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`.
221 .. option:: --server=args
223 Start a backend server, with `args` specifying what to listen to.
224 See `Client/Server`_ section.
226 .. option:: --daemonize=pidfile
228 Background a fio server, writing the pid to the given `pidfile` file.
230 .. option:: --client=hostname
232 Instead of running the jobs locally, send and run them on the given `hostname`
233 or set of `hostname`s. See `Client/Server`_ section.
235 .. option:: --remote-config=file
237 Tell fio server to load this local `file`.
239 .. option:: --idle-prof=option
241 Report CPU idleness. `option` is one of the following:
244 Run unit work calibration only and exit.
247 Show aggregate system idleness and unit work.
250 As **system** but also show per CPU idleness.
252 .. option:: --inflate-log=log
254 Inflate and output compressed `log`.
256 .. option:: --trigger-file=file
258 Execute trigger command when `file` exists.
260 .. option:: --trigger-timeout=time
262 Execute trigger at this `time`.
264 .. option:: --trigger=command
266 Set this `command` as local trigger.
268 .. option:: --trigger-remote=command
270 Set this `command` as remote trigger.
272 .. option:: --aux-path=path
274 Use this `path` for fio state generated files.
276 Any parameters following the options will be assumed to be job files, unless
277 they match a job file parameter. Multiple job files can be listed and each job
278 file will be regarded as a separate group. Fio will :option:`stonewall`
279 execution between each group.
285 As previously described, fio accepts one or more job files describing what it is
286 supposed to do. The job file format is the classic ini file, where the names
287 enclosed in [] brackets define the job name. You are free to use any ASCII name
288 you want, except *global* which has special meaning. Following the job name is
289 a sequence of zero or more parameters, one per line, that define the behavior of
290 the job. If the first character in a line is a ';' or a '#', the entire line is
291 discarded as a comment.
293 A *global* section sets defaults for the jobs described in that file. A job may
294 override a *global* section parameter, and a job file may even have several
295 *global* sections if so desired. A job is only affected by a *global* section
298 The :option:`--cmdhelp` option also lists all options. If used with a `command`
299 argument, :option:`--cmdhelp` will detail the given `command`.
301 See the `examples/` directory for inspiration on how to write job files. Note
302 the copyright and license requirements currently apply to `examples/` files.
304 So let's look at a really simple job file that defines two processes, each
305 randomly reading from a 128MiB file:
309 ; -- start job file --
320 As you can see, the job file sections themselves are empty as all the described
321 parameters are shared. As no :option:`filename` option is given, fio makes up a
322 `filename` for each of the jobs as it sees fit. On the command line, this job
323 would look as follows::
325 $ fio --name=global --rw=randread --size=128m --name=job1 --name=job2
328 Let's look at an example that has a number of processes writing randomly to
333 ; -- start job file --
344 Here we have no *global* section, as we only have one job defined anyway. We
345 want to use async I/O here, with a depth of 4 for each file. We also increased
346 the buffer size used to 32KiB and define numjobs to 4 to fork 4 identical
347 jobs. The result is 4 processes each randomly writing to their own 64MiB
348 file. Instead of using the above job file, you could have given the parameters
349 on the command line. For this case, you would specify::
351 $ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4
353 When fio is utilized as a basis of any reasonably large test suite, it might be
354 desirable to share a set of standardized settings across multiple job files.
355 Instead of copy/pasting such settings, any section may pull in an external
356 :file:`filename.fio` file with *include filename* directive, as in the following
359 ; -- start job file including.fio --
363 include glob-include.fio
370 include test-include.fio
371 ; -- end job file including.fio --
375 ; -- start job file glob-include.fio --
378 ; -- end job file glob-include.fio --
382 ; -- start job file test-include.fio --
385 ; -- end job file test-include.fio --
387 Settings pulled into a section apply to that section only (except *global*
388 section). Include directives may be nested in that any included file may contain
389 further include directive(s). Include files may not contain [] sections.
392 Environment variables
393 ~~~~~~~~~~~~~~~~~~~~~
395 Fio also supports environment variable expansion in job files. Any sub-string of
396 the form ``${VARNAME}`` as part of an option value (in other words, on the right
397 of the '='), will be expanded to the value of the environment variable called
398 `VARNAME`. If no such environment variable is defined, or `VARNAME` is the
399 empty string, the empty string will be substituted.
401 As an example, let's look at a sample fio invocation and job file::
403 $ SIZE=64m NUMJOBS=4 fio jobfile.fio
407 ; -- start job file --
414 This will expand to the following equivalent job file at runtime:
418 ; -- start job file --
425 Fio ships with a few example job files, you can also look there for inspiration.
430 Additionally, fio has a set of reserved keywords that will be replaced
431 internally with the appropriate value. Those keywords are:
435 The architecture page size of the running system.
439 Megabytes of total memory in the system.
443 Number of online available CPUs.
445 These can be used on the command line or in the job file, and will be
446 automatically substituted with the current system values when the job is
447 run. Simple math is also supported on these keywords, so you can perform actions
452 and get that properly expanded to 8 times the size of memory in the machine.
458 This section describes in details each parameter associated with a job. Some
459 parameters take an option of a given type, such as an integer or a
460 string. Anywhere a numeric value is required, an arithmetic expression may be
461 used, provided it is surrounded by parentheses. Supported operators are:
470 For time values in expressions, units are microseconds by default. This is
471 different than for time values not in expressions (not enclosed in
472 parentheses). The following types are used:
479 String: A sequence of alphanumeric characters.
482 Integer with possible time suffix. Without a unit value is interpreted as
483 seconds unless otherwise specified. Accepts a suffix of 'd' for days, 'h' for
484 hours, 'm' for minutes, 's' for seconds, 'ms' (or 'msec') for milliseconds and
485 'us' (or 'usec') for microseconds. For example, use 10m for 10 minutes.
490 Integer. A whole number value, which may contain an integer prefix
491 and an integer suffix:
493 [*integer prefix*] **number** [*integer suffix*]
495 The optional *integer prefix* specifies the number's base. The default
496 is decimal. *0x* specifies hexadecimal.
498 The optional *integer suffix* specifies the number's units, and includes an
499 optional unit prefix and an optional unit. For quantities of data, the
500 default unit is bytes. For quantities of time, the default unit is seconds
501 unless otherwise specified.
503 With :option:`kb_base`\=1000, fio follows international standards for unit
504 prefixes. To specify power-of-10 decimal values defined in the
505 International System of Units (SI):
507 * *K* -- means kilo (K) or 1000
508 * *M* -- means mega (M) or 1000**2
509 * *G* -- means giga (G) or 1000**3
510 * *T* -- means tera (T) or 1000**4
511 * *P* -- means peta (P) or 1000**5
513 To specify power-of-2 binary values defined in IEC 80000-13:
515 * *Ki* -- means kibi (Ki) or 1024
516 * *Mi* -- means mebi (Mi) or 1024**2
517 * *Gi* -- means gibi (Gi) or 1024**3
518 * *Ti* -- means tebi (Ti) or 1024**4
519 * *Pi* -- means pebi (Pi) or 1024**5
521 With :option:`kb_base`\=1024 (the default), the unit prefixes are opposite
522 from those specified in the SI and IEC 80000-13 standards to provide
523 compatibility with old scripts. For example, 4k means 4096.
525 For quantities of data, an optional unit of 'B' may be included
526 (e.g., 'kB' is the same as 'k').
528 The *integer suffix* is not case sensitive (e.g., m/mi mean mebi/mega,
529 not milli). 'b' and 'B' both mean byte, not bit.
531 Examples with :option:`kb_base`\=1000:
533 * *4 KiB*: 4096, 4096b, 4096B, 4ki, 4kib, 4kiB, 4Ki, 4KiB
534 * *1 MiB*: 1048576, 1mi, 1024ki
535 * *1 MB*: 1000000, 1m, 1000k
536 * *1 TiB*: 1099511627776, 1ti, 1024gi, 1048576mi
537 * *1 TB*: 1000000000, 1t, 1000m, 1000000k
539 Examples with :option:`kb_base`\=1024 (default):
541 * *4 KiB*: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB
542 * *1 MiB*: 1048576, 1m, 1024k
543 * *1 MB*: 1000000, 1mi, 1000ki
544 * *1 TiB*: 1099511627776, 1t, 1024g, 1048576m
545 * *1 TB*: 1000000000, 1ti, 1000mi, 1000000ki
547 To specify times (units are not case sensitive):
551 * *M* -- means minutes
552 * *s* -- or sec means seconds (default)
553 * *ms* -- or *msec* means milliseconds
554 * *us* -- or *usec* means microseconds
556 If the option accepts an upper and lower range, use a colon ':' or
557 minus '-' to separate such values. See :ref:`irange <irange>`.
558 If the lower value specified happens to be larger than the upper value
559 the two values are swapped.
564 Boolean. Usually parsed as an integer, however only defined for
565 true and false (1 and 0).
570 Integer range with suffix. Allows value range to be given, such as
571 1024-4096. A colon may also be used as the separator, e.g. 1k:4k. If the
572 option allows two sets of ranges, they can be specified with a ',' or '/'
573 delimiter: 1k-4k/8k-32k. Also see :ref:`int <int>`.
576 A list of floating point numbers, separated by a ':' character.
578 With the above in mind, here follows the complete list of fio job parameters.
584 .. option:: kb_base=int
586 Select the interpretation of unit prefixes in input parameters.
589 Inputs comply with IEC 80000-13 and the International
590 System of Units (SI). Use:
592 - power-of-2 values with IEC prefixes (e.g., KiB)
593 - power-of-10 values with SI prefixes (e.g., kB)
596 Compatibility mode (default). To avoid breaking old scripts:
598 - power-of-2 values with SI prefixes
599 - power-of-10 values with IEC prefixes
601 See :option:`bs` for more details on input parameters.
603 Outputs always use correct prefixes. Most outputs include both
606 bw=2383.3kB/s (2327.4KiB/s)
608 If only one value is reported, then kb_base selects the one to use:
610 **1000** -- SI prefixes
612 **1024** -- IEC prefixes
614 .. option:: unit_base=int
616 Base unit for reporting. Allowed values are:
619 Use auto-detection (default).
631 ASCII name of the job. This may be used to override the name printed by fio
632 for this job. Otherwise the job name is used. On the command line this
633 parameter has the special purpose of also signaling the start of a new job.
635 .. option:: description=str
637 Text description of the job. Doesn't do anything except dump this text
638 description when this job is run. It's not parsed.
640 .. option:: loops=int
642 Run the specified number of iterations of this job. Used to repeat the same
643 workload a given number of times. Defaults to 1.
645 .. option:: numjobs=int
647 Create the specified number of clones of this job. Each clone of job
648 is spawned as an independent thread or process. May be used to setup a
649 larger number of threads/processes doing the same thing. Each thread is
650 reported separately; to see statistics for all clones as a whole, use
651 :option:`group_reporting` in conjunction with :option:`new_group`.
652 See :option:`--max-jobs`. Default: 1.
655 Time related parameters
656 ~~~~~~~~~~~~~~~~~~~~~~~
658 .. option:: runtime=time
660 Tell fio to terminate processing after the specified period of time. It
661 can be quite hard to determine for how long a specified job will run, so
662 this parameter is handy to cap the total runtime to a given time. When
663 the unit is omitted, the value is intepreted in seconds.
665 .. option:: time_based
667 If set, fio will run for the duration of the :option:`runtime` specified
668 even if the file(s) are completely read or written. It will simply loop over
669 the same workload as many times as the :option:`runtime` allows.
671 .. option:: startdelay=irange(time)
673 Delay the start of job for the specified amount of time. Can be a single
674 value or a range. When given as a range, each thread will choose a value
675 randomly from within the range. Value is in seconds if a unit is omitted.
677 .. option:: ramp_time=time
679 If set, fio will run the specified workload for this amount of time before
680 logging any performance numbers. Useful for letting performance settle
681 before logging results, thus minimizing the runtime required for stable
682 results. Note that the ``ramp_time`` is considered lead in time for a job,
683 thus it will increase the total runtime if a special timeout or
684 :option:`runtime` is specified. When the unit is omitted, the value is
687 .. option:: clocksource=str
689 Use the given clocksource as the base of timing. The supported options are:
692 :manpage:`gettimeofday(2)`
695 :manpage:`clock_gettime(2)`
698 Internal CPU clock source
700 cpu is the preferred clocksource if it is reliable, as it is very fast (and
701 fio is heavy on time calls). Fio will automatically use this clocksource if
702 it's supported and considered reliable on the system it is running on,
703 unless another clocksource is specifically set. For x86/x86-64 CPUs, this
704 means supporting TSC Invariant.
706 .. option:: gtod_reduce=bool
708 Enable all of the :manpage:`gettimeofday(2)` reducing options
709 (:option:`disable_clat`, :option:`disable_slat`, :option:`disable_bw_measurement`) plus
710 reduce precision of the timeout somewhat to really shrink the
711 :manpage:`gettimeofday(2)` call count. With this option enabled, we only do
712 about 0.4% of the :manpage:`gettimeofday(2)` calls we would have done if all
713 time keeping was enabled.
715 .. option:: gtod_cpu=int
717 Sometimes it's cheaper to dedicate a single thread of execution to just
718 getting the current time. Fio (and databases, for instance) are very
719 intensive on :manpage:`gettimeofday(2)` calls. With this option, you can set
720 one CPU aside for doing nothing but logging current time to a shared memory
721 location. Then the other threads/processes that run I/O workloads need only
722 copy that segment, instead of entering the kernel with a
723 :manpage:`gettimeofday(2)` call. The CPU set aside for doing these time
724 calls will be excluded from other uses. Fio will manually clear it from the
725 CPU mask of other jobs.
731 .. option:: directory=str
733 Prefix filenames with this directory. Used to place files in a different
734 location than :file:`./`. You can specify a number of directories by
735 separating the names with a ':' character. These directories will be
736 assigned equally distributed to job clones created by :option:`numjobs` as
737 long as they are using generated filenames. If specific `filename(s)` are
738 set fio will use the first listed directory, and thereby matching the
739 `filename` semantic which generates a file each clone if not specified, but
740 let all clones use the same if set.
742 See the :option:`filename` option for information on how to escape "``:``" and
743 "``\``" characters within the directory path itself.
745 .. option:: filename=str
747 Fio normally makes up a `filename` based on the job name, thread number, and
748 file number (see :option:`filename_format`). If you want to share files
749 between threads in a job or several
750 jobs with fixed file paths, specify a `filename` for each of them to override
751 the default. If the ioengine is file based, you can specify a number of files
752 by separating the names with a ':' colon. So if you wanted a job to open
753 :file:`/dev/sda` and :file:`/dev/sdb` as the two working files, you would use
754 ``filename=/dev/sda:/dev/sdb``. This also means that whenever this option is
755 specified, :option:`nrfiles` is ignored. The size of regular files specified
756 by this option will be :option:`size` divided by number of files unless an
757 explicit size is specified by :option:`filesize`.
