that some platforms don't allow
921 writing against a mounted device regardless of this option. Default: false.
923 .. option:: pre_read=bool
925 If this is given, files will be pre-read into memory before starting the
926 given I/O operation. This will also clear the :option:`invalidate` flag,
927 since it is pointless to pre-read and then drop the cache. This will only
928 work for I/O engines that are seek-able, since they allow you to read the
929 same data multiple times. Thus it will not work on non-seekable I/O engines
930 (e.g. network, splice). Default: false.
932 .. option:: unlink=bool
934 Unlink the job files when done. Not the default, as repeated runs of that
935 job would then waste time recreating the file set again and again. Default:
938 .. option:: unlink_each_loop=bool
940 Unlink job files after each iteration or loop. Default: false.
942 .. option:: zonesize=int
944 Divide a file into zones of the specified size. See :option:`zoneskip`.
946 .. option:: zonerange=int
948 Give size of an I/O zone. See :option:`zoneskip`.
950 .. option:: zoneskip=int
952 Skip the specified number of bytes when :option:`zonesize` data has been
953 read. The two zone options can be used to only do I/O on zones of a file.
959 .. option:: direct=bool
961 If value is true, use non-buffered I/O. This is usually O_DIRECT. Note that
962 OpenBSD and ZFS on Solaris don't support direct I/O. On Windows the synchronous
963 ioengines don't support direct I/O. Default: false.
965 .. option:: atomic=bool
967 If value is true, attempt to use atomic direct I/O. Atomic writes are
968 guaranteed to be stable once acknowledged by the operating system. Only
969 Linux supports O_ATOMIC right now.
971 .. option:: buffered=bool
973 If value is true, use buffered I/O. This is the opposite of the
974 :option:`direct` option. Defaults to true.
976 .. option:: readwrite=str, rw=str
978 Type of I/O pattern. Accepted values are:
985 Sequential trims (Linux block devices only).
991 Random trims (Linux block devices only).
993 Sequential mixed reads and writes.
995 Random mixed reads and writes.
997 Sequential trim+write sequences. Blocks will be trimmed first,
998 then the same blocks will be written to.
1000 Fio defaults to read if the option is not specified. For the mixed I/O
1001 types, the default is to split them 50/50. For certain types of I/O the
1002 result may still be skewed a bit, since the speed may be different.
1004 It is possible to specify the number of I/Os to do before getting a new
1005 offset by appending ``:<nr>`` to the end of the string given. For a
1006 random read, it would look like ``rw=randread:8`` for passing in an offset
1007 modifier with a value of 8. If the suffix is used with a sequential I/O
1008 pattern, then the *<nr>* value specified will be **added** to the generated
1009 offset for each I/O turning sequential I/O into sequential I/O with holes.
1010 For instance, using ``rw=write:4k`` will skip 4k for every write. Also see
1011 the :option:`rw_sequencer` option.
1013 .. option:: rw_sequencer=str
1015 If an offset modifier is given by appending a number to the ``rw=<str>``
1016 line, then this option controls how that number modifies the I/O offset
1017 being generated. Accepted values are:
1020 Generate sequential offset.
1022 Generate the same offset.
1024 ``sequential`` is only useful for random I/O, where fio would normally
1025 generate a new random offset for every I/O. If you append e.g. 8 to randread,
1026 you would get a new random offset for every 8 I/Os. The result would be a
1027 seek for only every 8 I/Os, instead of for every I/O. Use ``rw=randread:8``
1028 to specify that. As sequential I/O is already sequential, setting
1029 ``sequential`` for that would not result in any differences. ``identical``
1030 behaves in a similar fashion, except it sends the same offset 8 number of
1031 times before generating a new offset.
1033 .. option:: unified_rw_reporting=bool
1035 Fio normally reports statistics on a per data direction basis, meaning that
1036 reads, writes, and trims are accounted and reported separately. If this
1037 option is set fio sums the results and report them as "mixed" instead.
1039 .. option:: randrepeat=bool
1041 Seed the random number generator used for random I/O patterns in a
1042 predictable way so the pattern is repeatable across runs. Default: true.
1044 .. option:: allrandrepeat=bool
1046 Seed all random number generators in a predictable way so results are
1047 repeatable across runs. Default: false.
1049 .. option:: randseed=int
1051 Seed the random number generators based on this seed value, to be able to
1052 control what sequence of output is being generated. If not set, the random
1053 sequence depends on the :option:`randrepeat` setting.
1055 .. option:: fallocate=str
1057 Whether pre-allocation is performed when laying down files.
1058 Accepted values are:
1061 Do not pre-allocate space.
1064 Use a platform's native pre-allocation call but fall back to
1065 **none** behavior if it fails/is not implemented.
1068 Pre-allocate via :manpage:`posix_fallocate(3)`.
1071 Pre-allocate via :manpage:`fallocate(2)` with
1072 FALLOC_FL_KEEP_SIZE set.
1075 Backward-compatible alias for **none**.
1078 Backward-compatible alias for **posix**.
1080 May not be available on all supported platforms. **keep** is only available
1081 on Linux. If using ZFS on Solaris this cannot be set to **posix**
1082 because ZFS doesn't support pre-allocation. Default: **native** if any
1083 pre-allocation methods are available, **none** if not.
1085 .. option:: fadvise_hint=str
1087 Use :manpage:`posix_fadvise(2)` to advise the kernel on what I/O patterns
1088 are likely to be issued. Accepted values are:
1091 Backwards-compatible hint for "no hint".
1094 Backwards compatible hint for "advise with fio workload type". This
1095 uses **FADV_RANDOM** for a random workload, and **FADV_SEQUENTIAL**
1096 for a sequential workload.
1099 Advise using **FADV_SEQUENTIAL**.
1102 Advise using **FADV_RANDOM**.
1104 .. option:: write_hint=str
1106 Use :manpage:`fcntl(2)` to advise the kernel what life time to expect
1107 from a write. Only supported on Linux, as of version 4.13. Accepted
1111 No particular life time associated with this file.
1114 Data written to this file has a short life time.
1117 Data written to this file has a medium life time.
1120 Data written to this file has a long life time.
1123 Data written to this file has a very long life time.
1125 The values are all relative to each other, and no absolute meaning
1126 should be associated with them.
1128 .. option:: offset=int
1130 Start I/O at the provided offset in the file, given as either a fixed size in
1131 bytes or a percentage. If a percentage is given, the generated offset will be
1132 aligned to the minimum ``blocksize`` or to the value of ``offset_align`` if
1133 provided. Data before the given offset will not be touched. This
1134 effectively caps the file size at `real_size - offset`. Can be combined with
1135 :option:`size` to constrain the start and end range of the I/O workload.
1136 A percentage can be specified by a number between 1 and 100 followed by '%',
1137 for example, ``offset=20%`` to specify 20%.
1139 .. option:: offset_align=int
1141 If set to non-zero value, the byte offset generated by a percentage ``offset``
1142 is aligned upwards to this value. Defaults to 0 meaning that a percentage
1143 offset is aligned to the minimum block size.
1145 .. option:: offset_increment=int
1147 If this is provided, then the real offset becomes `offset + offset_increment
1148 * thread_number`, where the thread number is a counter that starts at 0 and
1149 is incremented for each sub-job (i.e. when :option:`numjobs` option is
1150 specified). This option is useful if there are several jobs which are
1151 intended to operate on a file in parallel disjoint segments, with even
1152 spacing between the starting points.
1154 .. option:: number_ios=int
1156 Fio will normally perform I/Os until it has exhausted the size of the region
1157 set by :option:`size`, or if it exhaust the allocated time (or hits an error
1158 condition). With this setting, the range/size can be set independently of
1159 the number of I/Os to perform. When fio reaches this number, it will exit
1160 normally and report status. Note that this does not extend the amount of I/O
1161 that will be done, it will only stop fio if this condition is met before
1162 other end-of-job criteria.
1164 .. option:: fsync=int
1166 If writing to a file, issue an :manpage:`fsync(2)` (or its equivalent) of
1167 the dirty data for every number of blocks given. For example, if you give 32
1168 as a parameter, fio will sync the file after every 32 writes issued. If fio is
1169 using non-buffered I/O, we may not sync the file. The exception is the sg
1170 I/O engine, which synchronizes the disk cache anyway. Defaults to 0, which
1171 means fio does not periodically issue and wait for a sync to complete. Also
1172 see :option:`end_fsync` and :option:`fsync_on_close`.
1174 .. option:: fdatasync=int
1176 Like :option:`fsync` but uses :manpage:`fdatasync(2)` to only sync data and
1177 not metadata blocks. In Windows, FreeBSD, and DragonFlyBSD there is no
1178 :manpage:`fdatasync(2)` so this falls back to using :manpage:`fsync(2)`.
1179 Defaults to 0, which means fio does not periodically issue and wait for a
1180 data-only sync to complete.
1182 .. option:: write_barrier=int
1184 Make every `N-th` write a barrier write.
1186 .. option:: sync_file_range=str:int
1188 Use :manpage:`sync_file_range(2)` for every `int` number of write
1189 operations. Fio will track range of writes that have happened since the last
1190 :manpage:`sync_file_range(2)` call. `str` can currently be one or more of:
1193 SYNC_FILE_RANGE_WAIT_BEFORE
1195 SYNC_FILE_RANGE_WRITE
1197 SYNC_FILE_RANGE_WAIT_AFTER
1199 So if you do ``sync_file_range=wait_before,write:8``, fio would use
1200 ``SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE`` for every 8
1201 writes. Also see the :manpage:`sync_file_range(2)` man page. This option is
1204 .. option:: overwrite=bool
1206 If true, writes to a file will always overwrite existing data. If the file
1207 doesn't already exist, it will be created before the write phase begins. If
1208 the file exists and is large enough for the specified write phase, nothing
1209 will be done. Default: false.
1211 .. option:: end_fsync=bool
1213 If true, :manpage:`fsync(2)` file contents when a write stage has completed.
1216 .. option:: fsync_on_close=bool
1218 If true, fio will :manpage:`fsync(2)` a dirty file on close. This differs
1219 from :option:`end_fsync` in that it will happen on every file close, not
1220 just at the end of the job. Default: false.
1222 .. option:: rwmixread=int
1224 Percentage of a mixed workload that should be reads. Default: 50.
1226 .. option:: rwmixwrite=int
1228 Percentage of a mixed workload that should be writes. If both
1229 :option:`rwmixread` and :option:`rwmixwrite` is given and the values do not
1230 add up to 100%, the latter of the two will be used to override the
1231 first. This may interfere with a given rate setting, if fio is asked to
1232 limit reads or writes to a certain rate. If that is the case, then the
1233 distribution may be skewed. Default: 50.
1235 .. option:: random_distribution=str:float[,str:float][,str:float]
1237 By default, fio will use a completely uniform random distribution when asked
1238 to perform random I/O. Sometimes it is useful to skew the distribution in
1239 specific ways, ensuring that some parts of the data is more hot than others.
1240 fio includes the following distribution models:
1243 Uniform random distribution
1252 Normal (Gaussian) distribution
1255 Zoned random distribution
1258 Zone absolute random distribution
1260 When using a **zipf** or **pareto** distribution, an input value is also
1261 needed to define the access pattern. For **zipf**, this is the `Zipf
1262 theta`. For **pareto**, it's the `Pareto power`. Fio includes a test
1263 program, :command:`fio-genzipf`, that can be used visualize what the given input
1264 values will yield in terms of hit rates. If you wanted to use **zipf** with
1265 a `theta` of 1.2, you would use ``random_distribution=zipf:1.2`` as the
1266 option. If a non-uniform model is used, fio will disable use of the random
1267 map. For the **normal** distribution, a normal (Gaussian) deviation is
1268 supplied as a value between 0 and 100.
1270 For a **zoned** distribution, fio supports specifying percentages of I/O
1271 access that should fall within what range of the file or device. For
1272 example, given a criteria of:
1274 * 60% of accesses should be to the first 10%
1275 * 30% of accesses should be to the next 20%
1276 * 8% of accesses should be to the next 30%
1277 * 2% of accesses should be to the next 40%
1279 we can define that through zoning of the random accesses. For the above
1280 example, the user would do::
1282 random_distribution=zoned:60/10:30/20:8/30:2/40
1284 A **zoned_abs** distribution works exactly like the **zoned**, except
1285 that it takes absolute sizes. For example, let's say you wanted to
1286 define access according to the following criteria:
1288 * 60% of accesses should be to the first 20G
1289 * 30% of accesses should be to the next 100G
1290 * 10% of accesses should be to the next 500G
1292 we can define an absolute zoning distribution with:
1294 random_distribution=zoned_abs=60/20G:30/100G:10/500g
1296 Similarly to how :option:`bssplit` works for setting ranges and
1297 percentages of block sizes. Like :option:`bssplit`, it's possible to
1298 specify separate zones for reads, writes, and trims. If just one set
1299 is given, it'll apply to all of them. This goes for both **zoned**
1300 **zoned_abs** distributions.
