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 full status dump every `time` period passed. When the unit is
186 omitted, the value is interpreted in seconds.
188 .. option:: --section=name
190 Only run specified section `name` in job file. Multiple sections can be specified.
191 The ``--section`` option allows one to combine related jobs into one file.
192 E.g. one job file could define light, moderate, and heavy sections. Tell
193 fio to run only the "heavy" section by giving ``--section=heavy``
194 command line option. One can also specify the "write" operations in one
195 section and "verify" operation in another section. The ``--section`` option
196 only applies to job sections. The reserved *global* section is always
199 .. option:: --alloc-size=kb
201 Set the internal smalloc pool size to `kb` in KiB. The
202 ``--alloc-size`` switch allows one to use a larger pool size for smalloc.
203 If running large jobs with randommap enabled, fio can run out of memory.
204 Smalloc is an internal allocator for shared structures from a fixed size
205 memory pool and can grow to 16 pools. The pool size defaults to 16MiB.
207 NOTE: While running :file:`.fio_smalloc.*` backing store files are visible
210 .. option:: --warnings-fatal
212 All fio parser warnings are fatal, causing fio to exit with an
215 .. option:: --max-jobs=nr
217 Set the maximum number of threads/processes to support to `nr`.
219 .. option:: --server=args
221 Start a backend server, with `args` specifying what to listen to.
222 See `Client/Server`_ section.
224 .. option:: --daemonize=pidfile
226 Background a fio server, writing the pid to the given `pidfile` file.
228 .. option:: --client=hostname
230 Instead of running the jobs locally, send and run them on the given `hostname`
231 or set of `hostname`s. See `Client/Server`_ section.
233 .. option:: --remote-config=file
235 Tell fio server to load this local `file`.
237 .. option:: --idle-prof=option
239 Report CPU idleness. `option` is one of the following:
242 Run unit work calibration only and exit.
245 Show aggregate system idleness and unit work.
248 As **system** but also show per CPU idleness.
250 .. option:: --inflate-log=log
252 Inflate and output compressed `log`.
254 .. option:: --trigger-file=file
256 Execute trigger command when `file` exists.
258 .. option:: --trigger-timeout=time
260 Execute trigger at this `time`.
262 .. option:: --trigger=command
264 Set this `command` as local trigger.
266 .. option:: --trigger-remote=command
268 Set this `command` as remote trigger.
270 .. option:: --aux-path=path
272 Use this `path` for fio state generated files.
274 Any parameters following the options will be assumed to be job files, unless
275 they match a job file parameter. Multiple job files can be listed and each job
276 file will be regarded as a separate group. Fio will :option:`stonewall`
277 execution between each group.
283 As previously described, fio accepts one or more job files describing what it is
284 supposed to do. The job file format is the classic ini file, where the names
285 enclosed in [] brackets define the job name. You are free to use any ASCII name
286 you want, except *global* which has special meaning. Following the job name is
287 a sequence of zero or more parameters, one per line, that define the behavior of
288 the job. If the first character in a line is a ';' or a '#', the entire line is
289 discarded as a comment.
291 A *global* section sets defaults for the jobs described in that file. A job may
292 override a *global* section parameter, and a job file may even have several
293 *global* sections if so desired. A job is only affected by a *global* section
296 The :option:`--cmdhelp` option also lists all options. If used with a `command`
297 argument, :option:`--cmdhelp` will detail the given `command`.
299 See the `examples/` directory for inspiration on how to write job files. Note
300 the copyright and license requirements currently apply to `examples/` files.
302 So let's look at a really simple job file that defines two processes, each
303 randomly reading from a 128MiB file:
307 ; -- start job file --
318 As you can see, the job file sections themselves are empty as all the described
319 parameters are shared. As no :option:`filename` option is given, fio makes up a
320 `filename` for each of the jobs as it sees fit. On the command line, this job
321 would look as follows::
323 $ fio --name=global --rw=randread --size=128m --name=job1 --name=job2
326 Let's look at an example that has a number of processes writing randomly to
331 ; -- start job file --
342 Here we have no *global* section, as we only have one job defined anyway. We
343 want to use async I/O here, with a depth of 4 for each file. We also increased
344 the buffer size used to 32KiB and define numjobs to 4 to fork 4 identical
345 jobs. The result is 4 processes each randomly writing to their own 64MiB
346 file. Instead of using the above job file, you could have given the parameters
347 on the command line. For this case, you would specify::
349 $ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4
351 When fio is utilized as a basis of any reasonably large test suite, it might be
352 desirable to share a set of standardized settings across multiple job files.