759 Each colon and backslash in the wanted path must be escaped with a ``\``
760 character. For instance, if the path is :file:`/dev/dsk/foo@3,0:c` then you
761 would use ``filename=/dev/dsk/foo@3,0\:c`` and if the path is
762 :file:`F:\\filename` then you would use ``filename=F\:\\filename``.
764 On Windows, disk devices are accessed as :file:`\\\\.\\PhysicalDrive0` for
765 the first device, :file:`\\\\.\\PhysicalDrive1` for the second etc.
766 Note: Windows and FreeBSD prevent write access to areas
767 of the disk containing in-use data (e.g. filesystems).
769 The filename "`-`" is a reserved name, meaning *stdin* or *stdout*. Which
770 of the two depends on the read/write direction set.
772 .. option:: filename_format=str
774 If sharing multiple files between jobs, it is usually necessary to have fio
775 generate the exact names that you want. By default, fio will name a file
776 based on the default file format specification of
777 :file:`jobname.jobnumber.filenumber`. With this option, that can be
778 customized. Fio will recognize and replace the following keywords in this
782 The name of the worker thread or process.
784 The incremental number of the worker thread or process.
786 The incremental number of the file for that worker thread or
789 To have dependent jobs share a set of files, this option can be set to have
790 fio generate filenames that are shared between the two. For instance, if
791 :file:`testfiles.$filenum` is specified, file number 4 for any job will be
792 named :file:`testfiles.4`. The default of :file:`$jobname.$jobnum.$filenum`
793 will be used if no other format specifier is given.
795 .. option:: unique_filename=bool
797 To avoid collisions between networked clients, fio defaults to prefixing any
798 generated filenames (with a directory specified) with the source of the
799 client connecting. To disable this behavior, set this option to 0.
801 .. option:: opendir=str
803 Recursively open any files below directory `str`.
805 .. option:: lockfile=str
807 Fio defaults to not locking any files before it does I/O to them. If a file
808 or file descriptor is shared, fio can serialize I/O to that file to make the
809 end result consistent. This is usual for emulating real workloads that share
810 files. The lock modes are:
813 No locking. The default.
815 Only one thread or process may do I/O at a time, excluding all
818 Read-write locking on the file. Many readers may
819 access the file at the same time, but writes get exclusive access.
821 .. option:: nrfiles=int
823 Number of files to use for this job. Defaults to 1. The size of files
824 will be :option:`size` divided by this unless explicit size is specified by
825 :option:`filesize`. Files are created for each thread separately, and each
826 file will have a file number within its name by default, as explained in
827 :option:`filename` section.
830 .. option:: openfiles=int
832 Number of files to keep open at the same time. Defaults to the same as
833 :option:`nrfiles`, can be set smaller to limit the number simultaneous
836 .. option:: file_service_type=str
838 Defines how fio decides which file from a job to service next. The following
842 Choose a file at random.
845 Round robin over opened files. This is the default.
848 Finish one file before moving on to the next. Multiple files can
849 still be open depending on :option:`openfiles`.
852 Use a *Zipf* distribution to decide what file to access.
855 Use a *Pareto* distribution to decide what file to access.
858 Use a *Gaussian* (normal) distribution to decide what file to
864 For *random*, *roundrobin*, and *sequential*, a postfix can be appended to
865 tell fio how many I/Os to issue before switching to a new file. For example,
866 specifying ``file_service_type=random:8`` would cause fio to issue
867 8 I/Os before selecting a new file at random. For the non-uniform
868 distributions, a floating point postfix can be given to influence how the
869 distribution is skewed. See :option:`random_distribution` for a description
870 of how that would work.
872 .. option:: ioscheduler=str
874 Attempt to switch the device hosting the file to the specified I/O scheduler
877 .. option:: create_serialize=bool
879 If true, serialize the file creation for the jobs. This may be handy to
880 avoid interleaving of data files, which may greatly depend on the filesystem
881 used and even the number of processors in the system. Default: true.
883 .. option:: create_fsync=bool
885 :manpage:`fsync(2)` the data file after creation. This is the default.
887 .. option:: create_on_open=bool
889 If true, don't pre-create files but allow the job's open() to create a file
890 when it's time to do I/O. Default: false -- pre-create all necessary files
893 .. option:: create_only=bool
895 If true, fio will only run the setup phase of the job. If files need to be
896 laid out or updated on disk, only that will be done -- the actual job contents
897 are not executed. Default: false.
899 .. option:: allow_file_create=bool
901 If true, fio is permitted to create files as part of its workload. If this
902 option is false, then fio will error out if
903 the files it needs to use don't already exist. Default: true.
905 .. option:: allow_mounted_write=bool
907 If this isn't set, fio will abort jobs that are destructive (e.g. that write)
908 to what appears to be a mounted device or partition. This should help catch
909 creating inadvertently destructive tests, not realizing that the test will
910 destroy data on the mounted file system. Note that some platforms don't allow
911 writing against a mounted device regardless of this option. Default: false.
913 .. option:: pre_read=bool
915 If this is given, files will be pre-read into memory before starting the
916 given I/O operation. This will also clear the :option:`invalidate` flag,
917 since it is pointless to pre-read and then drop the cache. This will only
918 work for I/O engines that are seek-able, since they allow you to read the
919 same data multiple times. Thus it will not work on non-seekable I/O engines
920 (e.g. network, splice). Default: false.
922 .. option:: unlink=bool
924 Unlink the job files when done. Not the default, as repeated runs of that
925 job would then waste time recreating the file set again and again. Default:
928 .. option:: unlink_each_loop=bool
930 Unlink job files after each iteration or loop. Default: false.
932 .. option:: zonesize=int
934 Divide a file into zones of the specified size. See :option:`zoneskip`.
936 .. option:: zonerange=int
938 Give size of an I/O zone. See :option:`zoneskip`.
940 .. option:: zoneskip=int
942 Skip the specified number of bytes when :option:`zonesize` data has been
943 read. The two zone options can be used to only do I/O on zones of a file.
949 .. option:: direct=bool
951 If value is true, use non-buffered I/O. This is usually O_DIRECT. Note that
952 OpenBSD and ZFS on Solaris don't support direct I/O. On Windows the synchronous
953 ioengines don't support direct I/O. Default: false.
955 .. option:: atomic=bool
957 If value is true, attempt to use atomic direct I/O. Atomic writes are
958 guaranteed to be stable once acknowledged by the operating system. Only
959 Linux supports O_ATOMIC right now.
961 .. option:: buffered=bool
963 If value is true, use buffered I/O. This is the opposite of the
964 :option:`direct` option. Defaults to true.
966 .. option:: readwrite=str, rw=str
968 Type of I/O pattern. Accepted values are:
975 Sequential trims (Linux block devices only).
981 Random trims (Linux block devices only).
983 Sequential mixed reads and writes.
985 Random mixed reads and writes.
987 Sequential trim+write sequences. Blocks will be trimmed first,
988 then the same blocks will be written to.
990 Fio defaults to read if the option is not specified. For the mixed I/O
991 types, the default is to split them 50/50. For certain types of I/O the
992 result may still be skewed a bit, since the speed may be different.
994 It is possible to specify the number of I/Os to do before getting a new
995 offset by appending ``:<nr>`` to the end of the string given. For a
996 random read, it would look like ``rw=randread:8`` for passing in an offset
997 modifier with a value of 8. If the suffix is used with a sequential I/O
998 pattern, then the *<nr>* value specified will be **added** to the generated
999 offset for each I/O turning sequential I/O into sequential I/O with holes.
1000 For instance, using ``rw=write:4k`` will skip 4k for every write. Also see
1001 the :option:`rw_sequencer` option.
1003 .. option:: rw_sequencer=str
1005 If an offset modifier is given by appending a number to the ``rw=<str>``
1006 line, then this option controls how that number modifies the I/O offset
1007 being generated. Accepted values are:
1010 Generate sequential offset.
1012 Generate the same offset.
1014 ``sequential`` is only useful for random I/O, where fio would normally
1015 generate a new random offset for every I/O. If you append e.g. 8 to randread,
1016 you would get a new random offset for every 8 I/Os. The result would be a
1017 seek for only every 8 I/Os, instead of for every I/O. Use ``rw=randread:8``
1018 to specify that. As sequential I/O is already sequential, setting
1019 ``sequential`` for that would not result in any differences. ``identical``
1020 behaves in a similar fashion, except it sends the same offset 8 number of
1021 times before generating a new offset.
1023 .. option:: unified_rw_reporting=bool
1025 Fio normally reports statistics on a per data direction basis, meaning that
1026 reads, writes, and trims are accounted and reported separately. If this
1027 option is set fio sums the results and report them as "mixed" instead.
1029 .. option:: randrepeat=bool
1031 Seed the random number generator used for random I/O patterns in a
1032 predictable way so the pattern is repeatable across runs. Default: true.
1034 .. option:: allrandrepeat=bool
1036 Seed all random number generators in a predictable way so results are
1037 repeatable across runs. Default: false.
1039 .. option:: randseed=int
1041 Seed the random number generators based on this seed value, to be able to
1042 control what sequence of output is being generated. If not set, the random
1043 sequence depends on the :option:`randrepeat` setting.
1045 .. option:: fallocate=str
1047 Whether pre-allocation is performed when laying down files.
1048 Accepted values are:
1051 Do not pre-allocate space.
1054 Use a platform's native pre-allocation call but fall back to
1055 **none** behavior if it fails/is not implemented.
1058 Pre-allocate via :manpage:`posix_fallocate(3)`.
1061 Pre-allocate via :manpage:`fallocate(2)` with
1062 FALLOC_FL_KEEP_SIZE set.
1065 Backward-compatible alias for **none**.
1068 Backward-compatible alias for **posix**.
1070 May not be available on all supported platforms. **keep** is only available
1071 on Linux. If using ZFS on Solaris this cannot be set to **posix**
1072 because ZFS doesn't support pre-allocation. Default: **native** if any
1073 pre-allocation methods are available, **none** if not.
1075 .. option:: fadvise_hint=str
1077 Use :manpage:`posix_fadvise(2)` to advise the kernel on what I/O patterns
1078 are likely to be issued. Accepted values are:
1081 Backwards-compatible hint for "no hint".
1084 Backwards compatible hint for "advise with fio workload type". This
1085 uses **FADV_RANDOM** for a random workload, and **FADV_SEQUENTIAL**
1086 for a sequential workload.
1089 Advise using **FADV_SEQUENTIAL**.
1092 Advise using **FADV_RANDOM**.
1094 .. option:: write_hint=str
1096 Use :manpage:`fcntl(2)` to advise the kernel what life time to expect
1097 from a write. Only supported on Linux, as of version 4.13. Accepted
1101 No particular life time associated with this file.
1104 Data written to this file has a short life time.
1107 Data written to this file has a medium life time.
1110 Data written to this file has a long life time.
1113 Data written to this file has a very long life time.
1115 The values are all relative to each other, and no absolute meaning
1116 should be associated with them.
1118 .. option:: offset=int
1120 Start I/O at the provided offset in the file, given as either a fixed size in
1121 bytes or a percentage. If a percentage is given, the next ``blockalign``-ed
1122 offset will be used. Data before the given offset will not be touched. This
1123 effectively caps the file size at `real_size - offset`. Can be combined with
1124 :option:`size` to constrain the start and end range of the I/O workload.
1125 A percentage can be specified by a number between 1 and 100 followed by '%',
1126 for example, ``offset=20%`` to specify 20%.
1128 .. option:: offset_increment=int
1130 If this is provided, then the real offset becomes `offset + offset_increment
1131 * thread_number`, where the thread number is a counter that starts at 0 and
1132 is incremented for each sub-job (i.e. when :option:`numjobs` option is
1133 specified). This option is useful if there are several jobs which are
1134 intended to operate on a file in parallel disjoint segments, with even
1135 spacing between the starting points.
1137 .. option:: number_ios=int
1139 Fio will normally perform I/Os until it has exhausted the size of the region
1140 set by :option:`size`, or if it exhaust the allocated time (or hits an error
1141 condition). With this setting, the range/size can be set independently of
1142 the number of I/Os to perform. When fio reaches this number, it will exit
1143 normally and report status. Note that this does not extend the amount of I/O
1144 that will be done, it will only stop fio if this condition is met before
1145 other end-of-job criteria.
1147 .. option:: fsync=int
1149 If writing to a file, issue an :manpage:`fsync(2)` (or its equivalent) of
1150 the dirty data for every number of blocks given. For example, if you give 32
1151 as a parameter, fio will sync the file after every 32 writes issued. If fio is
1152 using non-buffered I/O, we may not sync the file. The exception is the sg
1153 I/O engine, which synchronizes the disk cache anyway. Defaults to 0, which
1154 means fio does not periodically issue and wait for a sync to complete. Also
1155 see :option:`end_fsync` and :option:`fsync_on_close`.
1157 .. option:: fdatasync=int
1159 Like :option:`fsync` but uses :manpage:`fdatasync(2)` to only sync data and
1160 not metadata blocks. In Windows, FreeBSD, and DragonFlyBSD there is no
1161 :manpage:`fdatasync(2)` so this falls back to using :manpage:`fsync(2)`.
1162 Defaults to 0, which means fio does not periodically issue and wait for a
1163 data-only sync to complete.
1165 .. option:: write_barrier=int
1167 Make every `N-th` write a barrier write.
1169 .. option:: sync_file_range=str:int
1171 Use :manpage:`sync_file_range(2)` for every `int` number of write
1172 operations. Fio will track range of writes that have happened since the last
1173 :manpage:`sync_file_range(2)` call. `str` can currently be one or more of:
1176 SYNC_FILE_RANGE_WAIT_BEFORE
1178 SYNC_FILE_RANGE_WRITE
1180 SYNC_FILE_RANGE_WAIT_AFTER
1182 So if you do ``sync_file_range=wait_before,write:8``, fio would use
1183 ``SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE`` for every 8
1184 writes. Also see the :manpage:`sync_file_range(2)` man page. This option is
1187 .. option:: overwrite=bool
1189 If true, writes to a file will always overwrite existing data. If the file
1190 doesn't already exist, it will be created before the write phase begins. If
1191 the file exists and is large enough for the specified write phase, nothing
1192 will be done. Default: false.
1194 .. option:: end_fsync=bool
1196 If true, :manpage:`fsync(2)` file contents when a write stage has completed.
1199 .. option:: fsync_on_close=bool
1201 If true, fio will :manpage:`fsync(2)` a dirty file on close. This differs
1202 from :option:`end_fsync` in that it will happen on every file close, not
1203 just at the end of the job. Default: false.