1302 .. option:: percentage_random=int[,int][,int]
1304 For a random workload, set how big a percentage should be random. This
1305 defaults to 100%, in which case the workload is fully random. It can be set
1306 from anywhere from 0 to 100. Setting it to 0 would make the workload fully
1307 sequential. Any setting in between will result in a random mix of sequential
1308 and random I/O, at the given percentages. Comma-separated values may be
1309 specified for reads, writes, and trims as described in :option:`blocksize`.
1311 .. option:: norandommap
1313 Normally fio will cover every block of the file when doing random I/O. If
1314 this option is given, fio will just get a new random offset without looking
1315 at past I/O history. This means that some blocks may not be read or written,
1316 and that some blocks may be read/written more than once. If this option is
1317 used with :option:`verify` and multiple blocksizes (via :option:`bsrange`),
1318 only intact blocks are verified, i.e., partially-overwritten blocks are
1321 .. option:: softrandommap=bool
1323 See :option:`norandommap`. If fio runs with the random block map enabled and
1324 it fails to allocate the map, if this option is set it will continue without
1325 a random block map. As coverage will not be as complete as with random maps,
1326 this option is disabled by default.
1328 .. option:: random_generator=str
1330 Fio supports the following engines for generating I/O offsets for random I/O:
1333 Strong 2^88 cycle random number generator.
1335 Linear feedback shift register generator.
1337 Strong 64-bit 2^258 cycle random number generator.
1339 **tausworthe** is a strong random number generator, but it requires tracking
1340 on the side if we want to ensure that blocks are only read or written
1341 once. **lfsr** guarantees that we never generate the same offset twice, and
1342 it's also less computationally expensive. It's not a true random generator,
1343 however, though for I/O purposes it's typically good enough. **lfsr** only
1344 works with single block sizes, not with workloads that use multiple block
1345 sizes. If used with such a workload, fio may read or write some blocks
1346 multiple times. The default value is **tausworthe**, unless the required
1347 space exceeds 2^32 blocks. If it does, then **tausworthe64** is
1348 selected automatically.
1354 .. option:: blocksize=int[,int][,int], bs=int[,int][,int]
1356 The block size in bytes used for I/O units. Default: 4096. A single value
1357 applies to reads, writes, and trims. Comma-separated values may be
1358 specified for reads, writes, and trims. A value not terminated in a comma
1359 applies to subsequent types.
1364 means 256k for reads, writes and trims.
1367 means 8k for reads, 32k for writes and trims.
1370 means 8k for reads, 32k for writes, and default for trims.
1373 means default for reads, 8k for writes and trims.
1376 means default for reads, 8k for writes, and default for trims.
1378 .. option:: blocksize_range=irange[,irange][,irange], bsrange=irange[,irange][,irange]
1380 A range of block sizes in bytes for I/O units. The issued I/O unit will
1381 always be a multiple of the minimum size, unless
1382 :option:`blocksize_unaligned` is set.
1384 Comma-separated ranges may be specified for reads, writes, and trims as
1385 described in :option:`blocksize`.
1387 Example: ``bsrange=1k-4k,2k-8k``.
1389 .. option:: bssplit=str[,str][,str]
1391 Sometimes you want even finer grained control of the block sizes issued, not
1392 just an even split between them. This option allows you to weight various
1393 block sizes, so that you are able to define a specific amount of block sizes
1394 issued. The format for this option is::
1396 bssplit=blocksize/percentage:blocksize/percentage
1398 for as many block sizes as needed. So if you want to define a workload that
1399 has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would write::
1401 bssplit=4k/10:64k/50:32k/40
1403 Ordering does not matter. If the percentage is left blank, fio will fill in
1404 the remaining values evenly. So a bssplit option like this one::
1406 bssplit=4k/50:1k/:32k/
1408 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always add up
1409 to 100, if bssplit is given a range that adds up to more, it will error out.
1411 Comma-separated values may be specified for reads, writes, and trims as
1412 described in :option:`blocksize`.
1414 If you want a workload that has 50% 2k reads and 50% 4k reads, while having
1415 90% 4k writes and 10% 8k writes, you would specify::
1417 bssplit=2k/50:4k/50,4k/90,8k/10
1419 .. option:: blocksize_unaligned, bs_unaligned
1421 If set, fio will issue I/O units with any size within
1422 :option:`blocksize_range`, not just multiples of the minimum size. This
1423 typically won't work with direct I/O, as that normally requires sector
1426 .. option:: bs_is_seq_rand=bool
1428 If this option is set, fio will use the normal read,write blocksize settings
1429 as sequential,random blocksize settings instead. Any random read or write
1430 will use the WRITE blocksize settings, and any sequential read or write will
1431 use the READ blocksize settings.
1433 .. option:: blockalign=int[,int][,int], ba=int[,int][,int]
1435 Boundary to which fio will align random I/O units. Default:
1436 :option:`blocksize`. Minimum alignment is typically 512b for using direct
1437 I/O, though it usually depends on the hardware block size. This option is
1438 mutually exclusive with using a random map for files, so it will turn off
1439 that option. Comma-separated values may be specified for reads, writes, and
1440 trims as described in :option:`blocksize`.
1446 .. option:: zero_buffers
1448 Initialize buffers with all zeros. Default: fill buffers with random data.
1450 .. option:: refill_buffers
1452 If this option is given, fio will refill the I/O buffers on every
1453 submit. Only makes sense if :option:`zero_buffers` isn't specified,
1454 naturally. Defaults to being unset i.e., the buffer is only filled at
1455 init time and the data in it is reused when possible but if any of
1456 :option:`verify`, :option:`buffer_compress_percentage` or
1457 :option:`dedupe_percentage` are enabled then `refill_buffers` is also
1458 automatically enabled.
1460 .. option:: scramble_buffers=bool
1462 If :option:`refill_buffers` is too costly and the target is using data
1463 deduplication, then setting this option will slightly modify the I/O buffer
1464 contents to defeat normal de-dupe attempts. This is not enough to defeat
1465 more clever block compression attempts, but it will stop naive dedupe of
1466 blocks. Default: true.
1468 .. option:: buffer_compress_percentage=int
1470 If this is set, then fio will attempt to provide I/O buffer content
1471 (on WRITEs) that compresses to the specified level. Fio does this by
1472 providing a mix of random data followed by fixed pattern data. The
1473 fixed pattern is either zeros, or the pattern specified by
1474 :option:`buffer_pattern`. If the `buffer_pattern` option is used, it
1475 might skew the compression ratio slightly. Setting
1476 `buffer_compress_percentage` to a value other than 100 will also
1477 enable :option:`refill_buffers` in order to reduce the likelihood that
1478 adjacent blocks are so similar that they over compress when seen
1479 together. See :option:`buffer_compress_chunk` for how to set a finer or
1480 coarser granularity for the random/fixed data region. Defaults to unset
1481 i.e., buffer data will not adhere to any compression level.
1483 .. option:: buffer_compress_chunk=int
1485 This setting allows fio to manage how big the random/fixed data region
1486 is when using :option:`buffer_compress_percentage`. When
1487 `buffer_compress_chunk` is set to some non-zero value smaller than the
1488 block size, fio can repeat the random/fixed region throughout the I/O
1489 buffer at the specified interval (which particularly useful when
1490 bigger block sizes are used for a job). When set to 0, fio will use a
1491 chunk size that matches the block size resulting in a single
1492 random/fixed region within the I/O buffer. Defaults to 512. When the
1493 unit is omitted, the value is interpreted in bytes.
1495 .. option:: buffer_pattern=str
1497 If set, fio will fill the I/O buffers with this pattern or with the contents
1498 of a file. If not set, the contents of I/O buffers are defined by the other
1499 options related to buffer contents. The setting can be any pattern of bytes,
1500 and can be prefixed with 0x for hex values. It may also be a string, where
1501 the string must then be wrapped with ``""``. Or it may also be a filename,
1502 where the filename must be wrapped with ``''`` in which case the file is
1503 opened and read. Note that not all the file contents will be read if that
1504 would cause the buffers to overflow. So, for example::
1506 buffer_pattern='filename'
1510 buffer_pattern="abcd"
1518 buffer_pattern=0xdeadface
1520 Also you can combine everything together in any order::
1522 buffer_pattern=0xdeadface"abcd"-12'filename'
1524 .. option:: dedupe_percentage=int
1526 If set, fio will generate this percentage of identical buffers when
1527 writing. These buffers will be naturally dedupable. The contents of the
1528 buffers depend on what other buffer compression settings have been set. It's
1529 possible to have the individual buffers either fully compressible, or not at
1530 all -- this option only controls the distribution of unique buffers. Setting
1531 this option will also enable :option:`refill_buffers` to prevent every buffer
1534 .. option:: invalidate=bool
1536 Invalidate the buffer/page cache parts of the files to be used prior to
1537 starting I/O if the platform and file type support it. Defaults to true.
1538 This will be ignored if :option:`pre_read` is also specified for the
1541 .. option:: sync=bool
1543 Use synchronous I/O for buffered writes. For the majority of I/O engines,
1544 this means using O_SYNC. Default: false.
1546 .. option:: iomem=str, mem=str
1548 Fio can use various types of memory as the I/O unit buffer. The allowed
1552 Use memory from :manpage:`malloc(3)` as the buffers. Default memory
1556 Use shared memory as the buffers. Allocated through
1557 :manpage:`shmget(2)`.
1560 Same as shm, but use huge pages as backing.
1563 Use :manpage:`mmap(2)` to allocate buffers. May either be anonymous memory, or can
1564 be file backed if a filename is given after the option. The format
1565 is `mem=mmap:/path/to/file`.
1568 Use a memory mapped huge file as the buffer backing. Append filename
1569 after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file`.
1572 Same as mmap, but use a MMAP_SHARED mapping.
1575 Use GPU memory as the buffers for GPUDirect RDMA benchmark.
1576 The :option:`ioengine` must be `rdma`.
1578 The area allocated is a function of the maximum allowed bs size for the job,
1579 multiplied by the I/O depth given. Note that for **shmhuge** and
1580 **mmaphuge** to work, the system must have free huge pages allocated. This
1581 can normally be checked and set by reading/writing
1582 :file:`/proc/sys/vm/nr_hugepages` on a Linux system. Fio assumes a huge page
1583 is 4MiB in size. So to calculate the number of huge pages you need for a
1584 given job file, add up the I/O depth of all jobs (normally one unless
1585 :option:`iodepth` is used) and multiply by the maximum bs set. Then divide
1586 that number by the huge page size. You can see the size of the huge pages in
1587 :file:`/proc/meminfo`. If no huge pages are allocated by having a non-zero
1588 number in `nr_hugepages`, using **mmaphuge** or **shmhuge** will fail. Also
1589 see :option:`hugepage-size`.
1591 **mmaphuge** also needs to have hugetlbfs mounted and the file location
1592 should point there. So if it's mounted in :file:`/huge`, you would use
1593 `mem=mmaphuge:/huge/somefile`.
1595 .. option:: iomem_align=int, mem_align=int
1597 This indicates the memory alignment of the I/O memory buffers. Note that
1598 the given alignment is applied to the first I/O unit buffer, if using
1599 :option:`iodepth` the alignment of the following buffers are given by the
1600 :option:`bs` used. In other words, if using a :option:`bs` that is a
1601 multiple of the page sized in the system, all buffers will be aligned to
1602 this value. If using a :option:`bs` that is not page aligned, the alignment
1603 of subsequent I/O memory buffers is the sum of the :option:`iomem_align` and
1606 .. option:: hugepage-size=int
1608 Defines the size of a huge page. Must at least be equal to the system
1609 setting, see :file:`/proc/meminfo`. Defaults to 4MiB. Should probably
1610 always be a multiple of megabytes, so using ``hugepage-size=Xm`` is the
1611 preferred way to set this to avoid setting a non-pow-2 bad value.
1613 .. option:: lockmem=int
1615 Pin the specified amount of memory with :manpage:`mlock(2)`. Can be used to
1616 simulate a smaller amount of memory. The amount specified is per worker.
1622 .. option:: size=int
1624 The total size of file I/O for each thread of this job. Fio will run until
1625 this many bytes has been transferred, unless runtime is limited by other options
1626 (such as :option:`runtime`, for instance, or increased/decreased by :option:`io_size`).
1627 Fio will divide this size between the available files determined by options
1628 such as :option:`nrfiles`, :option:`filename`, unless :option:`filesize` is
1629 specified by the job. If the result of division happens to be 0, the size is
1630 set to the physical size of the given files or devices if they exist.