1205 .. option:: rwmixread=int
1207 Percentage of a mixed workload that should be reads. Default: 50.
1209 .. option:: rwmixwrite=int
1211 Percentage of a mixed workload that should be writes. If both
1212 :option:`rwmixread` and :option:`rwmixwrite` is given and the values do not
1213 add up to 100%, the latter of the two will be used to override the
1214 first. This may interfere with a given rate setting, if fio is asked to
1215 limit reads or writes to a certain rate. If that is the case, then the
1216 distribution may be skewed. Default: 50.
1218 .. option:: random_distribution=str:float[,str:float][,str:float]
1220 By default, fio will use a completely uniform random distribution when asked
1221 to perform random I/O. Sometimes it is useful to skew the distribution in
1222 specific ways, ensuring that some parts of the data is more hot than others.
1223 fio includes the following distribution models:
1226 Uniform random distribution
1235 Normal (Gaussian) distribution
1238 Zoned random distribution
1240 When using a **zipf** or **pareto** distribution, an input value is also
1241 needed to define the access pattern. For **zipf**, this is the `Zipf
1242 theta`. For **pareto**, it's the `Pareto power`. Fio includes a test
1243 program, :command:`fio-genzipf`, that can be used visualize what the given input
1244 values will yield in terms of hit rates. If you wanted to use **zipf** with
1245 a `theta` of 1.2, you would use ``random_distribution=zipf:1.2`` as the
1246 option. If a non-uniform model is used, fio will disable use of the random
1247 map. For the **normal** distribution, a normal (Gaussian) deviation is
1248 supplied as a value between 0 and 100.
1250 For a **zoned** distribution, fio supports specifying percentages of I/O
1251 access that should fall within what range of the file or device. For
1252 example, given a criteria of:
1254 * 60% of accesses should be to the first 10%
1255 * 30% of accesses should be to the next 20%
1256 * 8% of accesses should be to the next 30%
1257 * 2% of accesses should be to the next 40%
1259 we can define that through zoning of the random accesses. For the above
1260 example, the user would do::
1262 random_distribution=zoned:60/10:30/20:8/30:2/40
1264 similarly to how :option:`bssplit` works for setting ranges and percentages
1265 of block sizes. Like :option:`bssplit`, it's possible to specify separate
1266 zones for reads, writes, and trims. If just one set is given, it'll apply to
1269 .. option:: percentage_random=int[,int][,int]
1271 For a random workload, set how big a percentage should be random. This
1272 defaults to 100%, in which case the workload is fully random. It can be set
1273 from anywhere from 0 to 100. Setting it to 0 would make the workload fully
1274 sequential. Any setting in between will result in a random mix of sequential
1275 and random I/O, at the given percentages. Comma-separated values may be
1276 specified for reads, writes, and trims as described in :option:`blocksize`.
1278 .. option:: norandommap
1280 Normally fio will cover every block of the file when doing random I/O. If
1281 this option is given, fio will just get a new random offset without looking
1282 at past I/O history. This means that some blocks may not be read or written,
1283 and that some blocks may be read/written more than once. If this option is
1284 used with :option:`verify` and multiple blocksizes (via :option:`bsrange`),
1285 only intact blocks are verified, i.e., partially-overwritten blocks are
1288 .. option:: softrandommap=bool
1290 See :option:`norandommap`. If fio runs with the random block map enabled and
1291 it fails to allocate the map, if this option is set it will continue without
1292 a random block map. As coverage will not be as complete as with random maps,
1293 this option is disabled by default.
1295 .. option:: random_generator=str
1297 Fio supports the following engines for generating I/O offsets for random I/O:
1300 Strong 2^88 cycle random number generator.
1302 Linear feedback shift register generator.
1304 Strong 64-bit 2^258 cycle random number generator.
1306 **tausworthe** is a strong random number generator, but it requires tracking
1307 on the side if we want to ensure that blocks are only read or written
1308 once. **lfsr** guarantees that we never generate the same offset twice, and
1309 it's also less computationally expensive. It's not a true random generator,
1310 however, though for I/O purposes it's typically good enough. **lfsr** only
1311 works with single block sizes, not with workloads that use multiple block
1312 sizes. If used with such a workload, fio may read or write some blocks
1313 multiple times. The default value is **tausworthe**, unless the required
1314 space exceeds 2^32 blocks. If it does, then **tausworthe64** is
1315 selected automatically.
1321 .. option:: blocksize=int[,int][,int], bs=int[,int][,int]
1323 The block size in bytes used for I/O units. Default: 4096. A single value
1324 applies to reads, writes, and trims. Comma-separated values may be
1325 specified for reads, writes, and trims. A value not terminated in a comma
1326 applies to subsequent types.
1331 means 256k for reads, writes and trims.
1334 means 8k for reads, 32k for writes and trims.
1337 means 8k for reads, 32k for writes, and default for trims.
1340 means default for reads, 8k for writes and trims.
1343 means default for reads, 8k for writes, and default for trims.
1345 .. option:: blocksize_range=irange[,irange][,irange], bsrange=irange[,irange][,irange]
1347 A range of block sizes in bytes for I/O units. The issued I/O unit will
1348 always be a multiple of the minimum size, unless
1349 :option:`blocksize_unaligned` is set.
1351 Comma-separated ranges may be specified for reads, writes, and trims as
1352 described in :option:`blocksize`.
1354 Example: ``bsrange=1k-4k,2k-8k``.
1356 .. option:: bssplit=str[,str][,str]
1358 Sometimes you want even finer grained control of the block sizes issued, not
1359 just an even split between them. This option allows you to weight various
1360 block sizes, so that you are able to define a specific amount of block sizes
1361 issued. The format for this option is::
1363 bssplit=blocksize/percentage:blocksize/percentage
1365 for as many block sizes as needed. So if you want to define a workload that
1366 has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would write::
1368 bssplit=4k/10:64k/50:32k/40
1370 Ordering does not matter. If the percentage is left blank, fio will fill in
1371 the remaining values evenly. So a bssplit option like this one::
1373 bssplit=4k/50:1k/:32k/
1375 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always add up
1376 to 100, if bssplit is given a range that adds up to more, it will error out.
1378 Comma-separated values may be specified for reads, writes, and trims as
1379 described in :option:`blocksize`.
1381 If you want a workload that has 50% 2k reads and 50% 4k reads, while having
1382 90% 4k writes and 10% 8k writes, you would specify::
1384 bssplit=2k/50:4k/50,4k/90,8k/10
1386 .. option:: blocksize_unaligned, bs_unaligned
1388 If set, fio will issue I/O units with any size within
1389 :option:`blocksize_range`, not just multiples of the minimum size. This
1390 typically won't work with direct I/O, as that normally requires sector
1393 .. option:: bs_is_seq_rand=bool
1395 If this option is set, fio will use the normal read,write blocksize settings
1396 as sequential,random blocksize settings instead. Any random read or write
1397 will use the WRITE blocksize settings, and any sequential read or write will
1398 use the READ blocksize settings.
1400 .. option:: blockalign=int[,int][,int], ba=int[,int][,int]
1402 Boundary to which fio will align random I/O units. Default:
1403 :option:`blocksize`. Minimum alignment is typically 512b for using direct
1404 I/O, though it usually depends on the hardware block size. This option is
1405 mutually exclusive with using a random map for files, so it will turn off
1406 that option. Comma-separated values may be specified for reads, writes, and
1407 trims as described in :option:`blocksize`.
1413 .. option:: zero_buffers
1415 Initialize buffers with all zeros. Default: fill buffers with random data.
1417 .. option:: refill_buffers
1419 If this option is given, fio will refill the I/O buffers on every
1420 submit. The default is to only fill it at init time and reuse that
1421 data. Only makes sense if zero_buffers isn't specified, naturally. If data
1422 verification is enabled, `refill_buffers` is also automatically enabled.
1424 .. option:: scramble_buffers=bool
1426 If :option:`refill_buffers` is too costly and the target is using data
1427 deduplication, then setting this option will slightly modify the I/O buffer
1428 contents to defeat normal de-dupe attempts. This is not enough to defeat
1429 more clever block compression attempts, but it will stop naive dedupe of
1430 blocks. Default: true.
1432 .. option:: buffer_compress_percentage=int
1434 If this is set, then fio will attempt to provide I/O buffer content (on
1435 WRITEs) that compresses to the specified level. Fio does this by providing a
1436 mix of random data and a fixed pattern. The fixed pattern is either zeros,
1437 or the pattern specified by :option:`buffer_pattern`. If the pattern option
1438 is used, it might skew the compression ratio slightly. Note that this is per
1439 block size unit, for file/disk wide compression level that matches this
1440 setting, you'll also want to set :option:`refill_buffers`.
1442 .. option:: buffer_compress_chunk=int
1444 See :option:`buffer_compress_percentage`. This setting allows fio to manage
1445 how big the ranges of random data and zeroed data is. Without this set, fio
1446 will provide :option:`buffer_compress_percentage` of blocksize random data,
1447 followed by the remaining zeroed. With this set to some chunk size smaller
1448 than the block size, fio can alternate random and zeroed data throughout the
1451 .. option:: buffer_pattern=str
1453 If set, fio will fill the I/O buffers with this pattern or with the contents
1454 of a file. If not set, the contents of I/O buffers are defined by the other
1455 options related to buffer contents. The setting can be any pattern of bytes,
1456 and can be prefixed with 0x for hex values. It may also be a string, where
1457 the string must then be wrapped with ``""``. Or it may also be a filename,
1458 where the filename must be wrapped with ``''`` in which case the file is
1459 opened and read. Note that not all the file contents will be read if that
1460 would cause the buffers to overflow. So, for example::
1462 buffer_pattern='filename'
1466 buffer_pattern="abcd"
1474 buffer_pattern=0xdeadface
1476 Also you can combine everything together in any order::
1478 buffer_pattern=0xdeadface"abcd"-12'filename'
1480 .. option:: dedupe_percentage=int
1482 If set, fio will generate this percentage of identical buffers when
1483 writing. These buffers will be naturally dedupable. The contents of the
1484 buffers depend on what other buffer compression settings have been set. It's
1485 possible to have the individual buffers either fully compressible, or not at
1486 all. This option only controls the distribution of unique buffers.
1488 .. option:: invalidate=bool
1490 Invalidate the buffer/page cache parts of the files to be used prior to
1491 starting I/O if the platform and file type support it. Defaults to true.
1492 This will be ignored if :option:`pre_read` is also specified for the
1495 .. option:: sync=bool
1497 Use synchronous I/O for buffered writes. For the majority of I/O engines,
1498 this means using O_SYNC. Default: false.
1500 .. option:: iomem=str, mem=str
1502 Fio can use various types of memory as the I/O unit buffer. The allowed
1506 Use memory from :manpage:`malloc(3)` as the buffers. Default memory
1510 Use shared memory as the buffers. Allocated through
1511 :manpage:`shmget(2)`.
1514 Same as shm, but use huge pages as backing.
1517 Use :manpage:`mmap(2)` to allocate buffers. May either be anonymous memory, or can
1518 be file backed if a filename is given after the option. The format
1519 is `mem=mmap:/path/to/file`.
1522 Use a memory mapped huge file as the buffer backing. Append filename
1523 after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file`.
1526 Same as mmap, but use a MMAP_SHARED mapping.
1529 Use GPU memory as the buffers for GPUDirect RDMA benchmark.
1530 The :option:`ioengine` must be `rdma`.
1532 The area allocated is a function of the maximum allowed bs size for the job,
1533 multiplied by the I/O depth given. Note that for **shmhuge** and
1534 **mmaphuge** to work, the system must have free huge pages allocated. This
1535 can normally be checked and set by reading/writing
1536 :file:`/proc/sys/vm/nr_hugepages` on a Linux system. Fio assumes a huge page
1537 is 4MiB in size. So to calculate the number of huge pages you need for a
1538 given job file, add up the I/O depth of all jobs (normally one unless
1539 :option:`iodepth` is used) and multiply by the maximum bs set. Then divide
1540 that number by the huge page size. You can see the size of the huge pages in
1541 :file:`/proc/meminfo`. If no huge pages are allocated by having a non-zero
1542 number in `nr_hugepages`, using **mmaphuge** or **shmhuge** will fail. Also
1543 see :option:`hugepage-size`.
1545 **mmaphuge** also needs to have hugetlbfs mounted and the file location
1546 should point there. So if it's mounted in :file:`/huge`, you would use
1547 `mem=mmaphuge:/huge/somefile`.
1549 .. option:: iomem_align=int, mem_align=int
1551 This indicates the memory alignment of the I/O memory buffers. Note that
1552 the given alignment is applied to the first I/O unit buffer, if using
1553 :option:`iodepth` the alignment of the following buffers are given by the
1554 :option:`bs` used. In other words, if using a :option:`bs` that is a
1555 multiple of the page sized in the system, all buffers will be aligned to
1556 this value. If using a :option:`bs` that is not page aligned, the alignment
1557 of subsequent I/O memory buffers is the sum of the :option:`iomem_align` and
1560 .. option:: hugepage-size=int
1562 Defines the size of a huge page. Must at least be equal to the system
1563 setting, see :file:`/proc/meminfo`. Defaults to 4MiB. Should probably
1564 always be a multiple of megabytes, so using ``hugepage-size=Xm`` is the
1565 preferred way to set this to avoid setting a non-pow-2 bad value.
1567 .. option:: lockmem=int
1569 Pin the specified amount of memory with :manpage:`mlock(2)`. Can be used to
1570 simulate a smaller amount of memory. The amount specified is per worker.
1576 .. option:: size=int
1578 The total size of file I/O for each thread of this job. Fio will run until
1579 this many bytes has been transferred, unless runtime is limited by other options
1580 (such as :option:`runtime`, for instance, or increased/decreased by :option:`io_size`).
1581 Fio will divide this size between the available files determined by options
1582 such as :option:`nrfiles`, :option:`filename`, unless :option:`filesize` is
1583 specified by the job. If the result of division happens to be 0, the size is
1584 set to the physical size of the given files or devices if they exist.
1585 If this option is not specified, fio will use the full size of the given
1586 files or devices. If the files do not exist, size must be given. It is also
1587 possible to give size as a percentage between 1 and 100. If ``size=20%`` is
1588 given, fio will use 20% of the full size of the given files or devices.
1589 Can be combined with :option:`offset` to constrain the start and end range
1590 that I/O will be done within.
1592 .. option:: io_size=int, io_limit=int
1594 Normally fio operates within the region set by :option:`size`, which means
1595 that the :option:`size` option sets both the region and size of I/O to be
1596 performed. Sometimes that is not what you want. With this option, it is
1597 possible to define just the amount of I/O that fio should do. For instance,
1598 if :option:`size` is set to 20GiB and :option:`io_size` is set to 5GiB, fio
1599 will perform I/O within the first 20GiB but exit when 5GiB have been
1600 done. The opposite is also possible -- if :option:`size` is set to 20GiB,
1601 and :option:`io_size` is set to 40GiB, then fio will do 40GiB of I/O within
1602 the 0..20GiB region.