1631 If this option is not specified, fio will use the full size of the given
1632 files or devices. If the files do not exist, size must be given. It is also
1633 possible to give size as a percentage between 1 and 100. If ``size=20%`` is
1634 given, fio will use 20% of the full size of the given files or devices.
1635 Can be combined with :option:`offset` to constrain the start and end range
1636 that I/O will be done within.
1638 .. option:: io_size=int, io_limit=int
1640 Normally fio operates within the region set by :option:`size`, which means
1641 that the :option:`size` option sets both the region and size of I/O to be
1642 performed. Sometimes that is not what you want. With this option, it is
1643 possible to define just the amount of I/O that fio should do. For instance,
1644 if :option:`size` is set to 20GiB and :option:`io_size` is set to 5GiB, fio
1645 will perform I/O within the first 20GiB but exit when 5GiB have been
1646 done. The opposite is also possible -- if :option:`size` is set to 20GiB,
1647 and :option:`io_size` is set to 40GiB, then fio will do 40GiB of I/O within
1648 the 0..20GiB region.
1650 .. option:: filesize=irange(int)
1652 Individual file sizes. May be a range, in which case fio will select sizes
1653 for files at random within the given range and limited to :option:`size` in
1654 total (if that is given). If not given, each created file is the same size.
1655 This option overrides :option:`size` in terms of file size, which means
1656 this value is used as a fixed size or possible range of each file.
1658 .. option:: file_append=bool
1660 Perform I/O after the end of the file. Normally fio will operate within the
1661 size of a file. If this option is set, then fio will append to the file
1662 instead. This has identical behavior to setting :option:`offset` to the size
1663 of a file. This option is ignored on non-regular files.
1665 .. option:: fill_device=bool, fill_fs=bool
1667 Sets size to something really large and waits for ENOSPC (no space left on
1668 device) as the terminating condition. Only makes sense with sequential
1669 write. For a read workload, the mount point will be filled first then I/O
1670 started on the result. This option doesn't make sense if operating on a raw
1671 device node, since the size of that is already known by the file system.
1672 Additionally, writing beyond end-of-device will not return ENOSPC there.
1678 .. option:: ioengine=str
1680 Defines how the job issues I/O to the file. The following types are defined:
1683 Basic :manpage:`read(2)` or :manpage:`write(2)`
1684 I/O. :manpage:`lseek(2)` is used to position the I/O location.
1685 See :option:`fsync` and :option:`fdatasync` for syncing write I/Os.
1688 Basic :manpage:`pread(2)` or :manpage:`pwrite(2)` I/O. Default on
1689 all supported operating systems except for Windows.
1692 Basic :manpage:`readv(2)` or :manpage:`writev(2)` I/O. Will emulate
1693 queuing by coalescing adjacent I/Os into a single submission.
1696 Basic :manpage:`preadv(2)` or :manpage:`pwritev(2)` I/O.
1699 Basic :manpage:`preadv2(2)` or :manpage:`pwritev2(2)` I/O.
1702 Linux native asynchronous I/O. Note that Linux may only support
1703 queued behavior with non-buffered I/O (set ``direct=1`` or
1705 This engine defines engine specific options.
1708 POSIX asynchronous I/O using :manpage:`aio_read(3)` and
1709 :manpage:`aio_write(3)`.
1712 Solaris native asynchronous I/O.
1715 Windows native asynchronous I/O. Default on Windows.
1718 File is memory mapped with :manpage:`mmap(2)` and data copied
1719 to/from using :manpage:`memcpy(3)`.
1722 :manpage:`splice(2)` is used to transfer the data and
1723 :manpage:`vmsplice(2)` to transfer data from user space to the
1727 SCSI generic sg v3 I/O. May either be synchronous using the SG_IO
1728 ioctl, or if the target is an sg character device we use
1729 :manpage:`read(2)` and :manpage:`write(2)` for asynchronous
1730 I/O. Requires :option:`filename` option to specify either block or
1734 Doesn't transfer any data, just pretends to. This is mainly used to
1735 exercise fio itself and for debugging/testing purposes.
1738 Transfer over the network to given ``host:port``. Depending on the
1739 :option:`protocol` used, the :option:`hostname`, :option:`port`,
1740 :option:`listen` and :option:`filename` options are used to specify
1741 what sort of connection to make, while the :option:`protocol` option
1742 determines which protocol will be used. This engine defines engine
1746 Like **net**, but uses :manpage:`splice(2)` and
1747 :manpage:`vmsplice(2)` to map data and send/receive.
1748 This engine defines engine specific options.
1751 Doesn't transfer any data, but burns CPU cycles according to the
1752 :option:`cpuload` and :option:`cpuchunks` options. Setting
1753 :option:`cpuload`\=85 will cause that job to do nothing but burn 85%
1754 of the CPU. In case of SMP machines, use :option:`numjobs`\=<nr_of_cpu>
1755 to get desired CPU usage, as the cpuload only loads a
1756 single CPU at the desired rate. A job never finishes unless there is
1757 at least one non-cpuio job.
1760 The GUASI I/O engine is the Generic Userspace Asyncronous Syscall
1761 Interface approach to async I/O. See
1763 http://www.xmailserver.org/guasi-lib.html
1765 for more info on GUASI.
1768 The RDMA I/O engine supports both RDMA memory semantics
1769 (RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the
1770 InfiniBand, RoCE and iWARP protocols. This engine defines engine
1774 I/O engine that does regular fallocate to simulate data transfer as
1778 does fallocate(,mode = FALLOC_FL_KEEP_SIZE,).
1781 does fallocate(,mode = 0).
1784 does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE).
1787 I/O engine that sends :manpage:`ftruncate(2)` operations in response
1788 to write (DDIR_WRITE) events. Each ftruncate issued sets the file's
1789 size to the current block offset. :option:`blocksize` is ignored.
1792 I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate
1793 defragment activity in request to DDIR_WRITE event.
1796 I/O engine supporting direct access to Ceph Rados Block Devices
1797 (RBD) via librbd without the need to use the kernel rbd driver. This
1798 ioengine defines engine specific options.
1801 Using GlusterFS libgfapi sync interface to direct access to
1802 GlusterFS volumes without having to go through FUSE. This ioengine
1803 defines engine specific options.
1806 Using GlusterFS libgfapi async interface to direct access to
1807 GlusterFS volumes without having to go through FUSE. This ioengine
1808 defines engine specific options.
1811 Read and write through Hadoop (HDFS). The :option:`filename` option
1812 is used to specify host,port of the hdfs name-node to connect. This
1813 engine interprets offsets a little differently. In HDFS, files once
1814 created cannot be modified so random writes are not possible. To
1815 imitate this the libhdfs engine expects a bunch of small files to be
1816 created over HDFS and will randomly pick a file from them
1817 based on the offset generated by fio backend (see the example
1818 job file to create such files, use ``rw=write`` option). Please
1819 note, it may be necessary to set environment variables to work
1820 with HDFS/libhdfs properly. Each job uses its own connection to
1824 Read, write and erase an MTD character device (e.g.,
1825 :file:`/dev/mtd0`). Discards are treated as erases. Depending on the
1826 underlying device type, the I/O may have to go in a certain pattern,
1827 e.g., on NAND, writing sequentially to erase blocks and discarding
1828 before overwriting. The `trimwrite` mode works well for this
1832 Read and write using filesystem DAX to a file on a filesystem
1833 mounted with DAX on a persistent memory device through the NVML
1837 Read and write using device DAX to a persistent memory device (e.g.,
1838 /dev/dax0.0) through the NVML libpmem library.
1841 Prefix to specify loading an external I/O engine object file. Append
1842 the engine filename, e.g. ``ioengine=external:/tmp/foo.o`` to load
1843 ioengine :file:`foo.o` in :file:`/tmp`. The path can be either
1844 absolute or relative. See :file:`engines/skeleton_external.c` for
1845 details of writing an external I/O engine.
1848 Simply create the files and do no I/O to them. You still need to
1849 set `filesize` so that all the accounting still occurs, but no
1850 actual I/O will be done other than creating the file.
1853 Read and write using mmap I/O to a file on a filesystem
1854 mounted with DAX on a persistent memory device through the NVML
1857 I/O engine specific parameters
1858 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1860 In addition, there are some parameters which are only valid when a specific
1861 :option:`ioengine` is in use. These are used identically to normal parameters,
1862 with the caveat that when used on the command line, they must come after the
1863 :option:`ioengine` that defines them is selected.
1865 .. option:: userspace_reap : [libaio]
1867 Normally, with the libaio engine in use, fio will use the
1868 :manpage:`io_getevents(2)` system call to reap newly returned events. With
1869 this flag turned on, the AIO ring will be read directly from user-space to
1870 reap events. The reaping mode is only enabled when polling for a minimum of
1871 0 events (e.g. when :option:`iodepth_batch_complete` `=0`).
1873 .. option:: hipri : [pvsync2]
1875 Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority
1878 .. option:: hipri_percentage : [pvsync2]
1880 When hipri is set this determines the probability of a pvsync2 I/O being high
1881 priority. The default is 100%.
1883 .. option:: cpuload=int : [cpuio]
1885 Attempt to use the specified percentage of CPU cycles. This is a mandatory
1886 option when using cpuio I/O engine.
1888 .. option:: cpuchunks=int : [cpuio]
1890 Split the load into cycles of the given time. In microseconds.
1892 .. option:: exit_on_io_done=bool : [cpuio]
1894 Detect when I/O threads are done, then exit.
1896 .. option:: namenode=str : [libhdfs]
1898 The hostname or IP address of a HDFS cluster namenode to contact.
1900 .. option:: port=int
1904 The listening port of the HFDS cluster namenode.
1908 The TCP or UDP port to bind to or connect to. If this is used with
1909 :option:`numjobs` to spawn multiple instances of the same job type, then
1910 this will be the starting port number since fio will use a range of
1915 The port to use for RDMA-CM communication. This should be the same value
1916 on the client and the server side.
1918 .. option:: hostname=str : [netsplice] [net] [rdma]
1920 The hostname or IP address to use for TCP, UDP or RDMA-CM based I/O. If the job
1921 is a TCP listener or UDP reader, the hostname is not used and must be omitted
1922 unless it is a valid UDP multicast address.
1924 .. option:: interface=str : [netsplice] [net]
1926 The IP address of the network interface used to send or receive UDP
1929 .. option:: ttl=int : [netsplice] [net]
1931 Time-to-live value for outgoing UDP multicast packets. Default: 1.
1933 .. option:: nodelay=bool : [netsplice] [net]
1935 Set TCP_NODELAY on TCP connections.
1937 .. option:: protocol=str, proto=str : [netsplice] [net]
1939 The network protocol to use. Accepted values are:
1942 Transmission control protocol.
1944 Transmission control protocol V6.
1946 User datagram protocol.
1948 User datagram protocol V6.
1952 When the protocol is TCP or UDP, the port must also be given, as well as the
1953 hostname if the job is a TCP listener or UDP reader. For unix sockets, the
1954 normal :option:`filename` option should be used and the port is invalid.
1956 .. option:: listen : [netsplice] [net]
1958 For TCP network connections, tell fio to listen for incoming connections
1959 rather than initiating an outgoing connection. The :option:`hostname` must
1960 be omitted if this option is used.
1962 .. option:: pingpong : [netsplice] [net]
1964 Normally a network writer will just continue writing data, and a network
1965 reader will just consume packages. If ``pingpong=1`` is set, a writer will
1966 send its normal payload to the reader, then wait for the reader to send the
1967 same payload back. This allows fio to measure network latencies. The
1968 submission and completion latencies then measure local time spent sending or
1969 receiving, and the completion latency measures how long it took for the
1970 other end to receive and send back. For UDP multicast traffic
1971 ``pingpong=1`` should only be set for a single reader when multiple readers
1972 are listening to the same address.
1974 .. option:: window_size : [netsplice] [net]
1976 Set the desired socket buffer size for the connection.
1978 .. option:: mss : [netsplice] [net]
1980 Set the TCP maximum segment size (TCP_MAXSEG).
1982 .. option:: donorname=str : [e4defrag]
1984 File will be used as a block donor (swap extents between files).
1986 .. option:: inplace=int : [e4defrag]
1988 Configure donor file blocks allocation strategy:
1991 Default. Preallocate donor's file on init.
1993 Allocate space immediately inside defragment event, and free right
1996 .. option:: clustername=str : [rbd]
1998 Specifies the name of the Ceph cluster.
2000 .. option:: rbdname=str : [rbd]
2002 Specifies the name of the RBD.
2004 .. option:: pool=str : [rbd]
2006 Specifies the name of the Ceph pool containing RBD.
2008 .. option:: clientname=str : [rbd]
2010 Specifies the username (without the 'client.' prefix) used to access the
2011 Ceph cluster. If the *clustername* is specified, the *clientname* shall be
2012 the full *type.id* string. If no type. prefix is given, fio will add
2013 'client.' by default.
2015 .. option:: skip_bad=bool : [mtd]
2017 Skip operations against known bad blocks.
2019 .. option:: hdfsdirectory : [libhdfs]
2021 libhdfs will create chunk in this HDFS directory.
2023 .. option:: chunk_size : [libhdfs]
2025 The size of the chunk to use for each file.
2027 .. option:: verb=str : [rdma]
2029 The RDMA verb to use on this side of the RDMA ioengine connection. Valid
2030 values are write, read, send and recv. These correspond to the equivalent
2031 RDMA verbs (e.g. write = rdma_write etc.). Note that this only needs to be
2032 specified on the client side of the connection. See the examples folder.
2034 .. option:: bindname=str : [rdma]
2036 The name to use to bind the local RDMA-CM connection to a local RDMA device.
2037 This could be a hostname or an IPv4 or IPv6 address. On the server side this
2038 will be passed into the rdma_bind_addr() function and on the client site it
2039 will be used in the rdma_resolve_add() function. This can be useful when
2040 multiple paths exist between the client and the server or in certain loopback
2046 .. option:: iodepth=int
2048 Number of I/O units to keep in flight against the file. Note that
2049 increasing *iodepth* beyond 1 will not affect synchronous ioengines (except
2050 for small degrees when :option:`verify_async` is in use). Even async
2051 engines may impose OS restrictions causing the desired depth not to be
2052 achieved. This may happen on Linux when using libaio and not setting
2053 :option:`direct`\=1, since buffered I/O is not async on that OS. Keep an
2054 eye on the I/O depth distribution in the fio output to verify that the
2055 achieved depth is as expected. Default: 1.
2057 .. option:: iodepth_batch_submit=int, iodepth_batch=int
2059 This defines how many pieces of I/O to submit at once. It defaults to 1
2060 which means that we submit each I/O as soon as it is available, but can be
2061 raised to submit bigger batches of I/O at the time. If it is set to 0 the
2062 :option:`iodepth` value will be used.
2064 .. option:: iodepth_batch_complete_min=int, iodepth_batch_complete=int
2066 This defines how many pieces of I/O to retrieve at once. It defaults to 1
2067 which means that we'll ask for a minimum of 1 I/O in the retrieval process
2068 from the kernel. The I/O retrieval will go on until we hit the limit set by
2069 :option:`iodepth_low`. If this variable is set to 0, then fio will always
2070 check for completed events before queuing more I/O. This helps reduce I/O
2071 latency, at the cost of more retrieval system calls.
2073 .. option:: iodepth_batch_complete_max=int
2075 This defines maximum pieces of I/O to retrieve at once. This variable should
2076 be used along with :option:`iodepth_batch_complete_min`\=int variable,
2077 specifying the range of min and max amount of I/O which should be
2078 retrieved. By default it is equal to the :option:`iodepth_batch_complete_min`
2083 iodepth_batch_complete_min=1
2084 iodepth_batch_complete_max=<iodepth>
2086 which means that we will retrieve at least 1 I/O and up to the whole
2087 submitted queue depth. If none of I/O has been completed yet, we will wait.
2091 iodepth_batch_complete_min=0
2092 iodepth_batch_complete_max=<iodepth>
2094 which means that we can retrieve up to the whole submitted queue depth, but
2095 if none of I/O has been completed yet, we will NOT wait and immediately exit
2096 the system call. In this example we simply do polling.
2098 .. option:: iodepth_low=int
2100 The low water mark indicating when to start filling the queue
2101 again. Defaults to the same as :option:`iodepth`, meaning that fio will
2102 attempt to keep the queue full at all times. If :option:`iodepth` is set to
2103 e.g. 16 and *iodepth_low* is set to 4, then after fio has filled the queue of
2104 16 requests, it will let the depth drain down to 4 before starting to fill
2107 .. option:: serialize_overlap=bool
2109 Serialize in-flight I/Os that might otherwise cause or suffer from data races.
2110 When two or more I/Os are submitted simultaneously, there is no guarantee that
2111 the I/Os will be processed or completed in the submitted order. Further, if
2112 two or more of those I/Os are writes, any overlapping region between them can
2113 become indeterminate/undefined on certain storage. These issues can cause
2114 verification to fail erratically when at least one of the racing I/Os is
2115 changing data and the overlapping region has a non-zero size. Setting
2116 ``serialize_overlap`` tells fio to avoid provoking this behavior by explicitly
2117 serializing in-flight I/Os that have a non-zero overlap. Note that setting
2118 this option can reduce both performance and the :option:`iodepth` achieved.
2119 Additionally this option does not work when :option:`io_submit_mode` is set to
2120 offload. Default: false.
2122 .. option:: io_submit_mode=str
2124 This option controls how fio submits the I/O to the I/O engine. The default
2125 is `inline`, which means that the fio job threads submit and reap I/O
2126 directly. If set to `offload`, the job threads will offload I/O submission
2127 to a dedicated pool of I/O threads. This requires some coordination and thus
2128 has a bit of extra overhead, especially for lower queue depth I/O where it
2129 can increase latencies. The benefit is that fio can manage submission rates
2130 independently of the device completion rates. This avoids skewed latency
2131 reporting if I/O gets backed up on the device side (the coordinated omission
2138 .. option:: thinktime=time
2140 Stall the job for the specified period of time after an I/O has completed before issuing the
2141 next. May be used to simulate processing being done by an application.
2142 When the unit is omitted, the value is interpreted in microseconds. See
2143 :option:`thinktime_blocks` and :option:`thinktime_spin`.
2145 .. option:: thinktime_spin=time
2147 Only valid if :option:`thinktime` is set - pretend to spend CPU time doing
2148 something with the data received, before falling back to sleeping for the
2149 rest of the period specified by :option:`thinktime`. When the unit is
2150 omitted, the value is interpreted in microseconds.
2152 .. option:: thinktime_blocks=int
2154 Only valid if :option:`thinktime` is set - control how many blocks to issue,
2155 before waiting :option:`thinktime` usecs. If not set, defaults to 1 which will make
2156 fio wait :option:`thinktime` usecs after every block. This effectively makes any
2157 queue depth setting redundant, since no more than 1 I/O will be queued
2158 before we have to complete it and do our :option:`thinktime`. In other words, this
2159 setting effectively caps the queue depth if the latter is larger.
2161 .. option:: rate=int[,int][,int]
2163 Cap the bandwidth used by this job. The number is in bytes/sec, the normal
2164 suffix rules apply. Comma-separated values may be specified for reads,
2165 writes, and trims as described in :option:`blocksize`.
2167 For example, using `rate=1m,500k` would limit reads to 1MiB/sec and writes to
2168 500KiB/sec. Capping only reads or writes can be done with `rate=,500k` or
2169 `rate=500k,` where the former will only limit writes (to 500KiB/sec) and the
2170 latter will only limit reads.
2172 .. option:: rate_min=int[,int][,int]
2174 Tell fio to do whatever it can to maintain at least this bandwidth. Failing
2175 to meet this requirement will cause the job to exit. Comma-separated values
2176 may be specified for reads, writes, and trims as described in
2177 :option:`blocksize`.
2179 .. option:: rate_iops=int[,int][,int]
2181 Cap the bandwidth to this number of IOPS. Basically the same as
2182 :option:`rate`, just specified independently of bandwidth. If the job is
2183 given a block size range instead of a fixed value, the smallest block size
2184 is used as the metric. Comma-separated values may be specified for reads,
2185 writes, and trims as described in :option:`blocksize`.
2187 .. option:: rate_iops_min=int[,int][,int]
2189 If fio doesn't meet this rate of I/O, it will cause the job to exit.
2190 Comma-separated values may be specified for reads, writes, and trims as
2191 described in :option:`blocksize`.
2193 .. option:: rate_process=str
2195 This option controls how fio manages rated I/O submissions. The default is
2196 `linear`, which submits I/O in a linear fashion with fixed delays between
2197 I/Os that gets adjusted based on I/O completion rates. If this is set to
2198 `poisson`, fio will submit I/O based on a more real world random request
2199 flow, known as the Poisson process
2200 (https://en.wikipedia.org/wiki/Poisson_point_process). The lambda will be
2201 10^6 / IOPS for the given workload.
2207 .. option:: latency_target=time
2209 If set, fio will attempt to find the max performance point that the given
2210 workload will run at while maintaining a latency below this target. When
2211 the unit is omitted, the value is interpreted in microseconds. See
2212 :option:`latency_window` and :option:`latency_percentile`.
2214 .. option:: latency_window=time
2216 Used with :option:`latency_target` to specify the sample window that the job
2217 is run at varying queue depths to test the performance. When the unit is
2218 omitted, the value is interpreted in microseconds.
2220 .. option:: latency_percentile=float
2222 The percentage of I/Os that must fall within the criteria specified by
2223 :option:`latency_target` and :option:`latency_window`. If not set, this
2224 defaults to 100.0, meaning that all I/Os must be equal or below to the value
2225 set by :option:`latency_target`.
2227 .. option:: max_latency=time
2229 If set, fio will exit the job with an ETIMEDOUT error if it exceeds this
2230 maximum latency. When the unit is omitted, the value is interpreted in
2233 .. option:: rate_cycle=int
2235 Average bandwidth for :option:`rate` and :option:`rate_min` over this number
2236 of milliseconds. Defaults to 1000.
2242 .. option:: write_iolog=str
2244 Write the issued I/O patterns to the specified file. See
2245 :option:`read_iolog`. Specify a separate file for each job, otherwise the
2246 iologs will be interspersed and the file may be corrupt.
2248 .. option:: read_iolog=str
2250 Open an iolog with the specified filename and replay the I/O patterns it
2251 contains. This can be used to store a workload and replay it sometime
2252 later. The iolog given may also be a blktrace binary file, which allows fio
2253 to replay a workload captured by :command:`blktrace`. See
2254 :manpage:`blktrace(8)` for how to capture such logging data. For blktrace
2255 replay, the file needs to be turned into a blkparse binary data file first
2256 (``blkparse <device> -o /dev/null -d file_for_fio.bin``).
2258 .. option:: replay_no_stall=bool
2260 When replaying I/O with :option:`read_iolog` the default behavior is to
2261 attempt to respect the timestamps within the log and replay them with the
2262 appropriate delay between IOPS. By setting this variable fio will not
2263 respect the timestamps and attempt to replay them as fast as possible while
2264 still respecting ordering. The result is the same I/O pattern to a given
2265 device, but different timings.
2267 .. option:: replay_redirect=str
2269 While replaying I/O patterns using :option:`read_iolog` the default behavior
2270 is to replay the IOPS onto the major/minor device that each IOP was recorded
2271 from. This is sometimes undesirable because on a different machine those
2272 major/minor numbers can map to a different device. Changing hardware on the
2273 same system can also result in a different major/minor mapping.
2274 ``replay_redirect`` causes all I/Os to be replayed onto the single specified
2275 device regardless of the device it was recorded
2276 from. i.e. :option:`replay_redirect`\= :file:`/dev/sdc` would cause all I/O
2277 in the blktrace or iolog to be replayed onto :file:`/dev/sdc`. This means
2278 multiple devices will be replayed onto a single device, if the trace
2279 contains multiple devices. If you want multiple devices to be replayed
2280 concurrently to multiple redirected devices you must blkparse your trace
2281 into separate traces and replay them with independent fio invocations.
2282 Unfortunately this also breaks the strict time ordering between multiple
2285 .. option:: replay_align=int
2287 Force alignment of I/O offsets and lengths in a trace to this power of 2
2290 .. option:: replay_scale=int
2292 Scale sector offsets down by this factor when replaying traces.
2295 Threads, processes and job synchronization
2296 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2300 Fio defaults to creating jobs by using fork, however if this option is
2301 given, fio will create jobs by using POSIX Threads' function
2302 :manpage:`pthread_create(3)` to create threads instead.
2304 .. option:: wait_for=str
2306 If set, the current job won't be started until all workers of the specified
2307 waitee job are done.
2309 ``wait_for`` operates on the job name basis, so there are a few
2310 limitations. First, the waitee must be defined prior to the waiter job
2311 (meaning no forward references). Second, if a job is being referenced as a
2312 waitee, it must have a unique name (no duplicate waitees).
2314 .. option:: nice=int
2316 Run the job with the given nice value. See man :manpage:`nice(2)`.
2318 On Windows, values less than -15 set the process class to "High"; -1 through
2319 -15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle"
2322 .. option:: prio=int
2324 Set the I/O priority value of this job. Linux limits us to a positive value
2325 between 0 and 7, with 0 being the highest. See man
2326 :manpage:`ionice(1)`. Refer to an appropriate manpage for other operating
2327 systems since meaning of priority may differ.