1604 .. option:: filesize=irange(int)
1606 Individual file sizes. May be a range, in which case fio will select sizes
1607 for files at random within the given range and limited to :option:`size` in
1608 total (if that is given). If not given, each created file is the same size.
1609 This option overrides :option:`size` in terms of file size, which means
1610 this value is used as a fixed size or possible range of each file.
1612 .. option:: file_append=bool
1614 Perform I/O after the end of the file. Normally fio will operate within the
1615 size of a file. If this option is set, then fio will append to the file
1616 instead. This has identical behavior to setting :option:`offset` to the size
1617 of a file. This option is ignored on non-regular files.
1619 .. option:: fill_device=bool, fill_fs=bool
1621 Sets size to something really large and waits for ENOSPC (no space left on
1622 device) as the terminating condition. Only makes sense with sequential
1623 write. For a read workload, the mount point will be filled first then I/O
1624 started on the result. This option doesn't make sense if operating on a raw
1625 device node, since the size of that is already known by the file system.
1626 Additionally, writing beyond end-of-device will not return ENOSPC there.
1632 .. option:: ioengine=str
1634 Defines how the job issues I/O to the file. The following types are defined:
1637 Basic :manpage:`read(2)` or :manpage:`write(2)`
1638 I/O. :manpage:`lseek(2)` is used to position the I/O location.
1639 See :option:`fsync` and :option:`fdatasync` for syncing write I/Os.
1642 Basic :manpage:`pread(2)` or :manpage:`pwrite(2)` I/O. Default on
1643 all supported operating systems except for Windows.
1646 Basic :manpage:`readv(2)` or :manpage:`writev(2)` I/O. Will emulate
1647 queuing by coalescing adjacent I/Os into a single submission.
1650 Basic :manpage:`preadv(2)` or :manpage:`pwritev(2)` I/O.
1653 Basic :manpage:`preadv2(2)` or :manpage:`pwritev2(2)` I/O.
1656 Linux native asynchronous I/O. Note that Linux may only support
1657 queued behavior with non-buffered I/O (set ``direct=1`` or
1659 This engine defines engine specific options.
1662 POSIX asynchronous I/O using :manpage:`aio_read(3)` and
1663 :manpage:`aio_write(3)`.
1666 Solaris native asynchronous I/O.
1669 Windows native asynchronous I/O. Default on Windows.
1672 File is memory mapped with :manpage:`mmap(2)` and data copied
1673 to/from using :manpage:`memcpy(3)`.
1676 :manpage:`splice(2)` is used to transfer the data and
1677 :manpage:`vmsplice(2)` to transfer data from user space to the
1681 SCSI generic sg v3 I/O. May either be synchronous using the SG_IO
1682 ioctl, or if the target is an sg character device we use
1683 :manpage:`read(2)` and :manpage:`write(2)` for asynchronous
1684 I/O. Requires :option:`filename` option to specify either block or
1688 Doesn't transfer any data, just pretends to. This is mainly used to
1689 exercise fio itself and for debugging/testing purposes.
1692 Transfer over the network to given ``host:port``. Depending on the
1693 :option:`protocol` used, the :option:`hostname`, :option:`port`,
1694 :option:`listen` and :option:`filename` options are used to specify
1695 what sort of connection to make, while the :option:`protocol` option
1696 determines which protocol will be used. This engine defines engine
1700 Like **net**, but uses :manpage:`splice(2)` and
1701 :manpage:`vmsplice(2)` to map data and send/receive.
1702 This engine defines engine specific options.
1705 Doesn't transfer any data, but burns CPU cycles according to the
1706 :option:`cpuload` and :option:`cpuchunks` options. Setting
1707 :option:`cpuload`\=85 will cause that job to do nothing but burn 85%
1708 of the CPU. In case of SMP machines, use :option:`numjobs`=<nr_of_cpu>
1709 to get desired CPU usage, as the cpuload only loads a
1710 single CPU at the desired rate. A job never finishes unless there is
1711 at least one non-cpuio job.
1714 The GUASI I/O engine is the Generic Userspace Asyncronous Syscall
1715 Interface approach to async I/O. See
1717 http://www.xmailserver.org/guasi-lib.html
1719 for more info on GUASI.
1722 The RDMA I/O engine supports both RDMA memory semantics
1723 (RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the
1724 InfiniBand, RoCE and iWARP protocols.
1727 I/O engine that does regular fallocate to simulate data transfer as
1731 does fallocate(,mode = FALLOC_FL_KEEP_SIZE,).
1734 does fallocate(,mode = 0).
1737 does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE).
1740 I/O engine that sends :manpage:`ftruncate(2)` operations in response
1741 to write (DDIR_WRITE) events. Each ftruncate issued sets the file's
1742 size to the current block offset. :option:`blocksize` is ignored.
1745 I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate
1746 defragment activity in request to DDIR_WRITE event.
1749 I/O engine supporting direct access to Ceph Rados Block Devices
1750 (RBD) via librbd without the need to use the kernel rbd driver. This
1751 ioengine defines engine specific options.
1754 Using GlusterFS libgfapi sync interface to direct access to
1755 GlusterFS volumes without having to go through FUSE. This ioengine
1756 defines engine specific options.
1759 Using GlusterFS libgfapi async interface to direct access to
1760 GlusterFS volumes without having to go through FUSE. This ioengine
1761 defines engine specific options.
1764 Read and write through Hadoop (HDFS). The :option:`filename` option
1765 is used to specify host,port of the hdfs name-node to connect. This
1766 engine interprets offsets a little differently. In HDFS, files once
1767 created cannot be modified so random writes are not possible. To
1768 imitate this the libhdfs engine expects a bunch of small files to be
1769 created over HDFS and will randomly pick a file from them
1770 based on the offset generated by fio backend (see the example
1771 job file to create such files, use ``rw=write`` option). Please
1772 note, it may be necessary to set environment variables to work
1773 with HDFS/libhdfs properly. Each job uses its own connection to
1777 Read, write and erase an MTD character device (e.g.,
1778 :file:`/dev/mtd0`). Discards are treated as erases. Depending on the
1779 underlying device type, the I/O may have to go in a certain pattern,
1780 e.g., on NAND, writing sequentially to erase blocks and discarding
1781 before overwriting. The `trimwrite` mode works well for this
1785 Read and write using filesystem DAX to a file on a filesystem
1786 mounted with DAX on a persistent memory device through the NVML
1790 Read and write using device DAX to a persistent memory device (e.g.,
1791 /dev/dax0.0) through the NVML libpmem library.
1794 Prefix to specify loading an external I/O engine object file. Append
1795 the engine filename, e.g. ``ioengine=external:/tmp/foo.o`` to load
1796 ioengine :file:`foo.o` in :file:`/tmp`. The path can be either
1797 absolute or relative. See :file:`engines/skeleton_external.c` for
1798 details of writing an external I/O engine.
1801 Simply create the files and do no IO to them. You still need to
1802 set `filesize` so that all the accounting still occurs, but no
1803 actual IO will be done other than creating the file.
1805 I/O engine specific parameters
1806 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1808 In addition, there are some parameters which are only valid when a specific
1809 :option:`ioengine` is in use. These are used identically to normal parameters,
1810 with the caveat that when used on the command line, they must come after the
1811 :option:`ioengine` that defines them is selected.
1813 .. option:: userspace_reap : [libaio]
1815 Normally, with the libaio engine in use, fio will use the
1816 :manpage:`io_getevents(2)` system call to reap newly returned events. With
1817 this flag turned on, the AIO ring will be read directly from user-space to
1818 reap events. The reaping mode is only enabled when polling for a minimum of
1819 0 events (e.g. when :option:`iodepth_batch_complete` `=0`).
1821 .. option:: hipri : [pvsync2]
1823 Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority
1826 .. option:: hipri_percentage : [pvsync2]
1828 When hipri is set this determines the probability of a pvsync2 I/O being high
1829 priority. The default is 100%.
1831 .. option:: cpuload=int : [cpuio]
1833 Attempt to use the specified percentage of CPU cycles. This is a mandatory
1834 option when using cpuio I/O engine.
1836 .. option:: cpuchunks=int : [cpuio]
1838 Split the load into cycles of the given time. In microseconds.
1840 .. option:: exit_on_io_done=bool : [cpuio]
1842 Detect when I/O threads are done, then exit.
1844 .. option:: namenode=str : [libhdfs]
1846 The hostname or IP address of a HDFS cluster namenode to contact.
1848 .. option:: port=int
1852 The listening port of the HFDS cluster namenode.
1856 The TCP or UDP port to bind to or connect to. If this is used with
1857 :option:`numjobs` to spawn multiple instances of the same job type, then
1858 this will be the starting port number since fio will use a range of
1861 .. option:: hostname=str : [netsplice] [net]
1863 The hostname or IP address to use for TCP or UDP based I/O. If the job is
1864 a TCP listener or UDP reader, the hostname is not used and must be omitted
1865 unless it is a valid UDP multicast address.
1867 .. option:: interface=str : [netsplice] [net]
1869 The IP address of the network interface used to send or receive UDP
1872 .. option:: ttl=int : [netsplice] [net]
1874 Time-to-live value for outgoing UDP multicast packets. Default: 1.
1876 .. option:: nodelay=bool : [netsplice] [net]
1878 Set TCP_NODELAY on TCP connections.
1880 .. option:: protocol=str, proto=str : [netsplice] [net]
1882 The network protocol to use. Accepted values are:
1885 Transmission control protocol.
1887 Transmission control protocol V6.
1889 User datagram protocol.
1891 User datagram protocol V6.
1895 When the protocol is TCP or UDP, the port must also be given, as well as the
1896 hostname if the job is a TCP listener or UDP reader. For unix sockets, the
1897 normal :option:`filename` option should be used and the port is invalid.
1899 .. option:: listen : [netsplice] [net]
1901 For TCP network connections, tell fio to listen for incoming connections
1902 rather than initiating an outgoing connection. The :option:`hostname` must
1903 be omitted if this option is used.
1905 .. option:: pingpong : [netsplice] [net]
1907 Normally a network writer will just continue writing data, and a network
1908 reader will just consume packages. If ``pingpong=1`` is set, a writer will
1909 send its normal payload to the reader, then wait for the reader to send the
1910 same payload back. This allows fio to measure network latencies. The
1911 submission and completion latencies then measure local time spent sending or
1912 receiving, and the completion latency measures how long it took for the
1913 other end to receive and send back. For UDP multicast traffic
1914 ``pingpong=1`` should only be set for a single reader when multiple readers
1915 are listening to the same address.
1917 .. option:: window_size : [netsplice] [net]
1919 Set the desired socket buffer size for the connection.
1921 .. option:: mss : [netsplice] [net]
1923 Set the TCP maximum segment size (TCP_MAXSEG).
1925 .. option:: donorname=str : [e4defrag]
1927 File will be used as a block donor (swap extents between files).
1929 .. option:: inplace=int : [e4defrag]
1931 Configure donor file blocks allocation strategy:
1934 Default. Preallocate donor's file on init.
1936 Allocate space immediately inside defragment event, and free right
1939 .. option:: clustername=str : [rbd]
1941 Specifies the name of the Ceph cluster.
1943 .. option:: rbdname=str : [rbd]
1945 Specifies the name of the RBD.
1947 .. option:: pool=str : [rbd]
1949 Specifies the name of the Ceph pool containing RBD.
1951 .. option:: clientname=str : [rbd]
1953 Specifies the username (without the 'client.' prefix) used to access the
1954 Ceph cluster. If the *clustername* is specified, the *clientname* shall be
1955 the full *type.id* string. If no type. prefix is given, fio will add
1956 'client.' by default.
1958 .. option:: skip_bad=bool : [mtd]
1960 Skip operations against known bad blocks.
1962 .. option:: hdfsdirectory : [libhdfs]
1964 libhdfs will create chunk in this HDFS directory.
1966 .. option:: chunk_size : [libhdfs]
1968 The size of the chunk to use for each file.
1974 .. option:: iodepth=int
1976 Number of I/O units to keep in flight against the file. Note that
1977 increasing *iodepth* beyond 1 will not affect synchronous ioengines (except
1978 for small degrees when :option:`verify_async` is in use). Even async
1979 engines may impose OS restrictions causing the desired depth not to be
1980 achieved. This may happen on Linux when using libaio and not setting
1981 :option:`direct`\=1, since buffered I/O is not async on that OS. Keep an
1982 eye on the I/O depth distribution in the fio output to verify that the
1983 achieved depth is as expected. Default: 1.
1985 .. option:: iodepth_batch_submit=int, iodepth_batch=int
1987 This defines how many pieces of I/O to submit at once. It defaults to 1
1988 which means that we submit each I/O as soon as it is available, but can be
1989 raised to submit bigger batches of I/O at the time. If it is set to 0 the
1990 :option:`iodepth` value will be used.
1992 .. option:: iodepth_batch_complete_min=int, iodepth_batch_complete=int
1994 This defines how many pieces of I/O to retrieve at once. It defaults to 1
1995 which means that we'll ask for a minimum of 1 I/O in the retrieval process
1996 from the kernel. The I/O retrieval will go on until we hit the limit set by
1997 :option:`iodepth_low`. If this variable is set to 0, then fio will always
1998 check for completed events before queuing more I/O. This helps reduce I/O
1999 latency, at the cost of more retrieval system calls.
2001 .. option:: iodepth_batch_complete_max=int
2003 This defines maximum pieces of I/O to retrieve at once. This variable should
2004 be used along with :option:`iodepth_batch_complete_min`\=int variable,
2005 specifying the range of min and max amount of I/O which should be
2006 retrieved. By default it is equal to the :option:`iodepth_batch_complete_min`
2011 iodepth_batch_complete_min=1
2012 iodepth_batch_complete_max=<iodepth>
2014 which means that we will retrieve at least 1 I/O and up to the whole
2015 submitted queue depth. If none of I/O has been completed yet, we will wait.
2019 iodepth_batch_complete_min=0
2020 iodepth_batch_complete_max=<iodepth>
2022 which means that we can retrieve up to the whole submitted queue depth, but
2023 if none of I/O has been completed yet, we will NOT wait and immediately exit
2024 the system call. In this example we simply do polling.
2026 .. option:: iodepth_low=int
2028 The low water mark indicating when to start filling the queue
2029 again. Defaults to the same as :option:`iodepth`, meaning that fio will
2030 attempt to keep the queue full at all times. If :option:`iodepth` is set to
2031 e.g. 16 and *iodepth_low* is set to 4, then after fio has filled the queue of
2032 16 requests, it will let the depth drain down to 4 before starting to fill
2035 .. option:: serialize_overlap=bool
2037 Serialize in-flight I/Os that might otherwise cause or suffer from data races.
2038 When two or more I/Os are submitted simultaneously, there is no guarantee that
2039 the I/Os will be processed or completed in the submitted order. Further, if
2040 two or more of those I/Os are writes, any overlapping region between them can
2041 become indeterminate/undefined on certain storage. These issues can cause
2042 verification to fail erratically when at least one of the racing I/Os is
2043 changing data and the overlapping region has a non-zero size. Setting
2044 ``serialize_overlap`` tells fio to avoid provoking this behavior by explicitly
2045 serializing in-flight I/Os that have a non-zero overlap. Note that setting
2046 this option can reduce both performance and the `:option:iodepth` achieved.