2329 .. option:: prioclass=int
2331 Set the I/O priority class. See man :manpage:`ionice(1)`.
2333 .. option:: cpumask=int
2335 Set the CPU affinity of this job. The parameter given is a bit mask of
2336 allowed CPUs the job may run on. So if you want the allowed CPUs to be 1
2337 and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man
2338 :manpage:`sched_setaffinity(2)`. This may not work on all supported
2339 operating systems or kernel versions. This option doesn't work well for a
2340 higher CPU count than what you can store in an integer mask, so it can only
2341 control cpus 1-32. For boxes with larger CPU counts, use
2342 :option:`cpus_allowed`.
2344 .. option:: cpus_allowed=str
2346 Controls the same options as :option:`cpumask`, but accepts a textual
2347 specification of the permitted CPUs instead. So to use CPUs 1 and 5 you
2348 would specify ``cpus_allowed=1,5``. This option also allows a range of CPUs
2349 to be specified -- say you wanted a binding to CPUs 1, 5, and 8 to 15, you
2350 would set ``cpus_allowed=1,5,8-15``.
2352 .. option:: cpus_allowed_policy=str
2354 Set the policy of how fio distributes the CPUs specified by
2355 :option:`cpus_allowed` or :option:`cpumask`. Two policies are supported:
2358 All jobs will share the CPU set specified.
2360 Each job will get a unique CPU from the CPU set.
2362 **shared** is the default behavior, if the option isn't specified. If
2363 **split** is specified, then fio will will assign one cpu per job. If not
2364 enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs
2367 .. option:: numa_cpu_nodes=str
2369 Set this job running on specified NUMA nodes' CPUs. The arguments allow
2370 comma delimited list of cpu numbers, A-B ranges, or `all`. Note, to enable
2371 NUMA options support, fio must be built on a system with libnuma-dev(el)
2374 .. option:: numa_mem_policy=str
2376 Set this job's memory policy and corresponding NUMA nodes. Format of the
2381 ``mode`` is one of the following memory poicies: ``default``, ``prefer``,
2382 ``bind``, ``interleave`` or ``local``. For ``default`` and ``local`` memory
2383 policies, no node needs to be specified. For ``prefer``, only one node is
2384 allowed. For ``bind`` and ``interleave`` the ``nodelist`` may be as
2385 follows: a comma delimited list of numbers, A-B ranges, or `all`.
2387 .. option:: cgroup=str
2389 Add job to this control group. If it doesn't exist, it will be created. The
2390 system must have a mounted cgroup blkio mount point for this to work. If
2391 your system doesn't have it mounted, you can do so with::
2393 # mount -t cgroup -o blkio none /cgroup
2395 .. option:: cgroup_weight=int
2397 Set the weight of the cgroup to this value. See the documentation that comes
2398 with the kernel, allowed values are in the range of 100..1000.
2400 .. option:: cgroup_nodelete=bool
2402 Normally fio will delete the cgroups it has created after the job
2403 completion. To override this behavior and to leave cgroups around after the
2404 job completion, set ``cgroup_nodelete=1``. This can be useful if one wants
2405 to inspect various cgroup files after job completion. Default: false.
2407 .. option:: flow_id=int
2409 The ID of the flow. If not specified, it defaults to being a global
2410 flow. See :option:`flow`.
2412 .. option:: flow=int
2414 Weight in token-based flow control. If this value is used, then there is a
2415 'flow counter' which is used to regulate the proportion of activity between
2416 two or more jobs. Fio attempts to keep this flow counter near zero. The
2417 ``flow`` parameter stands for how much should be added or subtracted to the
2418 flow counter on each iteration of the main I/O loop. That is, if one job has
2419 ``flow=8`` and another job has ``flow=-1``, then there will be a roughly 1:8
2420 ratio in how much one runs vs the other.
2422 .. option:: flow_watermark=int
2424 The maximum value that the absolute value of the flow counter is allowed to
2425 reach before the job must wait for a lower value of the counter.
2427 .. option:: flow_sleep=int
2429 The period of time, in microseconds, to wait after the flow watermark has
2430 been exceeded before retrying operations.
2432 .. option:: stonewall, wait_for_previous
2434 Wait for preceding jobs in the job file to exit, before starting this
2435 one. Can be used to insert serialization points in the job file. A stone
2436 wall also implies starting a new reporting group, see
2437 :option:`group_reporting`.
2441 By default, fio will continue running all other jobs when one job finishes
2442 but sometimes this is not the desired action. Setting ``exitall`` will
2443 instead make fio terminate all other jobs when one job finishes.
2445 .. option:: exec_prerun=str
2447 Before running this job, issue the command specified through
2448 :manpage:`system(3)`. Output is redirected in a file called
2449 :file:`jobname.prerun.txt`.
2451 .. option:: exec_postrun=str
2453 After the job completes, issue the command specified though
2454 :manpage:`system(3)`. Output is redirected in a file called
2455 :file:`jobname.postrun.txt`.
2459 Instead of running as the invoking user, set the user ID to this value
2460 before the thread/process does any work.
2464 Set group ID, see :option:`uid`.
2470 .. option:: verify_only
2472 Do not perform specified workload, only verify data still matches previous
2473 invocation of this workload. This option allows one to check data multiple
2474 times at a later date without overwriting it. This option makes sense only
2475 for workloads that write data, and does not support workloads with the
2476 :option:`time_based` option set.
2478 .. option:: do_verify=bool
2480 Run the verify phase after a write phase. Only valid if :option:`verify` is
2483 .. option:: verify=str
2485 If writing to a file, fio can verify the file contents after each iteration
2486 of the job. Each verification method also implies verification of special
2487 header, which is written to the beginning of each block. This header also
2488 includes meta information, like offset of the block, block number, timestamp
2489 when block was written, etc. :option:`verify` can be combined with
2490 :option:`verify_pattern` option. The allowed values are:
2493 Use an md5 sum of the data area and store it in the header of
2497 Use an experimental crc64 sum of the data area and store it in the
2498 header of each block.
2501 Use a crc32c sum of the data area and store it in the header of
2502 each block. This will automatically use hardware acceleration
2503 (e.g. SSE4.2 on an x86 or CRC crypto extensions on ARM64) but will
2504 fall back to software crc32c if none is found. Generally the
2505 fatest checksum fio supports when hardware accelerated.
2511 Use a crc32 sum of the data area and store it in the header of each
2515 Use a crc16 sum of the data area and store it in the header of each
2519 Use a crc7 sum of the data area and store it in the header of each
2523 Use xxhash as the checksum function. Generally the fastest software
2524 checksum that fio supports.
2527 Use sha512 as the checksum function.
2530 Use sha256 as the checksum function.
2533 Use optimized sha1 as the checksum function.
2536 Use optimized sha3-224 as the checksum function.
2539 Use optimized sha3-256 as the checksum function.
2542 Use optimized sha3-384 as the checksum function.
2545 Use optimized sha3-512 as the checksum function.
2548 This option is deprecated, since now meta information is included in
2549 generic verification header and meta verification happens by
2550 default. For detailed information see the description of the
2551 :option:`verify` setting. This option is kept because of
2552 compatibility's sake with old configurations. Do not use it.
2555 Verify a strict pattern. Normally fio includes a header with some
2556 basic information and checksumming, but if this option is set, only
2557 the specific pattern set with :option:`verify_pattern` is verified.
2560 Only pretend to verify. Useful for testing internals with
2561 :option:`ioengine`\=null, not for much else.
2563 This option can be used for repeated burn-in tests of a system to make sure
2564 that the written data is also correctly read back. If the data direction
2565 given is a read or random read, fio will assume that it should verify a
2566 previously written file. If the data direction includes any form of write,
2567 the verify will be of the newly written data.
2569 .. option:: verifysort=bool
2571 If true, fio will sort written verify blocks when it deems it faster to read
2572 them back in a sorted manner. This is often the case when overwriting an
2573 existing file, since the blocks are already laid out in the file system. You
2574 can ignore this option unless doing huge amounts of really fast I/O where
2575 the red-black tree sorting CPU time becomes significant. Default: true.
2577 .. option:: verifysort_nr=int
2579 Pre-load and sort verify blocks for a read workload.
2581 .. option:: verify_offset=int
2583 Swap the verification header with data somewhere else in the block before
2584 writing. It is swapped back before verifying.
2586 .. option:: verify_interval=int
2588 Write the verification header at a finer granularity than the
2589 :option:`blocksize`. It will be written for chunks the size of
2590 ``verify_interval``. :option:`blocksize` should divide this evenly.
2592 .. option:: verify_pattern=str
2594 If set, fio will fill the I/O buffers with this pattern. Fio defaults to
2595 filling with totally random bytes, but sometimes it's interesting to fill
2596 with a known pattern for I/O verification purposes. Depending on the width
2597 of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time (it can
2598 be either a decimal or a hex number). The ``verify_pattern`` if larger than
2599 a 32-bit quantity has to be a hex number that starts with either "0x" or
2600 "0X". Use with :option:`verify`. Also, ``verify_pattern`` supports %o
2601 format, which means that for each block offset will be written and then
2602 verified back, e.g.::
2606 Or use combination of everything::
2608 verify_pattern=0xff%o"abcd"-12
2610 .. option:: verify_fatal=bool
2612 Normally fio will keep checking the entire contents before quitting on a
2613 block verification failure. If this option is set, fio will exit the job on
2614 the first observed failure. Default: false.
2616 .. option:: verify_dump=bool
2618 If set, dump the contents of both the original data block and the data block
2619 we read off disk to files. This allows later analysis to inspect just what
2620 kind of data corruption occurred. Off by default.
2622 .. option:: verify_async=int
2624 Fio will normally verify I/O inline from the submitting thread. This option
2625 takes an integer describing how many async offload threads to create for I/O
2626 verification instead, causing fio to offload the duty of verifying I/O
2627 contents to one or more separate threads. If using this offload option, even
2628 sync I/O engines can benefit from using an :option:`iodepth` setting higher
2629 than 1, as it allows them to have I/O in flight while verifies are running.
2630 Defaults to 0 async threads, i.e. verification is not asynchronous.
2632 .. option:: verify_async_cpus=str
2634 Tell fio to set the given CPU affinity on the async I/O verification
2635 threads. See :option:`cpus_allowed` for the format used.
2637 .. option:: verify_backlog=int
2639 Fio will normally verify the written contents of a job that utilizes verify
2640 once that job has completed. In other words, everything is written then
2641 everything is read back and verified. You may want to verify continually
2642 instead for a variety of reasons. Fio stores the meta data associated with
2643 an I/O block in memory, so for large verify workloads, quite a bit of memory
2644 would be used up holding this meta data. If this option is enabled, fio will
2645 write only N blocks before verifying these blocks.
2647 .. option:: verify_backlog_batch=int
2649 Control how many blocks fio will verify if :option:`verify_backlog` is
2650 set. If not set, will default to the value of :option:`verify_backlog`
2651 (meaning the entire queue is read back and verified). If
2652 ``verify_backlog_batch`` is less than :option:`verify_backlog` then not all
2653 blocks will be verified, if ``verify_backlog_batch`` is larger than
2654 :option:`verify_backlog`, some blocks will be verified more than once.
2656 .. option:: verify_state_save=bool
2658 When a job exits during the write phase of a verify workload, save its
2659 current state. This allows fio to replay up until that point, if the verify
2660 state is loaded for the verify read phase. The format of the filename is,
2663 <type>-<jobname>-<jobindex>-verify.state.
2665 <type> is "local" for a local run, "sock" for a client/server socket
2666 connection, and "ip" (192.168.0.1, for instance) for a networked
2667 client/server connection. Defaults to true.
2669 .. option:: verify_state_load=bool
2671 If a verify termination trigger was used, fio stores the current write state
2672 of each thread. This can be used at verification time so that fio knows how
2673 far it should verify. Without this information, fio will run a full
2674 verification pass, according to the settings in the job file used. Default
2677 .. option:: trim_percentage=int
2679 Number of verify blocks to discard/trim.
2681 .. option:: trim_verify_zero=bool
2683 Verify that trim/discarded blocks are returned as zeros.
2685 .. option:: trim_backlog=int
2687 Trim after this number of blocks are written.
2689 .. option:: trim_backlog_batch=int
2691 Trim this number of I/O blocks.
2693 .. option:: experimental_verify=bool
2695 Enable experimental verification.
2700 .. option:: steadystate=str:float, ss=str:float
2702 Define the criterion and limit for assessing steady state performance. The
2703 first parameter designates the criterion whereas the second parameter sets
2704 the threshold. When the criterion falls below the threshold for the
2705 specified duration, the job will stop. For example, `iops_slope:0.1%` will
2706 direct fio to terminate the job when the least squares regression slope
2707 falls below 0.1% of the mean IOPS. If :option:`group_reporting` is enabled
2708 this will apply to all jobs in the group. Below is the list of available
2709 steady state assessment criteria. All assessments are carried out using only
2710 data from the rolling collection window. Threshold limits can be expressed
2711 as a fixed value or as a percentage of the mean in the collection window.
2714 Collect IOPS data. Stop the job if all individual IOPS measurements
2715 are within the specified limit of the mean IOPS (e.g., ``iops:2``
2716 means that all individual IOPS values must be within 2 of the mean,
2717 whereas ``iops:0.2%`` means that all individual IOPS values must be
2718 within 0.2% of the mean IOPS to terminate the job).
2721 Collect IOPS data and calculate the least squares regression
2722 slope. Stop the job if the slope falls below the specified limit.
2725 Collect bandwidth data. Stop the job if all individual bandwidth
2726 measurements are within the specified limit of the mean bandwidth.
2729 Collect bandwidth data and calculate the least squares regression
2730 slope. Stop the job if the slope falls below the specified limit.
2732 .. option:: steadystate_duration=time, ss_dur=time
2734 A rolling window of this duration will be used to judge whether steady state
2735 has been reached. Data will be collected once per second. The default is 0
2736 which disables steady state detection. When the unit is omitted, the
2737 value is interpreted in seconds.
2739 .. option:: steadystate_ramp_time=time, ss_ramp=time
2741 Allow the job to run for the specified duration before beginning data
2742 collection for checking the steady state job termination criterion. The
2743 default is 0. When the unit is omitted, the value is interpreted in seconds.
2746 Measurements and reporting
2747 ~~~~~~~~~~~~~~~~~~~~~~~~~~
2749 .. option:: per_job_logs=bool
2751 If set, this generates bw/clat/iops log with per file private filenames. If
2752 not set, jobs with identical names will share the log filename. Default:
2755 .. option:: group_reporting
2757 It may sometimes be interesting to display statistics for groups of jobs as
2758 a whole instead of for each individual job. This is especially true if
2759 :option:`numjobs` is used; looking at individual thread/process output
2760 quickly becomes unwieldy. To see the final report per-group instead of
2761 per-job, use :option:`group_reporting`. Jobs in a file will be part of the
2762 same reporting group, unless if separated by a :option:`stonewall`, or by
2763 using :option:`new_group`.
2765 .. option:: new_group
2767 Start a new reporting group. See: :option:`group_reporting`. If not given,
2768 all jobs in a file will be part of the same reporting group, unless
2769 separated by a :option:`stonewall`.
2771 .. option:: stats=bool
2773 By default, fio collects and shows final output results for all jobs
2774 that run. If this option is set to 0, then fio will ignore it in
2775 the final stat output.
2777 .. option:: write_bw_log=str
2779 If given, write a bandwidth log for this job. Can be used to store data of
2780 the bandwidth of the jobs in their lifetime.
2782 If no str argument is given, the default filename of
2783 :file:`jobname_type.x.log` is used. Even when the argument is given, fio
2784 will still append the type of log. So if one specifies::
2788 The actual log name will be :file:`foo_bw.x.log` where `x` is the index
2789 of the job (`1..N`, where `N` is the number of jobs). If
2790 :option:`per_job_logs` is false, then the filename will not include the
2793 The included :command:`fio_generate_plots` script uses :command:`gnuplot` to turn these
2794 text files into nice graphs. See `Log File Formats`_ for how data is
2795 structured within the file.
2797 .. option:: write_lat_log=str
2799 Same as :option:`write_bw_log`, except this option creates I/O
2800 submission (e.g., :file:`name_slat.x.log`), completion (e.g.,
2801 :file:`name_clat.x.log`), and total (e.g., :file:`name_lat.x.log`)
2802 latency files instead. See :option:`write_bw_log` for details about
2803 the filename format and `Log File Formats`_ for how data is structured
2806 .. option:: write_hist_log=str
2808 Same as :option:`write_bw_log` but writes an I/O completion latency
2809 histogram file (e.g., :file:`name_hist.x.log`) instead. Note that this
2810 file will be empty unless :option:`log_hist_msec` has also been set.
2811 See :option:`write_bw_log` for details about the filename format and
2812 `Log File Formats`_ for how data is structured within the file.
2814 .. option:: write_iops_log=str
2816 Same as :option:`write_bw_log`, but writes an IOPS file (e.g.
2817 :file:`name_iops.x.log`) instead. See :option:`write_bw_log` for
2818 details about the filename format and `Log File Formats`_ for how data
2819 is structured within the file.
2821 .. option:: log_avg_msec=int
2823 By default, fio will log an entry in the iops, latency, or bw log for every
2824 I/O that completes. When writing to the disk log, that can quickly grow to a
2825 very large size. Setting this option makes fio average the each log entry
2826 over the specified period of time, reducing the resolution of the log. See
2827 :option:`log_max_value` as well. Defaults to 0, logging all entries.
2828 Also see `Log File Formats`_.
2830 .. option:: log_hist_msec=int
2832 Same as :option:`log_avg_msec`, but logs entries for completion latency
2833 histograms. Computing latency percentiles from averages of intervals using
2834 :option:`log_avg_msec` is inaccurate. Setting this option makes fio log
2835 histogram entries over the specified period of time, reducing log sizes for
2836 high IOPS devices while retaining percentile accuracy. See
2837 :option:`log_hist_coarseness` and :option:`write_hist_log` as well.
2838 Defaults to 0, meaning histogram logging is disabled.
2840 .. option:: log_hist_coarseness=int
2842 Integer ranging from 0 to 6, defining the coarseness of the resolution of
2843 the histogram logs enabled with :option:`log_hist_msec`. For each increment
2844 in coarseness, fio outputs half as many bins. Defaults to 0, for which
2845 histogram logs contain 1216 latency bins. See :option:`write_hist_log`
2846 and `Log File Formats`_.
2848 .. option:: log_max_value=bool
2850 If :option:`log_avg_msec` is set, fio logs the average over that window. If
2851 you instead want to log the maximum value, set this option to 1. Defaults to
2852 0, meaning that averaged values are logged.
2854 .. option:: log_offset=bool
2856 If this is set, the iolog options will include the byte offset for the I/O
2857 entry as well as the other data values. Defaults to 0 meaning that
2858 offsets are not present in logs. Also see `Log File Formats`_.
2860 .. option:: log_compression=int
2862 If this is set, fio will compress the I/O logs as it goes, to keep the
2863 memory footprint lower. When a log reaches the specified size, that chunk is
2864 removed and compressed in the background. Given that I/O logs are fairly
2865 highly compressible, this yields a nice memory savings for longer runs. The
2866 downside is that the compression will consume some background CPU cycles, so
2867 it may impact the run. This, however, is also true if the logging ends up
2868 consuming most of the system memory. So pick your poison. The I/O logs are
2869 saved normally at the end of a run, by decompressing the chunks and storing
2870 them in the specified log file. This feature depends on the availability of
2873 .. option:: log_compression_cpus=str
2875 Define the set of CPUs that are allowed to handle online log compression for
2876 the I/O jobs. This can provide better isolation between performance
2877 sensitive jobs, and background compression work.
2879 .. option:: log_store_compressed=bool
2881 If set, fio will store the log files in a compressed format. They can be
2882 decompressed with fio, using the :option:`--inflate-log` command line
2883 parameter. The files will be stored with a :file:`.fz` suffix.
2885 .. option:: log_unix_epoch=bool
2887 If set, fio will log Unix timestamps to the log files produced by enabling
2888 write_type_log for each log type, instead of the default zero-based
2891 .. option:: block_error_percentiles=bool
2893 If set, record errors in trim block-sized units from writes and trims and
2894 output a histogram of how many trims it took to get to errors, and what kind
2895 of error was encountered.
2897 .. option:: bwavgtime=int
2899 Average the calculated bandwidth over the given time. Value is specified in
2900 milliseconds. If the job also does bandwidth logging through
2901 :option:`write_bw_log`, then the minimum of this option and
2902 :option:`log_avg_msec` will be used. Default: 500ms.
2904 .. option:: iopsavgtime=int
2906 Average the calculated IOPS over the given time. Value is specified in
2907 milliseconds. If the job also does IOPS logging through
2908 :option:`write_iops_log`, then the minimum of this option and
2909 :option:`log_avg_msec` will be used. Default: 500ms.
2911 .. option:: disk_util=bool
2913 Generate disk utilization statistics, if the platform supports it.
2916 .. option:: disable_lat=bool
2918 Disable measurements of total latency numbers. Useful only for cutting back
2919 the number of calls to :manpage:`gettimeofday(2)`, as that does impact
2920 performance at really high IOPS rates. Note that to really get rid of a
2921 large amount of these calls, this option must be used with
2922 :option:`disable_slat` and :option:`disable_bw_measurement` as well.
2924 .. option:: disable_clat=bool
2926 Disable measurements of completion latency numbers. See
2927 :option:`disable_lat`.
2929 .. option:: disable_slat=bool
2931 Disable measurements of submission latency numbers. See
2932 :option:`disable_lat`.
2934 .. option:: disable_bw_measurement=bool, disable_bw=bool
2936 Disable measurements of throughput/bandwidth numbers. See
2937 :option:`disable_lat`.
2939 .. option:: clat_percentiles=bool
2941 Enable the reporting of percentiles of completion latencies. This
2942 option is mutually exclusive with :option:`lat_percentiles`.
2944 .. option:: lat_percentiles=bool
2946 Enable the reporting of percentiles of I/O latencies. This is similar
2947 to :option:`clat_percentiles`, except that this includes the
2948 submission latency. This option is mutually exclusive with
2949 :option:`clat_percentiles`.
2951 .. option:: percentile_list=float_list
2953 Overwrite the default list of percentiles for completion latencies and the
2954 block error histogram. Each number is a floating number in the range
2955 (0,100], and the maximum length of the list is 20. Use ``:`` to separate the
2956 numbers, and list the numbers in ascending order. For example,
2957 ``--percentile_list=99.5:99.9`` will cause fio to report the values of
2958 completion latency below which 99.5% and 99.9% of the observed latencies
2961 .. option:: significant_figures=int
2963 If using :option:`--output-format` of `normal`, set the significant figures
2964 to this value. Higher values will yield more precise IOPS and throughput
2965 units, while lower values will round. Requires a minimum value of 1 and a
2966 maximum value of 10. Defaults to 4.
2972 .. option:: exitall_on_error
2974 When one job finishes in error, terminate the rest. The default is to wait
2975 for each job to finish.
2977 .. option:: continue_on_error=str
2979 Normally fio will exit the job on the first observed failure. If this option
2980 is set, fio will continue the job when there is a 'non-fatal error' (EIO or
2981 EILSEQ) until the runtime is exceeded or the I/O size specified is
2982 completed. If this option is used, there are two more stats that are
2983 appended, the total error count and the first error. The error field given
2984 in the stats is the first error that was hit during the run.
2986 The allowed values are:
2989 Exit on any I/O or verify errors.
2992 Continue on read errors, exit on all others.
2995 Continue on write errors, exit on all others.
2998 Continue on any I/O error, exit on all others.
3001 Continue on verify errors, exit on all others.
3004 Continue on all errors.
3007 Backward-compatible alias for 'none'.
3010 Backward-compatible alias for 'all'.
3012 .. option:: ignore_error=str
3014 Sometimes you want to ignore some errors during test in that case you can
3015 specify error list for each error type, instead of only being able to
3016 ignore the default 'non-fatal error' using :option:`continue_on_error`.
3017 ``ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST`` errors for
3018 given error type is separated with ':'. Error may be symbol ('ENOSPC',
3019 'ENOMEM') or integer. Example::
3021 ignore_error=EAGAIN,ENOSPC:122
3023 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from
3024 WRITE. This option works by overriding :option:`continue_on_error` with
3025 the list of errors for each error type if any.
3027 .. option:: error_dump=bool
3029 If set dump every error even if it is non fatal, true by default. If
3030 disabled only fatal error will be dumped.
3032 Running predefined workloads
3033 ----------------------------
3035 Fio includes predefined profiles that mimic the I/O workloads generated by
3038 .. option:: profile=str
3040 The predefined workload to run. Current profiles are:
3043 Threaded I/O bench (tiotest/tiobench) like workload.
3046 Aerospike Certification Tool (ACT) like workload.
3048 To view a profile's additional options use :option:`--cmdhelp` after specifying
3049 the profile. For example::
3051 $ fio --profile=act --cmdhelp
3056 .. option:: device-names=str
3061 .. option:: load=int
3064 ACT load multiplier. Default: 1.
3066 .. option:: test-duration=time
3069 How long the entire test takes to run. When the unit is omitted, the value
3070 is given in seconds. Default: 24h.