2047 Additionally this option does not work when :option:`io_submit_mode` is set to
2048 offload. Default: false.
2050 .. option:: io_submit_mode=str
2052 This option controls how fio submits the I/O to the I/O engine. The default
2053 is `inline`, which means that the fio job threads submit and reap I/O
2054 directly. If set to `offload`, the job threads will offload I/O submission
2055 to a dedicated pool of I/O threads. This requires some coordination and thus
2056 has a bit of extra overhead, especially for lower queue depth I/O where it
2057 can increase latencies. The benefit is that fio can manage submission rates
2058 independently of the device completion rates. This avoids skewed latency
2059 reporting if I/O gets backed up on the device side (the coordinated omission
2066 .. option:: thinktime=time
2068 Stall the job for the specified period of time after an I/O has completed before issuing the
2069 next. May be used to simulate processing being done by an application.
2070 When the unit is omitted, the value is interpreted in microseconds. See
2071 :option:`thinktime_blocks` and :option:`thinktime_spin`.
2073 .. option:: thinktime_spin=time
2075 Only valid if :option:`thinktime` is set - pretend to spend CPU time doing
2076 something with the data received, before falling back to sleeping for the
2077 rest of the period specified by :option:`thinktime`. When the unit is
2078 omitted, the value is interpreted in microseconds.
2080 .. option:: thinktime_blocks=int
2082 Only valid if :option:`thinktime` is set - control how many blocks to issue,
2083 before waiting :option:`thinktime` usecs. If not set, defaults to 1 which will make
2084 fio wait :option:`thinktime` usecs after every block. This effectively makes any
2085 queue depth setting redundant, since no more than 1 I/O will be queued
2086 before we have to complete it and do our :option:`thinktime`. In other words, this
2087 setting effectively caps the queue depth if the latter is larger.
2089 .. option:: rate=int[,int][,int]
2091 Cap the bandwidth used by this job. The number is in bytes/sec, the normal
2092 suffix rules apply. Comma-separated values may be specified for reads,
2093 writes, and trims as described in :option:`blocksize`.
2095 For example, using `rate=1m,500k` would limit reads to 1MiB/sec and writes to
2096 500KiB/sec. Capping only reads or writes can be done with `rate=,500k` or
2097 `rate=500k,` where the former will only limit writes (to 500KiB/sec) and the
2098 latter will only limit reads.
2100 .. option:: rate_min=int[,int][,int]
2102 Tell fio to do whatever it can to maintain at least this bandwidth. Failing
2103 to meet this requirement will cause the job to exit. Comma-separated values
2104 may be specified for reads, writes, and trims as described in
2105 :option:`blocksize`.
2107 .. option:: rate_iops=int[,int][,int]
2109 Cap the bandwidth to this number of IOPS. Basically the same as
2110 :option:`rate`, just specified independently of bandwidth. If the job is
2111 given a block size range instead of a fixed value, the smallest block size
2112 is used as the metric. Comma-separated values may be specified for reads,
2113 writes, and trims as described in :option:`blocksize`.
2115 .. option:: rate_iops_min=int[,int][,int]
2117 If fio doesn't meet this rate of I/O, it will cause the job to exit.
2118 Comma-separated values may be specified for reads, writes, and trims as
2119 described in :option:`blocksize`.
2121 .. option:: rate_process=str
2123 This option controls how fio manages rated I/O submissions. The default is
2124 `linear`, which submits I/O in a linear fashion with fixed delays between
2125 I/Os that gets adjusted based on I/O completion rates. If this is set to
2126 `poisson`, fio will submit I/O based on a more real world random request
2127 flow, known as the Poisson process
2128 (https://en.wikipedia.org/wiki/Poisson_point_process). The lambda will be
2129 10^6 / IOPS for the given workload.
2135 .. option:: latency_target=time
2137 If set, fio will attempt to find the max performance point that the given
2138 workload will run at while maintaining a latency below this target. When
2139 the unit is omitted, the value is interpreted in microseconds. See
2140 :option:`latency_window` and :option:`latency_percentile`.
2142 .. option:: latency_window=time
2144 Used with :option:`latency_target` to specify the sample window that the job
2145 is run at varying queue depths to test the performance. When the unit is
2146 omitted, the value is interpreted in microseconds.
2148 .. option:: latency_percentile=float
2150 The percentage of I/Os that must fall within the criteria specified by
2151 :option:`latency_target` and :option:`latency_window`. If not set, this
2152 defaults to 100.0, meaning that all I/Os must be equal or below to the value
2153 set by :option:`latency_target`.
2155 .. option:: max_latency=time
2157 If set, fio will exit the job with an ETIMEDOUT error if it exceeds this
2158 maximum latency. When the unit is omitted, the value is interpreted in
2161 .. option:: rate_cycle=int
2163 Average bandwidth for :option:`rate` and :option:`rate_min` over this number
2164 of milliseconds. Defaults to 1000.
2170 .. option:: write_iolog=str
2172 Write the issued I/O patterns to the specified file. See
2173 :option:`read_iolog`. Specify a separate file for each job, otherwise the
2174 iologs will be interspersed and the file may be corrupt.
2176 .. option:: read_iolog=str
2178 Open an iolog with the specified filename and replay the I/O patterns it
2179 contains. This can be used to store a workload and replay it sometime
2180 later. The iolog given may also be a blktrace binary file, which allows fio
2181 to replay a workload captured by :command:`blktrace`. See
2182 :manpage:`blktrace(8)` for how to capture such logging data. For blktrace
2183 replay, the file needs to be turned into a blkparse binary data file first
2184 (``blkparse <device> -o /dev/null -d file_for_fio.bin``).
2186 .. option:: replay_no_stall=bool
2188 When replaying I/O with :option:`read_iolog` the default behavior is to
2189 attempt to respect the timestamps within the log and replay them with the
2190 appropriate delay between IOPS. By setting this variable fio will not
2191 respect the timestamps and attempt to replay them as fast as possible while
2192 still respecting ordering. The result is the same I/O pattern to a given
2193 device, but different timings.
2195 .. option:: replay_redirect=str
2197 While replaying I/O patterns using :option:`read_iolog` the default behavior
2198 is to replay the IOPS onto the major/minor device that each IOP was recorded
2199 from. This is sometimes undesirable because on a different machine those
2200 major/minor numbers can map to a different device. Changing hardware on the
2201 same system can also result in a different major/minor mapping.
2202 ``replay_redirect`` causes all I/Os to be replayed onto the single specified
2203 device regardless of the device it was recorded
2204 from. i.e. :option:`replay_redirect`\= :file:`/dev/sdc` would cause all I/O
2205 in the blktrace or iolog to be replayed onto :file:`/dev/sdc`. This means
2206 multiple devices will be replayed onto a single device, if the trace
2207 contains multiple devices. If you want multiple devices to be replayed
2208 concurrently to multiple redirected devices you must blkparse your trace
2209 into separate traces and replay them with independent fio invocations.
2210 Unfortunately this also breaks the strict time ordering between multiple
2213 .. option:: replay_align=int
2215 Force alignment of I/O offsets and lengths in a trace to this power of 2
2218 .. option:: replay_scale=int
2220 Scale sector offsets down by this factor when replaying traces.
2223 Threads, processes and job synchronization
2224 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2228 Fio defaults to creating jobs by using fork, however if this option is
2229 given, fio will create jobs by using POSIX Threads' function
2230 :manpage:`pthread_create(3)` to create threads instead.
2232 .. option:: wait_for=str
2234 If set, the current job won't be started until all workers of the specified
2235 waitee job are done.
2237 ``wait_for`` operates on the job name basis, so there are a few
2238 limitations. First, the waitee must be defined prior to the waiter job
2239 (meaning no forward references). Second, if a job is being referenced as a
2240 waitee, it must have a unique name (no duplicate waitees).
2242 .. option:: nice=int
2244 Run the job with the given nice value. See man :manpage:`nice(2)`.
2246 On Windows, values less than -15 set the process class to "High"; -1 through
2247 -15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle"
2250 .. option:: prio=int
2252 Set the I/O priority value of this job. Linux limits us to a positive value
2253 between 0 and 7, with 0 being the highest. See man
2254 :manpage:`ionice(1)`. Refer to an appropriate manpage for other operating
2255 systems since meaning of priority may differ.
2257 .. option:: prioclass=int
2259 Set the I/O priority class. See man :manpage:`ionice(1)`.
2261 .. option:: cpumask=int
2263 Set the CPU affinity of this job. The parameter given is a bit mask of
2264 allowed CPUs the job may run on. So if you want the allowed CPUs to be 1
2265 and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man
2266 :manpage:`sched_setaffinity(2)`. This may not work on all supported
2267 operating systems or kernel versions. This option doesn't work well for a
2268 higher CPU count than what you can store in an integer mask, so it can only
2269 control cpus 1-32. For boxes with larger CPU counts, use
2270 :option:`cpus_allowed`.
2272 .. option:: cpus_allowed=str
2274 Controls the same options as :option:`cpumask`, but accepts a textual
2275 specification of the permitted CPUs instead. So to use CPUs 1 and 5 you
2276 would specify ``cpus_allowed=1,5``. This option also allows a range of CPUs
2277 to be specified -- say you wanted a binding to CPUs 1, 5, and 8 to 15, you
2278 would set ``cpus_allowed=1,5,8-15``.
2280 .. option:: cpus_allowed_policy=str
2282 Set the policy of how fio distributes the CPUs specified by
2283 :option:`cpus_allowed` or :option:`cpumask`. Two policies are supported:
2286 All jobs will share the CPU set specified.
2288 Each job will get a unique CPU from the CPU set.
2290 **shared** is the default behavior, if the option isn't specified. If
2291 **split** is specified, then fio will will assign one cpu per job. If not
2292 enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs
2295 .. option:: numa_cpu_nodes=str
2297 Set this job running on specified NUMA nodes' CPUs. The arguments allow
2298 comma delimited list of cpu numbers, A-B ranges, or `all`. Note, to enable
2299 NUMA options support, fio must be built on a system with libnuma-dev(el)
2302 .. option:: numa_mem_policy=str
2304 Set this job's memory policy and corresponding NUMA nodes. Format of the
2309 ``mode`` is one of the following memory poicies: ``default``, ``prefer``,
2310 ``bind``, ``interleave`` or ``local``. For ``default`` and ``local`` memory
2311 policies, no node needs to be specified. For ``prefer``, only one node is
2312 allowed. For ``bind`` and ``interleave`` the ``nodelist`` may be as
2313 follows: a comma delimited list of numbers, A-B ranges, or `all`.
2315 .. option:: cgroup=str
2317 Add job to this control group. If it doesn't exist, it will be created. The
2318 system must have a mounted cgroup blkio mount point for this to work. If
2319 your system doesn't have it mounted, you can do so with::
2321 # mount -t cgroup -o blkio none /cgroup
2323 .. option:: cgroup_weight=int
2325 Set the weight of the cgroup to this value. See the documentation that comes
2326 with the kernel, allowed values are in the range of 100..1000.
2328 .. option:: cgroup_nodelete=bool
2330 Normally fio will delete the cgroups it has created after the job
2331 completion. To override this behavior and to leave cgroups around after the
2332 job completion, set ``cgroup_nodelete=1``. This can be useful if one wants
2333 to inspect various cgroup files after job completion. Default: false.
2335 .. option:: flow_id=int
2337 The ID of the flow. If not specified, it defaults to being a global
2338 flow. See :option:`flow`.
2340 .. option:: flow=int
2342 Weight in token-based flow control. If this value is used, then there is a
2343 'flow counter' which is used to regulate the proportion of activity between
2344 two or more jobs. Fio attempts to keep this flow counter near zero. The
2345 ``flow`` parameter stands for how much should be added or subtracted to the
2346 flow counter on each iteration of the main I/O loop. That is, if one job has
2347 ``flow=8`` and another job has ``flow=-1``, then there will be a roughly 1:8
2348 ratio in how much one runs vs the other.
2350 .. option:: flow_watermark=int
2352 The maximum value that the absolute value of the flow counter is allowed to
2353 reach before the job must wait for a lower value of the counter.
2355 .. option:: flow_sleep=int
2357 The period of time, in microseconds, to wait after the flow watermark has
2358 been exceeded before retrying operations.
2360 .. option:: stonewall, wait_for_previous
2362 Wait for preceding jobs in the job file to exit, before starting this
2363 one. Can be used to insert serialization points in the job file. A stone
2364 wall also implies starting a new reporting group, see
2365 :option:`group_reporting`.
2369 By default, fio will continue running all other jobs when one job finishes
2370 but sometimes this is not the desired action. Setting ``exitall`` will
2371 instead make fio terminate all other jobs when one job finishes.
2373 .. option:: exec_prerun=str
2375 Before running this job, issue the command specified through
2376 :manpage:`system(3)`. Output is redirected in a file called
2377 :file:`jobname.prerun.txt`.
2379 .. option:: exec_postrun=str
2381 After the job completes, issue the command specified though
2382 :manpage:`system(3)`. Output is redirected in a file called
2383 :file:`jobname.postrun.txt`.
2387 Instead of running as the invoking user, set the user ID to this value
2388 before the thread/process does any work.
2392 Set group ID, see :option:`uid`.
2398 .. option:: verify_only
2400 Do not perform specified workload, only verify data still matches previous
2401 invocation of this workload. This option allows one to check data multiple
2402 times at a later date without overwriting it. This option makes sense only
2403 for workloads that write data, and does not support workloads with the
2404 :option:`time_based` option set.
2406 .. option:: do_verify=bool
2408 Run the verify phase after a write phase. Only valid if :option:`verify` is
2411 .. option:: verify=str
2413 If writing to a file, fio can verify the file contents after each iteration
2414 of the job. Each verification method also implies verification of special
2415 header, which is written to the beginning of each block. This header also
2416 includes meta information, like offset of the block, block number, timestamp
2417 when block was written, etc. :option:`verify` can be combined with
2418 :option:`verify_pattern` option. The allowed values are:
2421 Use an md5 sum of the data area and store it in the header of
2425 Use an experimental crc64 sum of the data area and store it in the
2426 header of each block.
2429 Use a crc32c sum of the data area and store it in the header of
2430 each block. This will automatically use hardware acceleration
2431 (e.g. SSE4.2 on an x86 or CRC crypto extensions on ARM64) but will
2432 fall back to software crc32c if none is found. Generally the
2433 fatest checksum fio supports when hardware accelerated.
2439 Use a crc32 sum of the data area and store it in the header of each
2443 Use a crc16 sum of the data area and store it in the header of each
2447 Use a crc7 sum of the data area and store it in the header of each
2451 Use xxhash as the checksum function. Generally the fastest software
2452 checksum that fio supports.
2455 Use sha512 as the checksum function.
2458 Use sha256 as the checksum function.
2461 Use optimized sha1 as the checksum function.
2464 Use optimized sha3-224 as the checksum function.
2467 Use optimized sha3-256 as the checksum function.
2470 Use optimized sha3-384 as the checksum function.
2473 Use optimized sha3-512 as the checksum function.
2476 This option is deprecated, since now meta information is included in
2477 generic verification header and meta verification happens by
2478 default. For detailed information see the description of the
2479 :option:`verify` setting. This option is kept because of
2480 compatibility's sake with old configurations. Do not use it.
2483 Verify a strict pattern. Normally fio includes a header with some
2484 basic information and checksumming, but if this option is set, only
2485 the specific pattern set with :option:`verify_pattern` is verified.
2488 Only pretend to verify. Useful for testing internals with
2489 :option:`ioengine`\=null, not for much else.
2491 This option can be used for repeated burn-in tests of a system to make sure
2492 that the written data is also correctly read back. If the data direction
2493 given is a read or random read, fio will assume that it should verify a
2494 previously written file. If the data direction includes any form of write,
2495 the verify will be of the newly written data.
2497 .. option:: verifysort=bool
2499 If true, fio will sort written verify blocks when it deems it faster to read
2500 them back in a sorted manner. This is often the case when overwriting an
2501 existing file, since the blocks are already laid out in the file system. You
2502 can ignore this option unless doing huge amounts of really fast I/O where
2503 the red-black tree sorting CPU time becomes significant. Default: true.
2505 .. option:: verifysort_nr=int
2507 Pre-load and sort verify blocks for a read workload.
2509 .. option:: verify_offset=int
2511 Swap the verification header with data somewhere else in the block before
2512 writing. It is swapped back before verifying.
2514 .. option:: verify_interval=int
2516 Write the verification header at a finer granularity than the
2517 :option:`blocksize`. It will be written for chunks the size of
2518 ``verify_interval``. :option:`blocksize` should divide this evenly.
2520 .. option:: verify_pattern=str
2522 If set, fio will fill the I/O buffers with this pattern. Fio defaults to
2523 filling with totally random bytes, but sometimes it's interesting to fill
2524 with a known pattern for I/O verification purposes. Depending on the width
2525 of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time (it can
2526 be either a decimal or a hex number). The ``verify_pattern`` if larger than
2527 a 32-bit quantity has to be a hex number that starts with either "0x" or
2528 "0X". Use with :option:`verify`. Also, ``verify_pattern`` supports %o
2529 format, which means that for each block offset will be written and then
2530 verified back, e.g.::
2534 Or use combination of everything::
2536 verify_pattern=0xff%o"abcd"-12
2538 .. option:: verify_fatal=bool
2540 Normally fio will keep checking the entire contents before quitting on a
2541 block verification failure. If this option is set, fio will exit the job on
2542 the first observed failure. Default: false.
2544 .. option:: verify_dump=bool
2546 If set, dump the contents of both the original data block and the data block
2547 we read off disk to files. This allows later analysis to inspect just what
2548 kind of data corruption occurred. Off by default.
2550 .. option:: verify_async=int
2552 Fio will normally verify I/O inline from the submitting thread. This option
2553 takes an integer describing how many async offload threads to create for I/O
2554 verification instead, causing fio to offload the duty of verifying I/O
2555 contents to one or more separate threads. If using this offload option, even
2556 sync I/O engines can benefit from using an :option:`iodepth` setting higher
2557 than 1, as it allows them to have I/O in flight while verifies are running.
2558 Defaults to 0 async threads, i.e. verification is not asynchronous.
2560 .. option:: verify_async_cpus=str
2562 Tell fio to set the given CPU affinity on the async I/O verification
2563 threads. See :option:`cpus_allowed` for the format used.
2565 .. option:: verify_backlog=int
2567 Fio will normally verify the written contents of a job that utilizes verify
2568 once that job has completed. In other words, everything is written then
2569 everything is read back and verified. You may want to verify continually
2570 instead for a variety of reasons. Fio stores the meta data associated with
2571 an I/O block in memory, so for large verify workloads, quite a bit of memory
2572 would be used up holding this meta data. If this option is enabled, fio will
2573 write only N blocks before verifying these blocks.
2575 .. option:: verify_backlog_batch=int
2577 Control how many blocks fio will verify if :option:`verify_backlog` is
2578 set. If not set, will default to the value of :option:`verify_backlog`
2579 (meaning the entire queue is read back and verified). If
2580 ``verify_backlog_batch`` is less than :option:`verify_backlog` then not all
2581 blocks will be verified, if ``verify_backlog_batch`` is larger than
2582 :option:`verify_backlog`, some blocks will be verified more than once.
2584 .. option:: verify_state_save=bool
2586 When a job exits during the write phase of a verify workload, save its
2587 current state. This allows fio to replay up until that point, if the verify
2588 state is loaded for the verify read phase. The format of the filename is,
2591 <type>-<jobname>-<jobindex>-verify.state.
2593 <type> is "local" for a local run, "sock" for a client/server socket
2594 connection, and "ip" (192.168.0.1, for instance) for a networked
2595 client/server connection. Defaults to true.
2597 .. option:: verify_state_load=bool
2599 If a verify termination trigger was used, fio stores the current write state
2600 of each thread. This can be used at verification time so that fio knows how
2601 far it should verify. Without this information, fio will run a full
2602 verification pass, according to the settings in the job file used. Default
2605 .. option:: trim_percentage=int
2607 Number of verify blocks to discard/trim.
2609 .. option:: trim_verify_zero=bool
2611 Verify that trim/discarded blocks are returned as zeros.
2613 .. option:: trim_backlog=int
2615 Trim after this number of blocks are written.
2617 .. option:: trim_backlog_batch=int
2619 Trim this number of I/O blocks.
2621 .. option:: experimental_verify=bool
2623 Enable experimental verification.
2628 .. option:: steadystate=str:float, ss=str:float
2630 Define the criterion and limit for assessing steady state performance. The
2631 first parameter designates the criterion whereas the second parameter sets
2632 the threshold. When the criterion falls below the threshold for the
2633 specified duration, the job will stop. For example, `iops_slope:0.1%` will
2634 direct fio to terminate the job when the least squares regression slope
2635 falls below 0.1% of the mean IOPS. If :option:`group_reporting` is enabled
2636 this will apply to all jobs in the group. Below is the list of available
2637 steady state assessment criteria. All assessments are carried out using only
2638 data from the rolling collection window. Threshold limits can be expressed
2639 as a fixed value or as a percentage of the mean in the collection window.
2642 Collect IOPS data. Stop the job if all individual IOPS measurements
2643 are within the specified limit of the mean IOPS (e.g., ``iops:2``
2644 means that all individual IOPS values must be within 2 of the mean,
2645 whereas ``iops:0.2%`` means that all individual IOPS values must be
2646 within 0.2% of the mean IOPS to terminate the job).
2649 Collect IOPS data and calculate the least squares regression
2650 slope. Stop the job if the slope falls below the specified limit.
2653 Collect bandwidth data. Stop the job if all individual bandwidth
2654 measurements are within the specified limit of the mean bandwidth.
2657 Collect bandwidth data and calculate the least squares regression
2658 slope. Stop the job if the slope falls below the specified limit.
2660 .. option:: steadystate_duration=time, ss_dur=time
2662 A rolling window of this duration will be used to judge whether steady state
2663 has been reached. Data will be collected once per second. The default is 0
2664 which disables steady state detection. When the unit is omitted, the
2665 value is interpreted in seconds.
2667 .. option:: steadystate_ramp_time=time, ss_ramp=time
2669 Allow the job to run for the specified duration before beginning data
2670 collection for checking the steady state job termination criterion. The
2671 default is 0. When the unit is omitted, the value is interpreted in seconds.
2674 Measurements and reporting
2675 ~~~~~~~~~~~~~~~~~~~~~~~~~~
2677 .. option:: per_job_logs=bool
2679 If set, this generates bw/clat/iops log with per file private filenames. If
2680 not set, jobs with identical names will share the log filename. Default:
2683 .. option:: group_reporting
2685 It may sometimes be interesting to display statistics for groups of jobs as
2686 a whole instead of for each individual job. This is especially true if
2687 :option:`numjobs` is used; looking at individual thread/process output
2688 quickly becomes unwieldy. To see the final report per-group instead of
2689 per-job, use :option:`group_reporting`. Jobs in a file will be part of the
2690 same reporting group, unless if separated by a :option:`stonewall`, or by
2691 using :option:`new_group`.
2693 .. option:: new_group
2695 Start a new reporting group. See: :option:`group_reporting`. If not given,
2696 all jobs in a file will be part of the same reporting group, unless
2697 separated by a :option:`stonewall`.
2699 .. option:: stats=bool
2701 By default, fio collects and shows final output results for all jobs
2702 that run. If this option is set to 0, then fio will ignore it in
2703 the final stat output.
2705 .. option:: write_bw_log=str
2707 If given, write a bandwidth log for this job. Can be used to store data of
2708 the bandwidth of the jobs in their lifetime. The included
2709 :command:`fio_generate_plots` script uses :command:`gnuplot` to turn these
2710 text files into nice graphs. See :option:`write_lat_log` for behavior of
2711 given filename. For this option, the postfix is :file:`_bw.x.log`, where `x`
2712 is the index of the job (`1..N`, where `N` is the number of jobs). If
2713 :option:`per_job_logs` is false, then the filename will not include the job
2714 index. See `Log File Formats`_.
2716 .. option:: write_lat_log=str
2718 Same as :option:`write_bw_log`, except that this option stores I/O
2719 submission, completion, and total latencies instead. If no filename is given
2720 with this option, the default filename of :file:`jobname_type.log` is
2721 used. Even if the filename is given, fio will still append the type of
2722 log. So if one specifies::
2726 The actual log names will be :file:`foo_slat.x.log`, :file:`foo_clat.x.log`,
2727 and :file:`foo_lat.x.log`, where `x` is the index of the job (`1..N`, where `N`
2728 is the number of jobs). This helps :command:`fio_generate_plots` find the
2729 logs automatically. If :option:`per_job_logs` is false, then the filename
2730 will not include the job index. See `Log File Formats`_.
2732 .. option:: write_hist_log=str
2734 Same as :option:`write_lat_log`, but writes I/O completion latency
2735 histograms. If no filename is given with this option, the default filename
2736 of :file:`jobname_clat_hist.x.log` is used, where `x` is the index of the
2737 job (`1..N`, where `N` is the number of jobs). Even if the filename is given,
2738 fio will still append the type of log. If :option:`per_job_logs` is false,
2739 then the filename will not include the job index. See `Log File Formats`_.
2741 .. option:: write_iops_log=str
2743 Same as :option:`write_bw_log`, but writes IOPS. If no filename is given
2744 with this option, the default filename of :file:`jobname_type.x.log` is
2745 used, where `x` is the index of the job (`1..N`, where `N` is the number of
2746 jobs). Even if the filename is given, fio will still append the type of
2747 log. If :option:`per_job_logs` is false, then the filename will not include
2748 the job index. See `Log File Formats`_.
2750 .. option:: log_avg_msec=int
2752 By default, fio will log an entry in the iops, latency, or bw log for every
2753 I/O that completes. When writing to the disk log, that can quickly grow to a
2754 very large size. Setting this option makes fio average the each log entry
2755 over the specified period of time, reducing the resolution of the log. See
2756 :option:`log_max_value` as well. Defaults to 0, logging all entries.
2757 Also see `Log File Formats`_.
2759 .. option:: log_hist_msec=int
2761 Same as :option:`log_avg_msec`, but logs entries for completion latency
2762 histograms. Computing latency percentiles from averages of intervals using
2763 :option:`log_avg_msec` is inaccurate. Setting this option makes fio log
2764 histogram entries over the specified period of time, reducing log sizes for
2765 high IOPS devices while retaining percentile accuracy. See
2766 :option:`log_hist_coarseness` as well. Defaults to 0, meaning histogram
2767 logging is disabled.
2769 .. option:: log_hist_coarseness=int
2771 Integer ranging from 0 to 6, defining the coarseness of the resolution of
2772 the histogram logs enabled with :option:`log_hist_msec`. For each increment
2773 in coarseness, fio outputs half as many bins. Defaults to 0, for which
2774 histogram logs contain 1216 latency bins. See `Log File Formats`_.
2776 .. option:: log_max_value=bool
2778 If :option:`log_avg_msec` is set, fio logs the average over that window. If
2779 you instead want to log the maximum value, set this option to 1. Defaults to
2780 0, meaning that averaged values are logged.
2782 .. option:: log_offset=bool
2784 If this is set, the iolog options will include the byte offset for the I/O
2785 entry as well as the other data values. Defaults to 0 meaning that
2786 offsets are not present in logs. Also see `Log File Formats`_.
2788 .. option:: log_compression=int
2790 If this is set, fio will compress the I/O logs as it goes, to keep the
2791 memory footprint lower. When a log reaches the specified size, that chunk is
2792 removed and compressed in the background. Given that I/O logs are fairly
2793 highly compressible, this yields a nice memory savings for longer runs. The
2794 downside is that the compression will consume some background CPU cycles, so
2795 it may impact the run. This, however, is also true if the logging ends up
2796 consuming most of the system memory. So pick your poison. The I/O logs are
2797 saved normally at the end of a run, by decompressing the chunks and storing
2798 them in the specified log file. This feature depends on the availability of
2801 .. option:: log_compression_cpus=str
2803 Define the set of CPUs that are allowed to handle online log compression for
2804 the I/O jobs. This can provide better isolation between performance
2805 sensitive jobs, and background compression work.
2807 .. option:: log_store_compressed=bool
2809 If set, fio will store the log files in a compressed format. They can be
2810 decompressed with fio, using the :option:`--inflate-log` command line
2811 parameter. The files will be stored with a :file:`.fz` suffix.
2813 .. option:: log_unix_epoch=bool
2815 If set, fio will log Unix timestamps to the log files produced by enabling
2816 write_type_log for each log type, instead of the default zero-based
2819 .. option:: block_error_percentiles=bool
2821 If set, record errors in trim block-sized units from writes and trims and
2822 output a histogram of how many trims it took to get to errors, and what kind
2823 of error was encountered.
2825 .. option:: bwavgtime=int
2827 Average the calculated bandwidth over the given time. Value is specified in
2828 milliseconds. If the job also does bandwidth logging through
2829 :option:`write_bw_log`, then the minimum of this option and
2830 :option:`log_avg_msec` will be used. Default: 500ms.