3072 .. option:: threads-per-queue=int
3075 Number of read I/O threads per device. Default: 8.
3077 .. option:: read-req-num-512-blocks=int
3080 Number of 512B blocks to read at the time. Default: 3.
3082 .. option:: large-block-op-kbytes=int
3085 Size of large block ops in KiB (writes). Default: 131072.
3090 Set to run ACT prep phase.
3092 Tiobench profile options
3093 ~~~~~~~~~~~~~~~~~~~~~~~~
3095 .. option:: size=str
3100 .. option:: block=int
3103 Block size in bytes. Default: 4096.
3105 .. option:: numruns=int
3115 .. option:: threads=int
3120 Interpreting the output
3121 -----------------------
3124 Example output was based on the following:
3125 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --time_based \
3126 --rate=1256k --bs=14K --name=quick --runtime=1s --name=mixed \
3127 --runtime=2m --rw=rw
3129 Fio spits out a lot of output. While running, fio will display the status of the
3130 jobs created. An example of that would be::
3132 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]
3134 The characters inside the first set of square brackets denote the current status of
3135 each thread. The first character is the first job defined in the job file, and so
3136 forth. The possible values (in typical life cycle order) are:
3138 +------+-----+-----------------------------------------------------------+
3140 +======+=====+===========================================================+
3141 | P | | Thread setup, but not started. |
3142 +------+-----+-----------------------------------------------------------+
3143 | C | | Thread created. |
3144 +------+-----+-----------------------------------------------------------+
3145 | I | | Thread initialized, waiting or generating necessary data. |
3146 +------+-----+-----------------------------------------------------------+
3147 | | p | Thread running pre-reading file(s). |
3148 +------+-----+-----------------------------------------------------------+
3149 | | / | Thread is in ramp period. |
3150 +------+-----+-----------------------------------------------------------+
3151 | | R | Running, doing sequential reads. |
3152 +------+-----+-----------------------------------------------------------+
3153 | | r | Running, doing random reads. |
3154 +------+-----+-----------------------------------------------------------+
3155 | | W | Running, doing sequential writes. |
3156 +------+-----+-----------------------------------------------------------+
3157 | | w | Running, doing random writes. |
3158 +------+-----+-----------------------------------------------------------+
3159 | | M | Running, doing mixed sequential reads/writes. |
3160 +------+-----+-----------------------------------------------------------+
3161 | | m | Running, doing mixed random reads/writes. |
3162 +------+-----+-----------------------------------------------------------+
3163 | | D | Running, doing sequential trims. |
3164 +------+-----+-----------------------------------------------------------+
3165 | | d | Running, doing random trims. |
3166 +------+-----+-----------------------------------------------------------+
3167 | | F | Running, currently waiting for :manpage:`fsync(2)`. |
3168 +------+-----+-----------------------------------------------------------+
3169 | | V | Running, doing verification of written data. |
3170 +------+-----+-----------------------------------------------------------+
3171 | f | | Thread finishing. |
3172 +------+-----+-----------------------------------------------------------+
3173 | E | | Thread exited, not reaped by main thread yet. |
3174 +------+-----+-----------------------------------------------------------+
3175 | _ | | Thread reaped. |
3176 +------+-----+-----------------------------------------------------------+
3177 | X | | Thread reaped, exited with an error. |
3178 +------+-----+-----------------------------------------------------------+
3179 | K | | Thread reaped, exited due to signal. |
3180 +------+-----+-----------------------------------------------------------+
3183 Example output was based on the following:
3184 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --runtime=58m \
3185 --time_based --rate=2512k --bs=256K --numjobs=10 \
3186 --name=readers --rw=read --name=writers --rw=write
3188 Fio will condense the thread string as not to take up more space on the command
3189 line than needed. For instance, if you have 10 readers and 10 writers running,
3190 the output would look like this::
3192 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]
3194 Note that the status string is displayed in order, so it's possible to tell which of
3195 the jobs are currently doing what. In the example above this means that jobs 1--10
3196 are readers and 11--20 are writers.
3198 The other values are fairly self explanatory -- number of threads currently
3199 running and doing I/O, the number of currently open files (f=), the estimated
3200 completion percentage, the rate of I/O since last check (read speed listed first,
3201 then write speed and optionally trim speed) in terms of bandwidth and IOPS,
3202 and time to completion for the current running group. It's impossible to estimate
3203 runtime of the following groups (if any).
3206 Example output was based on the following:
3207 TZ=UTC fio --iodepth=16 --ioengine=posixaio --filename=/tmp/fiofile \
3208 --direct=1 --size=100M --time_based --runtime=50s --rate_iops=89 \
3209 --bs=7K --name=Client1 --rw=write
3211 When fio is done (or interrupted by :kbd:`Ctrl-C`), it will show the data for
3212 each thread, group of threads, and disks in that order. For each overall thread (or
3213 group) the output looks like::
3215 Client1: (groupid=0, jobs=1): err= 0: pid=16109: Sat Jun 24 12:07:54 2017
3216 write: IOPS=88, BW=623KiB/s (638kB/s)(30.4MiB/50032msec)
3217 slat (nsec): min=500, max=145500, avg=8318.00, stdev=4781.50
3218 clat (usec): min=170, max=78367, avg=4019.02, stdev=8293.31
3219 lat (usec): min=174, max=78375, avg=4027.34, stdev=8291.79
3220 clat percentiles (usec):
3221 | 1.00th=[ 302], 5.00th=[ 326], 10.00th=[ 343], 20.00th=[ 363],
3222 | 30.00th=[ 392], 40.00th=[ 404], 50.00th=[ 416], 60.00th=[ 445],
3223 | 70.00th=[ 816], 80.00th=[ 6718], 90.00th=[12911], 95.00th=[21627],
3224 | 99.00th=[43779], 99.50th=[51643], 99.90th=[68682], 99.95th=[72877],
3226 bw ( KiB/s): min= 532, max= 686, per=0.10%, avg=622.87, stdev=24.82, samples= 100
3227 iops : min= 76, max= 98, avg=88.98, stdev= 3.54, samples= 100
3228 lat (usec) : 250=0.04%, 500=64.11%, 750=4.81%, 1000=2.79%
3229 lat (msec) : 2=4.16%, 4=1.84%, 10=4.90%, 20=11.33%, 50=5.37%
3230 lat (msec) : 100=0.65%
3231 cpu : usr=0.27%, sys=0.18%, ctx=12072, majf=0, minf=21
3232 IO depths : 1=85.0%, 2=13.1%, 4=1.8%, 8=0.1%, 16=0.0%, 32=0.0%, >=64=0.0%
3233 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
3234 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
3235 issued rwt: total=0,4450,0, short=0,0,0, dropped=0,0,0
3236 latency : target=0, window=0, percentile=100.00%, depth=8
3238 The job name (or first job's name when using :option:`group_reporting`) is printed,
3239 along with the group id, count of jobs being aggregated, last error id seen (which
3240 is 0 when there are no errors), pid/tid of that thread and the time the job/group
3241 completed. Below are the I/O statistics for each data direction performed (showing
3242 writes in the example above). In the order listed, they denote:
3245 The string before the colon shows the I/O direction the statistics
3246 are for. **IOPS** is the average I/Os performed per second. **BW**
3247 is the average bandwidth rate shown as: value in power of 2 format
3248 (value in power of 10 format). The last two values show: (**total
3249 I/O performed** in power of 2 format / **runtime** of that thread).
3252 Submission latency (**min** being the minimum, **max** being the
3253 maximum, **avg** being the average, **stdev** being the standard
3254 deviation). This is the time it took to submit the I/O. For
3255 sync I/O this row is not displayed as the slat is really the
3256 completion latency (since queue/complete is one operation there).
3257 This value can be in nanoseconds, microseconds or milliseconds ---
3258 fio will choose the most appropriate base and print that (in the
3259 example above nanoseconds was the best scale). Note: in :option:`--minimal` mode
3260 latencies are always expressed in microseconds.
3263 Completion latency. Same names as slat, this denotes the time from
3264 submission to completion of the I/O pieces. For sync I/O, clat will
3265 usually be equal (or very close) to 0, as the time from submit to
3266 complete is basically just CPU time (I/O has already been done, see slat
3270 Total latency. Same names as slat and clat, this denotes the time from
3271 when fio created the I/O unit to completion of the I/O operation.
3274 Bandwidth statistics based on samples. Same names as the xlat stats,
3275 but also includes the number of samples taken (**samples**) and an
3276 approximate percentage of total aggregate bandwidth this thread
3277 received in its group (**per**). This last value is only really
3278 useful if the threads in this group are on the same disk, since they
3279 are then competing for disk access.
3282 IOPS statistics based on samples. Same names as bw.
3284 **lat (nsec/usec/msec)**
3285 The distribution of I/O completion latencies. This is the time from when
3286 I/O leaves fio and when it gets completed. Unlike the separate
3287 read/write/trim sections above, the data here and in the remaining
3288 sections apply to all I/Os for the reporting group. 250=0.04% means that
3289 0.04% of the I/Os completed in under 250us. 500=64.11% means that 64.11%
3290 of the I/Os required 250 to 499us for completion.
3293 CPU usage. User and system time, along with the number of context
3294 switches this thread went through, usage of system and user time, and
3295 finally the number of major and minor page faults. The CPU utilization
3296 numbers are averages for the jobs in that reporting group, while the
3297 context and fault counters are summed.
3300 The distribution of I/O depths over the job lifetime. The numbers are
3301 divided into powers of 2 and each entry covers depths from that value
3302 up to those that are lower than the next entry -- e.g., 16= covers
3303 depths from 16 to 31. Note that the range covered by a depth
3304 distribution entry can be different to the range covered by the
3305 equivalent submit/complete distribution entry.
3308 How many pieces of I/O were submitting in a single submit call. Each
3309 entry denotes that amount and below, until the previous entry -- e.g.,
3310 16=100% means that we submitted anywhere between 9 to 16 I/Os per submit
3311 call. Note that the range covered by a submit distribution entry can
3312 be different to the range covered by the equivalent depth distribution
3316 Like the above submit number, but for completions instead.
3319 The number of read/write/trim requests issued, and how many of them were
3323 These values are for :option:`latency_target` and related options. When
3324 these options are engaged, this section describes the I/O depth required
3325 to meet the specified latency target.
3328 Example output was based on the following:
3329 TZ=UTC fio --ioengine=null --iodepth=2 --size=100M --numjobs=2 \
3330 --rate_process=poisson --io_limit=32M --name=read --bs=128k \
3331 --rate=11M --name=write --rw=write --bs=2k --rate=700k
3333 After each client has been listed, the group statistics are printed. They
3334 will look like this::
3336 Run status group 0 (all jobs):
3337 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
3338 WRITE: bw=1231KiB/s (1261kB/s), 616KiB/s-621KiB/s (630kB/s-636kB/s), io=64.0MiB (67.1MB), run=52747-53223msec
3340 For each data direction it prints:
3343 Aggregate bandwidth of threads in this group followed by the
3344 minimum and maximum bandwidth of all the threads in this group.
3345 Values outside of brackets are power-of-2 format and those
3346 within are the equivalent value in a power-of-10 format.
3348 Aggregate I/O performed of all threads in this group. The
3349 format is the same as bw.
3351 The smallest and longest runtimes of the threads in this group.
3353 And finally, the disk statistics are printed. This is Linux specific. They will look like this::
3355 Disk stats (read/write):
3356 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
3358 Each value is printed for both reads and writes, with reads first. The
3362 Number of I/Os performed by all groups.
3364 Number of merges performed by the I/O scheduler.
3366 Number of ticks we kept the disk busy.
3368 Total time spent in the disk queue.
3370 The disk utilization. A value of 100% means we kept the disk
3371 busy constantly, 50% would be a disk idling half of the time.
3373 It is also possible to get fio to dump the current output while it is running,
3374 without terminating the job. To do that, send fio the **USR1** signal. You can
3375 also get regularly timed dumps by using the :option:`--status-interval`
3376 parameter, or by creating a file in :file:`/tmp` named
3377 :file:`fio-dump-status`. If fio sees this file, it will unlink it and dump the
3378 current output status.