2832 .. option:: iopsavgtime=int
2834 Average the calculated IOPS over the given time. Value is specified in
2835 milliseconds. If the job also does IOPS logging through
2836 :option:`write_iops_log`, then the minimum of this option and
2837 :option:`log_avg_msec` will be used. Default: 500ms.
2839 .. option:: disk_util=bool
2841 Generate disk utilization statistics, if the platform supports it.
2844 .. option:: disable_lat=bool
2846 Disable measurements of total latency numbers. Useful only for cutting back
2847 the number of calls to :manpage:`gettimeofday(2)`, as that does impact
2848 performance at really high IOPS rates. Note that to really get rid of a
2849 large amount of these calls, this option must be used with
2850 :option:`disable_slat` and :option:`disable_bw_measurement` as well.
2852 .. option:: disable_clat=bool
2854 Disable measurements of completion latency numbers. See
2855 :option:`disable_lat`.
2857 .. option:: disable_slat=bool
2859 Disable measurements of submission latency numbers. See
2860 :option:`disable_lat`.
2862 .. option:: disable_bw_measurement=bool, disable_bw=bool
2864 Disable measurements of throughput/bandwidth numbers. See
2865 :option:`disable_lat`.
2867 .. option:: clat_percentiles=bool
2869 Enable the reporting of percentiles of completion latencies. This
2870 option is mutually exclusive with :option:`lat_percentiles`.
2872 .. option:: lat_percentiles=bool
2874 Enable the reporting of percentiles of IO latencies. This is similar
2875 to :option:`clat_percentiles`, except that this includes the
2876 submission latency. This option is mutually exclusive with
2877 :option:`clat_percentiles`.
2879 .. option:: percentile_list=float_list
2881 Overwrite the default list of percentiles for completion latencies and the
2882 block error histogram. Each number is a floating number in the range
2883 (0,100], and the maximum length of the list is 20. Use ``:`` to separate the
2884 numbers, and list the numbers in ascending order. For example,
2885 ``--percentile_list=99.5:99.9`` will cause fio to report the values of
2886 completion latency below which 99.5% and 99.9% of the observed latencies
2893 .. option:: exitall_on_error
2895 When one job finishes in error, terminate the rest. The default is to wait
2896 for each job to finish.
2898 .. option:: continue_on_error=str
2900 Normally fio will exit the job on the first observed failure. If this option
2901 is set, fio will continue the job when there is a 'non-fatal error' (EIO or
2902 EILSEQ) until the runtime is exceeded or the I/O size specified is
2903 completed. If this option is used, there are two more stats that are
2904 appended, the total error count and the first error. The error field given
2905 in the stats is the first error that was hit during the run.
2907 The allowed values are:
2910 Exit on any I/O or verify errors.
2913 Continue on read errors, exit on all others.
2916 Continue on write errors, exit on all others.
2919 Continue on any I/O error, exit on all others.
2922 Continue on verify errors, exit on all others.
2925 Continue on all errors.
2928 Backward-compatible alias for 'none'.
2931 Backward-compatible alias for 'all'.
2933 .. option:: ignore_error=str
2935 Sometimes you want to ignore some errors during test in that case you can
2936 specify error list for each error type, instead of only being able to
2937 ignore the default 'non-fatal error' using :option:`continue_on_error`.
2938 ``ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST`` errors for
2939 given error type is separated with ':'. Error may be symbol ('ENOSPC',
2940 'ENOMEM') or integer. Example::
2942 ignore_error=EAGAIN,ENOSPC:122
2944 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from
2945 WRITE. This option works by overriding :option:`continue_on_error` with
2946 the list of errors for each error type if any.
2948 .. option:: error_dump=bool
2950 If set dump every error even if it is non fatal, true by default. If
2951 disabled only fatal error will be dumped.
2953 Running predefined workloads
2954 ----------------------------
2956 Fio includes predefined profiles that mimic the I/O workloads generated by
2959 .. option:: profile=str
2961 The predefined workload to run. Current profiles are:
2964 Threaded I/O bench (tiotest/tiobench) like workload.
2967 Aerospike Certification Tool (ACT) like workload.
2969 To view a profile's additional options use :option:`--cmdhelp` after specifying
2970 the profile. For example::
2972 $ fio --profile=act --cmdhelp
2977 .. option:: device-names=str
2982 .. option:: load=int
2985 ACT load multiplier. Default: 1.
2987 .. option:: test-duration=time
2990 How long the entire test takes to run. When the unit is omitted, the value
2991 is given in seconds. Default: 24h.
2993 .. option:: threads-per-queue=int
2996 Number of read I/O threads per device. Default: 8.
2998 .. option:: read-req-num-512-blocks=int
3001 Number of 512B blocks to read at the time. Default: 3.
3003 .. option:: large-block-op-kbytes=int
3006 Size of large block ops in KiB (writes). Default: 131072.
3011 Set to run ACT prep phase.
3013 Tiobench profile options
3014 ~~~~~~~~~~~~~~~~~~~~~~~~
3016 .. option:: size=str
3021 .. option:: block=int
3024 Block size in bytes. Default: 4096.
3026 .. option:: numruns=int
3036 .. option:: threads=int
3041 Interpreting the output
3042 -----------------------
3045 Example output was based on the following:
3046 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --time_based \
3047 --rate=1256k --bs=14K --name=quick --runtime=1s --name=mixed \
3048 --runtime=2m --rw=rw
3050 Fio spits out a lot of output. While running, fio will display the status of the
3051 jobs created. An example of that would be::
3053 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]
3055 The characters inside the first set of square brackets denote the current status of
3056 each thread. The first character is the first job defined in the job file, and so
3057 forth. The possible values (in typical life cycle order) are:
3059 +------+-----+-----------------------------------------------------------+
3061 +======+=====+===========================================================+
3062 | P | | Thread setup, but not started. |
3063 +------+-----+-----------------------------------------------------------+
3064 | C | | Thread created. |
3065 +------+-----+-----------------------------------------------------------+
3066 | I | | Thread initialized, waiting or generating necessary data. |
3067 +------+-----+-----------------------------------------------------------+
3068 | | p | Thread running pre-reading file(s). |
3069 +------+-----+-----------------------------------------------------------+
3070 | | / | Thread is in ramp period. |
3071 +------+-----+-----------------------------------------------------------+
3072 | | R | Running, doing sequential reads. |
3073 +------+-----+-----------------------------------------------------------+
3074 | | r | Running, doing random reads. |
3075 +------+-----+-----------------------------------------------------------+
3076 | | W | Running, doing sequential writes. |
3077 +------+-----+-----------------------------------------------------------+
3078 | | w | Running, doing random writes. |
3079 +------+-----+-----------------------------------------------------------+
3080 | | M | Running, doing mixed sequential reads/writes. |
3081 +------+-----+-----------------------------------------------------------+
3082 | | m | Running, doing mixed random reads/writes. |
3083 +------+-----+-----------------------------------------------------------+
3084 | | D | Running, doing sequential trims. |
3085 +------+-----+-----------------------------------------------------------+
3086 | | d | Running, doing random trims. |
3087 +------+-----+-----------------------------------------------------------+
3088 | | F | Running, currently waiting for :manpage:`fsync(2)`. |
3089 +------+-----+-----------------------------------------------------------+
3090 | | V | Running, doing verification of written data. |
3091 +------+-----+-----------------------------------------------------------+
3092 | f | | Thread finishing. |
3093 +------+-----+-----------------------------------------------------------+
3094 | E | | Thread exited, not reaped by main thread yet. |
3095 +------+-----+-----------------------------------------------------------+
3096 | _ | | Thread reaped. |
3097 +------+-----+-----------------------------------------------------------+
3098 | X | | Thread reaped, exited with an error. |
3099 +------+-----+-----------------------------------------------------------+
3100 | K | | Thread reaped, exited due to signal. |
3101 +------+-----+-----------------------------------------------------------+
3104 Example output was based on the following:
3105 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --runtime=58m \
3106 --time_based --rate=2512k --bs=256K --numjobs=10 \
3107 --name=readers --rw=read --name=writers --rw=write
3109 Fio will condense the thread string as not to take up more space on the command
3110 line than needed. For instance, if you have 10 readers and 10 writers running,
3111 the output would look like this::
3113 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]
3115 Note that the status string is displayed in order, so it's possible to tell which of
3116 the jobs are currently doing what. In the example above this means that jobs 1--10
3117 are readers and 11--20 are writers.
3119 The other values are fairly self explanatory -- number of threads currently
3120 running and doing I/O, the number of currently open files (f=), the estimated
3121 completion percentage, the rate of I/O since last check (read speed listed first,
3122 then write speed and optionally trim speed) in terms of bandwidth and IOPS,
3123 and time to completion for the current running group. It's impossible to estimate
3124 runtime of the following groups (if any).
3127 Example output was based on the following:
3128 TZ=UTC fio --iodepth=16 --ioengine=posixaio --filename=/tmp/fiofile \
3129 --direct=1 --size=100M --time_based --runtime=50s --rate_iops=89 \
3130 --bs=7K --name=Client1 --rw=write
3132 When fio is done (or interrupted by :kbd:`Ctrl-C`), it will show the data for
3133 each thread, group of threads, and disks in that order. For each overall thread (or
3134 group) the output looks like::
3136 Client1: (groupid=0, jobs=1): err= 0: pid=16109: Sat Jun 24 12:07:54 2017
3137 write: IOPS=88, BW=623KiB/s (638kB/s)(30.4MiB/50032msec)
3138 slat (nsec): min=500, max=145500, avg=8318.00, stdev=4781.50
3139 clat (usec): min=170, max=78367, avg=4019.02, stdev=8293.31
3140 lat (usec): min=174, max=78375, avg=4027.34, stdev=8291.79
3141 clat percentiles (usec):
3142 | 1.00th=[ 302], 5.00th=[ 326], 10.00th=[ 343], 20.00th=[ 363],
3143 | 30.00th=[ 392], 40.00th=[ 404], 50.00th=[ 416], 60.00th=[ 445],
3144 | 70.00th=[ 816], 80.00th=[ 6718], 90.00th=[12911], 95.00th=[21627],
3145 | 99.00th=[43779], 99.50th=[51643], 99.90th=[68682], 99.95th=[72877],
3147 bw ( KiB/s): min= 532, max= 686, per=0.10%, avg=622.87, stdev=24.82, samples= 100
3148 iops : min= 76, max= 98, avg=88.98, stdev= 3.54, samples= 100
3149 lat (usec) : 250=0.04%, 500=64.11%, 750=4.81%, 1000=2.79%
3150 lat (msec) : 2=4.16%, 4=1.84%, 10=4.90%, 20=11.33%, 50=5.37%
3151 lat (msec) : 100=0.65%
3152 cpu : usr=0.27%, sys=0.18%, ctx=12072, majf=0, minf=21
3153 IO depths : 1=85.0%, 2=13.1%, 4=1.8%, 8=0.1%, 16=0.0%, 32=0.0%, >=64=0.0%
3154 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
3155 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
3156 issued rwt: total=0,4450,0, short=0,0,0, dropped=0,0,0
3157 latency : target=0, window=0, percentile=100.00%, depth=8
3159 The job name (or first job's name when using :option:`group_reporting`) is printed,
3160 along with the group id, count of jobs being aggregated, last error id seen (which
3161 is 0 when there are no errors), pid/tid of that thread and the time the job/group
3162 completed. Below are the I/O statistics for each data direction performed (showing
3163 writes in the example above). In the order listed, they denote:
3166 The string before the colon shows the I/O direction the statistics
3167 are for. **IOPS** is the average I/Os performed per second. **BW**
3168 is the average bandwidth rate shown as: value in power of 2 format
3169 (value in power of 10 format). The last two values show: (**total
3170 I/O performed** in power of 2 format / **runtime** of that thread).
3173 Submission latency (**min** being the minimum, **max** being the
3174 maximum, **avg** being the average, **stdev** being the standard
3175 deviation). This is the time it took to submit the I/O. For
3176 sync I/O this row is not displayed as the slat is really the
3177 completion latency (since queue/complete is one operation there).
3178 This value can be in nanoseconds, microseconds or milliseconds ---
3179 fio will choose the most appropriate base and print that (in the
3180 example above nanoseconds was the best scale). Note: in :option:`--minimal` mode
3181 latencies are always expressed in microseconds.
3184 Completion latency. Same names as slat, this denotes the time from
3185 submission to completion of the I/O pieces. For sync I/O, clat will
3186 usually be equal (or very close) to 0, as the time from submit to
3187 complete is basically just CPU time (I/O has already been done, see slat
3191 Total latency. Same names as slat and clat, this denotes the time from
3192 when fio created the I/O unit to completion of the I/O operation.
3195 Bandwidth statistics based on samples. Same names as the xlat stats,
3196 but also includes the number of samples taken (**samples**) and an
3197 approximate percentage of total aggregate bandwidth this thread
3198 received in its group (**per**). This last value is only really
3199 useful if the threads in this group are on the same disk, since they
3200 are then competing for disk access.
3203 IOPS statistics based on samples. Same names as bw.
3205 **lat (nsec/usec/msec)**
3206 The distribution of I/O completion latencies. This is the time from when
3207 I/O leaves fio and when it gets completed. Unlike the separate
3208 read/write/trim sections above, the data here and in the remaining
3209 sections apply to all I/Os for the reporting group. 250=0.04% means that
3210 0.04% of the I/Os completed in under 250us. 500=64.11% means that 64.11%
3211 of the I/Os required 250 to 499us for completion.
3214 CPU usage. User and system time, along with the number of context
3215 switches this thread went through, usage of system and user time, and
3216 finally the number of major and minor page faults. The CPU utilization
3217 numbers are averages for the jobs in that reporting group, while the
3218 context and fault counters are summed.
3221 The distribution of I/O depths over the job lifetime. The numbers are
3222 divided into powers of 2 and each entry covers depths from that value
3223 up to those that are lower than the next entry -- e.g., 16= covers
3224 depths from 16 to 31. Note that the range covered by a depth
3225 distribution entry can be different to the range covered by the
3226 equivalent submit/complete distribution entry.
3229 How many pieces of I/O were submitting in a single submit call. Each
3230 entry denotes that amount and below, until the previous entry -- e.g.,
3231 16=100% means that we submitted anywhere between 9 to 16 I/Os per submit
3232 call. Note that the range covered by a submit distribution entry can
3233 be different to the range covered by the equivalent depth distribution
3237 Like the above submit number, but for completions instead.
3240 The number of read/write/trim requests issued, and how many of them were
3244 These values are for `--latency-target` and related options. When
3245 these options are engaged, this section describes the I/O depth required
3246 to meet the specified latency target.
3249 Example output was based on the following:
3250 TZ=UTC fio --ioengine=null --iodepth=2 --size=100M --numjobs=2 \
3251 --rate_process=poisson --io_limit=32M --name=read --bs=128k \
3252 --rate=11M --name=write --rw=write --bs=2k --rate=700k
3254 After each client has been listed, the group statistics are printed. They
3255 will look like this::
3257 Run status group 0 (all jobs):
3258 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
3259 WRITE: bw=1231KiB/s (1261kB/s), 616KiB/s-621KiB/s (630kB/s-636kB/s), io=64.0MiB (67.1MB), run=52747-53223msec
3261 For each data direction it prints:
3264 Aggregate bandwidth of threads in this group followed by the
3265 minimum and maximum bandwidth of all the threads in this group.
3266 Values outside of brackets are power-of-2 format and those
3267 within are the equivalent value in a power-of-10 format.
3269 Aggregate I/O performed of all threads in this group. The
3270 format is the same as bw.
3272 The smallest and longest runtimes of the threads in this group.
3274 And finally, the disk statistics are printed. This is Linux specific. They will look like this::
3276 Disk stats (read/write):
3277 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
3279 Each value is printed for both reads and writes, with reads first. The
3283 Number of I/Os performed by all groups.
3285 Number of merges performed by the I/O scheduler.
3287 Number of ticks we kept the disk busy.
3289 Total time spent in the disk queue.
3291 The disk utilization. A value of 100% means we kept the disk
3292 busy constantly, 50% would be a disk idling half of the time.
3294 It is also possible to get fio to dump the current output while it is running,
3295 without terminating the job. To do that, send fio the **USR1** signal. You can
3296 also get regularly timed dumps by using the :option:`--status-interval`
3297 parameter, or by creating a file in :file:`/tmp` named
3298 :file:`fio-dump-status`. If fio sees this file, it will unlink it and dump the
3299 current output status.
3305 For scripted usage where you typically want to generate tables or graphs of the
3306 results, fio can output the results in a semicolon separated format. The format
3307 is one long line of values, such as::
3309 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%
3310 A description of this job goes here.
3312 The job description (if provided) follows on a second line.
3314 To enable terse output, use the :option:`--minimal` or
3315 :option:`--output-format`\=terse command line options. The
3316 first value is the version of the terse output format. If the output has to be
3317 changed for some reason, this number will be incremented by 1 to signify that
3320 Split up, the format is as follows (comments in brackets denote when a
3321 field was introduced or whether it's specific to some terse version):
3325 terse version, fio version [v3], jobname, groupid, error
3329 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3330 Submission latency: min, max, mean, stdev (usec)
3331 Completion latency: min, max, mean, stdev (usec)
3332 Completion latency percentiles: 20 fields (see below)
3333 Total latency: min, max, mean, stdev (usec)
3334 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
3335 IOPS [v5]: min, max, mean, stdev, number of samples
3341 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3342 Submission latency: min, max, mean, stdev (usec)
3343 Completion latency: min, max, mean, stdev (usec)
3344 Completion latency percentiles: 20 fields (see below)
3345 Total latency: min, max, mean, stdev (usec)
3346 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
3347 IOPS [v5]: min, max, mean, stdev, number of samples
3349 TRIM status [all but version 3]:
3351 Fields are similar to READ/WRITE status.
3355 user, system, context switches, major faults, minor faults
3359 <=1, 2, 4, 8, 16, 32, >=64
3361 I/O latencies microseconds::
3363 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
3365 I/O latencies milliseconds::
3367 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
3369 Disk utilization [v3]::
3371 disk name, read ios, write ios, read merges, write merges, read ticks, write ticks,
3372 time spent in queue, disk utilization percentage
3374 Additional Info (dependent on continue_on_error, default off)::
3376 total # errors, first error code
3378 Additional Info (dependent on description being set)::
3382 Completion latency percentiles can be a grouping of up to 20 sets, so for the
3383 terse output fio writes all of them. Each field will look like this::
3387 which is the Xth percentile, and the `usec` latency associated with it.
3389 For `Disk utilization`, all disks used by fio are shown. So for each disk there
3390 will be a disk utilization section.
3392 Below is a single line containing short names for each of the fields in the
3393 minimal output v3, separated by semicolons::
3395 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
3401 The `json` output format is intended to be both human readable and convenient
3402 for automated parsing. For the most part its sections mirror those of the
3403 `normal` output. The `runtime` value is reported in msec and the `bw` value is
3404 reported in 1024 bytes per second units.
3410 The `json+` output format is identical to the `json` output format except that it
3411 adds a full dump of the completion latency bins. Each `bins` object contains a
3412 set of (key, value) pairs where keys are latency durations and values count how
3413 many I/Os had completion latencies of the corresponding duration. For example,
3416 "bins" : { "87552" : 1, "89600" : 1, "94720" : 1, "96768" : 1, "97792" : 1, "99840" : 1, "100864" : 2, "103936" : 6, "104960" : 534, "105984" : 5995, "107008" : 7529, ... }
3418 This data indicates that one I/O required 87,552ns to complete, two I/Os required
3419 100,864ns to complete, and 7529 I/Os required 107,008ns to complete.
3421 Also included with fio is a Python script `fio_jsonplus_clat2csv` that takes
3422 json+ output and generates CSV-formatted latency data suitable for plotting.
3424 The latency durations actually represent the midpoints of latency intervals.
3425 For details refer to :file:`stat.h`.
3431 There are two trace file format that you can encounter. The older (v1) format is
3432 unsupported since version 1.20-rc3 (March 2008). It will still be described
3433 below in case that you get an old trace and want to understand it.
3435 In any case the trace is a simple text file with a single action per line.
3438 Trace file format v1
3439 ~~~~~~~~~~~~~~~~~~~~
3441 Each line represents a single I/O action in the following format::
3445 where `rw=0/1` for read/write, and the `offset` and `length` entries being in bytes.
3447 This format is not supported in fio versions >= 1.20-rc3.
3450 Trace file format v2
3451 ~~~~~~~~~~~~~~~~~~~~
3453 The second version of the trace file format was added in fio version 1.17. It
3454 allows to access more then one file per trace and has a bigger set of possible
3457 The first line of the trace file has to be::
3461 Following this can be lines in two different formats, which are described below.
3463 The file management format::
3467 The `filename` is given as an absolute path. The `action` can be one of these:
3470 Add the given `filename` to the trace.
3472 Open the file with the given `filename`. The `filename` has to have
3473 been added with the **add** action before.
3475 Close the file with the given `filename`. The file has to have been
3479 The file I/O action format::
3481 filename action offset length
3483 The `filename` is given as an absolute path, and has to have been added and
3484 opened before it can be used with this format. The `offset` and `length` are
3485 given in bytes. The `action` can be one of these:
3488 Wait for `offset` microseconds. Everything below 100 is discarded.
3489 The time is relative to the previous `wait` statement.
3491 Read `length` bytes beginning from `offset`.
3493 Write `length` bytes beginning from `offset`.
3495 :manpage:`fsync(2)` the file.
3497 :manpage:`fdatasync(2)` the file.
3499 Trim the given file from the given `offset` for `length` bytes.
3501 CPU idleness profiling
3502 ----------------------
3504 In some cases, we want to understand CPU overhead in a test. For example, we
3505 test patches for the specific goodness of whether they reduce CPU usage.
3506 Fio implements a balloon approach to create a thread per CPU that runs at idle
3507 priority, meaning that it only runs when nobody else needs the cpu.
3508 By measuring the amount of work completed by the thread, idleness of each CPU
3509 can be derived accordingly.
3511 An unit work is defined as touching a full page of unsigned characters. Mean and
3512 standard deviation of time to complete an unit work is reported in "unit work"
3513 section. Options can be chosen to report detailed percpu idleness or overall
3514 system idleness by aggregating percpu stats.
3517 Verification and triggers
3518 -------------------------
3520 Fio is usually run in one of two ways, when data verification is done. The first
3521 is a normal write job of some sort with verify enabled. When the write phase has
3522 completed, fio switches to reads and verifies everything it wrote. The second
3523 model is running just the write phase, and then later on running the same job
3524 (but with reads instead of writes) to repeat the same I/O patterns and verify
3525 the contents. Both of these methods depend on the write phase being completed,
3526 as fio otherwise has no idea how much data was written.
3528 With verification triggers, fio supports dumping the current write state to
3529 local files. Then a subsequent read verify workload can load this state and know
3530 exactly where to stop. This is useful for testing cases where power is cut to a
3531 server in a managed fashion, for instance.
3533 A verification trigger consists of two things:
3535 1) Storing the write state of each job.
3536 2) Executing a trigger command.
3538 The write state is relatively small, on the order of hundreds of bytes to single
3539 kilobytes. It contains information on the number of completions done, the last X
3542 A trigger is invoked either through creation ('touch') of a specified file in
3543 the system, or through a timeout setting. If fio is run with
3544 :option:`--trigger-file`\= :file:`/tmp/trigger-file`, then it will continually
3545 check for the existence of :file:`/tmp/trigger-file`. When it sees this file, it
3546 will fire off the trigger (thus saving state, and executing the trigger
3549 For client/server runs, there's both a local and remote trigger. If fio is
3550 running as a server backend, it will send the job states back to the client for
3551 safe storage, then execute the remote trigger, if specified. If a local trigger
3552 is specified, the server will still send back the write state, but the client
3553 will then execute the trigger.
3555 Verification trigger example
3556 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3558 Let's say we want to run a powercut test on the remote Linux machine 'server'.
3559 Our write workload is in :file:`write-test.fio`. We want to cut power to 'server' at
3560 some point during the run, and we'll run this test from the safety or our local
3561 machine, 'localbox'. On the server, we'll start the fio backend normally::
3563 server# fio --server
3565 and on the client, we'll fire off the workload::
3567 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger-remote="bash -c \"echo b > /proc/sysrq-triger\""
3569 We set :file:`/tmp/my-trigger` as the trigger file, and we tell fio to execute::
3571 echo b > /proc/sysrq-trigger
3573 on the server once it has received the trigger and sent us the write state. This
3574 will work, but it's not **really** cutting power to the server, it's merely
3575 abruptly rebooting it. If we have a remote way of cutting power to the server
3576 through IPMI or similar, we could do that through a local trigger command
3577 instead. Let's assume we have a script that does IPMI reboot of a given hostname,
3578 ipmi-reboot. On localbox, we could then have run fio with a local trigger
3581 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger="ipmi-reboot server"
3583 For this case, fio would wait for the server to send us the write state, then
3584 execute ``ipmi-reboot server`` when that happened.
3586 Loading verify state
3587 ~~~~~~~~~~~~~~~~~~~~
3589 To load stored write state, a read verification job file must contain the
3590 :option:`verify_state_load` option. If that is set, fio will load the previously
3591 stored state. For a local fio run this is done by loading the files directly,
3592 and on a client/server run, the server backend will ask the client to send the
3593 files over and load them from there.
3599 Fio supports a variety of log file formats, for logging latencies, bandwidth,
3600 and IOPS. The logs share a common format, which looks like this:
3602 *time* (`msec`), *value*, *data direction*, *block size* (`bytes`),
3605 *Time* for the log entry is always in milliseconds. The *value* logged depends
3606 on the type of log, it will be one of the following:
3609 Value is latency in nsecs
3615 *Data direction* is one of the following:
3624 The entry's *block size* is always in bytes. The *offset* is the offset, in bytes,
3625 from the start of the file, for that particular I/O. The logging of the offset can be
3626 toggled with :option:`log_offset`.
3628 Fio defaults to logging every individual I/O. When IOPS are logged for individual
3629 I/Os the *value* entry will always be 1. If windowed logging is enabled through
3630 :option:`log_avg_msec`, fio logs the average values over the specified period of time.
3631 If windowed logging is enabled and :option:`log_max_value` is set, then fio logs
3632 maximum values in that window instead of averages. Since *data direction*, *block
3633 size* and *offset* are per-I/O values, if windowed logging is enabled they
3634 aren't applicable and will be 0.
3639 Normally fio is invoked as a stand-alone application on the machine where the
3640 I/O workload should be generated. However, the backend and frontend of fio can
3641 be run separately i.e., the fio server can generate an I/O workload on the "Device
3642 Under Test" while being controlled by a client on another machine.
3644 Start the server on the machine which has access to the storage DUT::
3648 where `args` defines what fio listens to. The arguments are of the form
3649 ``type,hostname`` or ``IP,port``. *type* is either ``ip`` (or ip4) for TCP/IP
3650 v4, ``ip6`` for TCP/IP v6, or ``sock`` for a local unix domain socket.
3651 *hostname* is either a hostname or IP address, and *port* is the port to listen
3652 to (only valid for TCP/IP, not a local socket). Some examples:
3656 Start a fio server, listening on all interfaces on the default port (8765).
3658 2) ``fio --server=ip:hostname,4444``
3660 Start a fio server, listening on IP belonging to hostname and on port 4444.
3662 3) ``fio --server=ip6:::1,4444``
3664 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
3666 4) ``fio --server=,4444``
3668 Start a fio server, listening on all interfaces on port 4444.
3670 5) ``fio --server=1.2.3.4``
3672 Start a fio server, listening on IP 1.2.3.4 on the default port.
3674 6) ``fio --server=sock:/tmp/fio.sock``
3676 Start a fio server, listening on the local socket :file:`/tmp/fio.sock`.
3678 Once a server is running, a "client" can connect to the fio server with::
3680 fio <local-args> --client=<server> <remote-args> <job file(s)>
3682 where `local-args` are arguments for the client where it is running, `server`
3683 is the connect string, and `remote-args` and `job file(s)` are sent to the
3684 server. The `server` string follows the same format as it does on the server
3685 side, to allow IP/hostname/socket and port strings.
3687 Fio can connect to multiple servers this way::
3689 fio --client=<server1> <job file(s)> --client=<server2> <job file(s)>
3691 If the job file is located on the fio server, then you can tell the server to
3692 load a local file as well. This is done by using :option:`--remote-config` ::
3694 fio --client=server --remote-config /path/to/file.fio
3696 Then fio will open this local (to the server) job file instead of being passed
3697 one from the client.
3699 If you have many servers (example: 100 VMs/containers), you can input a pathname
3700 of a file containing host IPs/names as the parameter value for the
3701 :option:`--client` option. For example, here is an example :file:`host.list`
3702 file containing 2 hostnames::
3704 host1.your.dns.domain
3705 host2.your.dns.domain
3707 The fio command would then be::
3709 fio --client=host.list <job file(s)>
3711 In this mode, you cannot input server-specific parameters or job files -- all
3712 servers receive the same job file.
3714 In order to let ``fio --client`` runs use a shared filesystem from multiple
3715 hosts, ``fio --client`` now prepends the IP address of the server to the
3716 filename. For example, if fio is using the directory :file:`/mnt/nfs/fio` and is
3717 writing filename :file:`fileio.tmp`, with a :option:`--client` `hostfile`
3718 containing two hostnames ``h1`` and ``h2`` with IP addresses 192.168.10.120 and
3719 192.168.10.121, then fio will create two files::
3721 /mnt/nfs/fio/192.168.10.120.fileio.tmp
3722 /mnt/nfs/fio/192.168.10.121.fileio.tmp