3384 For scripted usage where you typically want to generate tables or graphs of the
3385 results, fio can output the results in a semicolon separated format. The format
3386 is one long line of values, such as::
3388 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%
3389 A description of this job goes here.
3391 The job description (if provided) follows on a second line.
3393 To enable terse output, use the :option:`--minimal` or
3394 :option:`--output-format`\=terse command line options. The
3395 first value is the version of the terse output format. If the output has to be
3396 changed for some reason, this number will be incremented by 1 to signify that
3399 Split up, the format is as follows (comments in brackets denote when a
3400 field was introduced or whether it's specific to some terse version):
3404 terse version, fio version [v3], jobname, groupid, error
3408 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3409 Submission latency: min, max, mean, stdev (usec)
3410 Completion latency: min, max, mean, stdev (usec)
3411 Completion latency percentiles: 20 fields (see below)
3412 Total latency: min, max, mean, stdev (usec)
3413 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
3414 IOPS [v5]: min, max, mean, stdev, number of samples
3420 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3421 Submission latency: min, max, mean, stdev (usec)
3422 Completion latency: min, max, mean, stdev (usec)
3423 Completion latency percentiles: 20 fields (see below)
3424 Total latency: min, max, mean, stdev (usec)
3425 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
3426 IOPS [v5]: min, max, mean, stdev, number of samples
3428 TRIM status [all but version 3]:
3430 Fields are similar to READ/WRITE status.
3434 user, system, context switches, major faults, minor faults
3438 <=1, 2, 4, 8, 16, 32, >=64
3440 I/O latencies microseconds::
3442 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
3444 I/O latencies milliseconds::
3446 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
3448 Disk utilization [v3]::
3450 disk name, read ios, write ios, read merges, write merges, read ticks, write ticks,
3451 time spent in queue, disk utilization percentage
3453 Additional Info (dependent on continue_on_error, default off)::
3455 total # errors, first error code
3457 Additional Info (dependent on description being set)::
3461 Completion latency percentiles can be a grouping of up to 20 sets, so for the
3462 terse output fio writes all of them. Each field will look like this::
3466 which is the Xth percentile, and the `usec` latency associated with it.
3468 For `Disk utilization`, all disks used by fio are shown. So for each disk there
3469 will be a disk utilization section.
3471 Below is a single line containing short names for each of the fields in the
3472 minimal output v3, separated by semicolons::
3474 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
3480 The `json` output format is intended to be both human readable and convenient
3481 for automated parsing. For the most part its sections mirror those of the
3482 `normal` output. The `runtime` value is reported in msec and the `bw` value is
3483 reported in 1024 bytes per second units.
3489 The `json+` output format is identical to the `json` output format except that it
3490 adds a full dump of the completion latency bins. Each `bins` object contains a
3491 set of (key, value) pairs where keys are latency durations and values count how
3492 many I/Os had completion latencies of the corresponding duration. For example,
3495 "bins" : { "87552" : 1, "89600" : 1, "94720" : 1, "96768" : 1, "97792" : 1, "99840" : 1, "100864" : 2, "103936" : 6, "104960" : 534, "105984" : 5995, "107008" : 7529, ... }
3497 This data indicates that one I/O required 87,552ns to complete, two I/Os required
3498 100,864ns to complete, and 7529 I/Os required 107,008ns to complete.
3500 Also included with fio is a Python script `fio_jsonplus_clat2csv` that takes
3501 json+ output and generates CSV-formatted latency data suitable for plotting.
3503 The latency durations actually represent the midpoints of latency intervals.
3504 For details refer to :file:`stat.h`.
3510 There are two trace file format that you can encounter. The older (v1) format is
3511 unsupported since version 1.20-rc3 (March 2008). It will still be described
3512 below in case that you get an old trace and want to understand it.
3514 In any case the trace is a simple text file with a single action per line.
3517 Trace file format v1
3518 ~~~~~~~~~~~~~~~~~~~~
3520 Each line represents a single I/O action in the following format::
3524 where `rw=0/1` for read/write, and the `offset` and `length` entries being in bytes.
3526 This format is not supported in fio versions >= 1.20-rc3.
3529 Trace file format v2
3530 ~~~~~~~~~~~~~~~~~~~~
3532 The second version of the trace file format was added in fio version 1.17. It
3533 allows to access more then one file per trace and has a bigger set of possible
3536 The first line of the trace file has to be::
3540 Following this can be lines in two different formats, which are described below.
3542 The file management format::
3546 The `filename` is given as an absolute path. The `action` can be one of these:
3549 Add the given `filename` to the trace.
3551 Open the file with the given `filename`. The `filename` has to have
3552 been added with the **add** action before.
3554 Close the file with the given `filename`. The file has to have been
3558 The file I/O action format::
3560 filename action offset length
3562 The `filename` is given as an absolute path, and has to have been added and
3563 opened before it can be used with this format. The `offset` and `length` are
3564 given in bytes. The `action` can be one of these:
3567 Wait for `offset` microseconds. Everything below 100 is discarded.
3568 The time is relative to the previous `wait` statement.
3570 Read `length` bytes beginning from `offset`.
3572 Write `length` bytes beginning from `offset`.
3574 :manpage:`fsync(2)` the file.
3576 :manpage:`fdatasync(2)` the file.
3578 Trim the given file from the given `offset` for `length` bytes.
3580 CPU idleness profiling
3581 ----------------------
3583 In some cases, we want to understand CPU overhead in a test. For example, we
3584 test patches for the specific goodness of whether they reduce CPU usage.
3585 Fio implements a balloon approach to create a thread per CPU that runs at idle
3586 priority, meaning that it only runs when nobody else needs the cpu.
3587 By measuring the amount of work completed by the thread, idleness of each CPU
3588 can be derived accordingly.
3590 An unit work is defined as touching a full page of unsigned characters. Mean and
3591 standard deviation of time to complete an unit work is reported in "unit work"
3592 section. Options can be chosen to report detailed percpu idleness or overall
3593 system idleness by aggregating percpu stats.
3596 Verification and triggers
3597 -------------------------
3599 Fio is usually run in one of two ways, when data verification is done. The first
3600 is a normal write job of some sort with verify enabled. When the write phase has
3601 completed, fio switches to reads and verifies everything it wrote. The second
3602 model is running just the write phase, and then later on running the same job
3603 (but with reads instead of writes) to repeat the same I/O patterns and verify
3604 the contents. Both of these methods depend on the write phase being completed,
3605 as fio otherwise has no idea how much data was written.
3607 With verification triggers, fio supports dumping the current write state to
3608 local files. Then a subsequent read verify workload can load this state and know
3609 exactly where to stop. This is useful for testing cases where power is cut to a
3610 server in a managed fashion, for instance.
3612 A verification trigger consists of two things:
3614 1) Storing the write state of each job.
3615 2) Executing a trigger command.
3617 The write state is relatively small, on the order of hundreds of bytes to single
3618 kilobytes. It contains information on the number of completions done, the last X
3621 A trigger is invoked either through creation ('touch') of a specified file in
3622 the system, or through a timeout setting. If fio is run with
3623 :option:`--trigger-file`\= :file:`/tmp/trigger-file`, then it will continually
3624 check for the existence of :file:`/tmp/trigger-file`. When it sees this file, it
3625 will fire off the trigger (thus saving state, and executing the trigger
3628 For client/server runs, there's both a local and remote trigger. If fio is
3629 running as a server backend, it will send the job states back to the client for
3630 safe storage, then execute the remote trigger, if specified. If a local trigger
3631 is specified, the server will still send back the write state, but the client
3632 will then execute the trigger.
3634 Verification trigger example
3635 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3637 Let's say we want to run a powercut test on the remote Linux machine 'server'.
3638 Our write workload is in :file:`write-test.fio`. We want to cut power to 'server' at
3639 some point during the run, and we'll run this test from the safety or our local
3640 machine, 'localbox'. On the server, we'll start the fio backend normally::
3642 server# fio --server
3644 and on the client, we'll fire off the workload::
3646 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger-remote="bash -c \"echo b > /proc/sysrq-triger\""
3648 We set :file:`/tmp/my-trigger` as the trigger file, and we tell fio to execute::
3650 echo b > /proc/sysrq-trigger
3652 on the server once it has received the trigger and sent us the write state. This
3653 will work, but it's not **really** cutting power to the server, it's merely
3654 abruptly rebooting it. If we have a remote way of cutting power to the server
3655 through IPMI or similar, we could do that through a local trigger command
3656 instead. Let's assume we have a script that does IPMI reboot of a given hostname,
3657 ipmi-reboot. On localbox, we could then have run fio with a local trigger
3660 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger="ipmi-reboot server"
3662 For this case, fio would wait for the server to send us the write state, then
3663 execute ``ipmi-reboot server`` when that happened.
3665 Loading verify state
3666 ~~~~~~~~~~~~~~~~~~~~
3668 To load stored write state, a read verification job file must contain the
3669 :option:`verify_state_load` option. If that is set, fio will load the previously
3670 stored state. For a local fio run this is done by loading the files directly,
3671 and on a client/server run, the server backend will ask the client to send the
3672 files over and load them from there.
3678 Fio supports a variety of log file formats, for logging latencies, bandwidth,
3679 and IOPS. The logs share a common format, which looks like this:
3681 *time* (`msec`), *value*, *data direction*, *block size* (`bytes`),
3684 *Time* for the log entry is always in milliseconds. The *value* logged depends
3685 on the type of log, it will be one of the following:
3688 Value is latency in nsecs
3694 *Data direction* is one of the following:
3703 The entry's *block size* is always in bytes. The *offset* is the offset, in bytes,
3704 from the start of the file, for that particular I/O. The logging of the offset can be
3705 toggled with :option:`log_offset`.
3707 Fio defaults to logging every individual I/O. When IOPS are logged for individual
3708 I/Os the *value* entry will always be 1. If windowed logging is enabled through
3709 :option:`log_avg_msec`, fio logs the average values over the specified period of time.
3710 If windowed logging is enabled and :option:`log_max_value` is set, then fio logs
3711 maximum values in that window instead of averages. Since *data direction*, *block
3712 size* and *offset* are per-I/O values, if windowed logging is enabled they
3713 aren't applicable and will be 0.
3718 Normally fio is invoked as a stand-alone application on the machine where the
3719 I/O workload should be generated. However, the backend and frontend of fio can
3720 be run separately i.e., the fio server can generate an I/O workload on the "Device
3721 Under Test" while being controlled by a client on another machine.
3723 Start the server on the machine which has access to the storage DUT::
3727 where `args` defines what fio listens to. The arguments are of the form
3728 ``type,hostname`` or ``IP,port``. *type* is either ``ip`` (or ip4) for TCP/IP
3729 v4, ``ip6`` for TCP/IP v6, or ``sock`` for a local unix domain socket.
3730 *hostname* is either a hostname or IP address, and *port* is the port to listen
3731 to (only valid for TCP/IP, not a local socket). Some examples:
3735 Start a fio server, listening on all interfaces on the default port (8765).
3737 2) ``fio --server=ip:hostname,4444``
3739 Start a fio server, listening on IP belonging to hostname and on port 4444.
3741 3) ``fio --server=ip6:::1,4444``
3743 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
3745 4) ``fio --server=,4444``
3747 Start a fio server, listening on all interfaces on port 4444.
3749 5) ``fio --server=1.2.3.4``
3751 Start a fio server, listening on IP 1.2.3.4 on the default port.
3753 6) ``fio --server=sock:/tmp/fio.sock``
3755 Start a fio server, listening on the local socket :file:`/tmp/fio.sock`.
3757 Once a server is running, a "client" can connect to the fio server with::
3759 fio <local-args> --client=<server> <remote-args> <job file(s)>
3761 where `local-args` are arguments for the client where it is running, `server`
3762 is the connect string, and `remote-args` and `job file(s)` are sent to the
3763 server. The `server` string follows the same format as it does on the server
3764 side, to allow IP/hostname/socket and port strings.
3766 Fio can connect to multiple servers this way::
3768 fio --client=<server1> <job file(s)> --client=<server2> <job file(s)>
3770 If the job file is located on the fio server, then you can tell the server to
3771 load a local file as well. This is done by using :option:`--remote-config` ::
3773 fio --client=server --remote-config /path/to/file.fio
3775 Then fio will open this local (to the server) job file instead of being passed
3776 one from the client.
3778 If you have many servers (example: 100 VMs/containers), you can input a pathname
3779 of a file containing host IPs/names as the parameter value for the
3780 :option:`--client` option. For example, here is an example :file:`host.list`
3781 file containing 2 hostnames::
3783 host1.your.dns.domain
3784 host2.your.dns.domain
3786 The fio command would then be::
3788 fio --client=host.list <job file(s)>
3790 In this mode, you cannot input server-specific parameters or job files -- all
3791 servers receive the same job file.
3793 In order to let ``fio --client`` runs use a shared filesystem from multiple
3794 hosts, ``fio --client`` now prepends the IP address of the server to the
3795 filename. For example, if fio is using the directory :file:`/mnt/nfs/fio` and is
3796 writing filename :file:`fileio.tmp`, with a :option:`--client` `hostfile`
3797 containing two hostnames ``h1`` and ``h2`` with IP addresses 192.168.10.120 and
3798 192.168.10.121, then fio will create two files::
3800 /mnt/nfs/fio/192.168.10.120.fileio.tmp
3801 /mnt/nfs/fio/192.168.10.121.fileio.tmp