353 Instead of copy/pasting such settings, any section may pull in an external
354 :file:`filename.fio` file with *include filename* directive, as in the following
357 ; -- start job file including.fio --
361 include glob-include.fio
368 include test-include.fio
369 ; -- end job file including.fio --
373 ; -- start job file glob-include.fio --
376 ; -- end job file glob-include.fio --
380 ; -- start job file test-include.fio --
383 ; -- end job file test-include.fio --
385 Settings pulled into a section apply to that section only (except *global*
386 section). Include directives may be nested in that any included file may contain
387 further include directive(s). Include files may not contain [] sections.
390 Environment variables
391 ~~~~~~~~~~~~~~~~~~~~~
393 Fio also supports environment variable expansion in job files. Any sub-string of
394 the form ``${VARNAME}`` as part of an option value (in other words, on the right
395 of the '='), will be expanded to the value of the environment variable called
396 `VARNAME`. If no such environment variable is defined, or `VARNAME` is the
397 empty string, the empty string will be substituted.
399 As an example, let's look at a sample fio invocation and job file::
401 $ SIZE=64m NUMJOBS=4 fio jobfile.fio
405 ; -- start job file --
412 This will expand to the following equivalent job file at runtime:
416 ; -- start job file --
423 Fio ships with a few example job files, you can also look there for inspiration.
428 Additionally, fio has a set of reserved keywords that will be replaced
429 internally with the appropriate value. Those keywords are:
433 The architecture page size of the running system.
437 Megabytes of total memory in the system.
441 Number of online available CPUs.
443 These can be used on the command line or in the job file, and will be
444 automatically substituted with the current system values when the job is
445 run. Simple math is also supported on these keywords, so you can perform actions
450 and get that properly expanded to 8 times the size of memory in the machine.
456 This section describes in details each parameter associated with a job. Some
457 parameters take an option of a given type, such as an integer or a
458 string. Anywhere a numeric value is required, an arithmetic expression may be
459 used, provided it is surrounded by parentheses. Supported operators are:
468 For time values in expressions, units are microseconds by default. This is
469 different than for time values not in expressions (not enclosed in
470 parentheses). The following types are used:
477 String: A sequence of alphanumeric characters.
480 Integer with possible time suffix. Without a unit value is interpreted as
481 seconds unless otherwise specified. Accepts a suffix of 'd' for days, 'h' for
482 hours, 'm' for minutes, 's' for seconds, 'ms' (or 'msec') for milliseconds and
483 'us' (or 'usec') for microseconds. For example, use 10m for 10 minutes.
488 Integer. A whole number value, which may contain an integer prefix
489 and an integer suffix:
491 [*integer prefix*] **number** [*integer suffix*]
493 The optional *integer prefix* specifies the number's base. The default
494 is decimal. *0x* specifies hexadecimal.
496 The optional *integer suffix* specifies the number's units, and includes an
497 optional unit prefix and an optional unit. For quantities of data, the
498 default unit is bytes. For quantities of time, the default unit is seconds
499 unless otherwise specified.
501 With :option:`kb_base`\=1000, fio follows international standards for unit
502 prefixes. To specify power-of-10 decimal values defined in the
503 International System of Units (SI):
505 * *K* -- means kilo (K) or 1000
506 * *M* -- means mega (M) or 1000**2
507 * *G* -- means giga (G) or 1000**3
508 * *T* -- means tera (T) or 1000**4
509 * *P* -- means peta (P) or 1000**5
511 To specify power-of-2 binary values defined in IEC 80000-13:
513 * *Ki* -- means kibi (Ki) or 1024
514 * *Mi* -- means mebi (Mi) or 1024**2
515 * *Gi* -- means gibi (Gi) or 1024**3
516 * *Ti* -- means tebi (Ti) or 1024**4
517 * *Pi* -- means pebi (Pi) or 1024**5
519 With :option:`kb_base`\=1024 (the default), the unit prefixes are opposite
520 from those specified in the SI and IEC 80000-13 standards to provide
521 compatibility with old scripts. For example, 4k means 4096.
523 For quantities of data, an optional unit of 'B' may be included
524 (e.g., 'kB' is the same as 'k').
526 The *integer suffix* is not case sensitive (e.g., m/mi mean mebi/mega,
527 not milli). 'b' and 'B' both mean byte, not bit.
529 Examples with :option:`kb_base`\=1000:
531 * *4 KiB*: 4096, 4096b, 4096B, 4ki, 4kib, 4kiB, 4Ki, 4KiB
532 * *1 MiB*: 1048576, 1mi, 1024ki
533 * *1 MB*: 1000000, 1m, 1000k
534 * *1 TiB*: 1099511627776, 1ti, 1024gi, 1048576mi
535 * *1 TB*: 1000000000, 1t, 1000m, 1000000k
537 Examples with :option:`kb_base`\=1024 (default):
539 * *4 KiB*: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB
540 * *1 MiB*: 1048576, 1m, 1024k
541 * *1 MB*: 1000000, 1mi, 1000ki
542 * *1 TiB*: 1099511627776, 1t, 1024g, 1048576m
543 * *1 TB*: 1000000000, 1ti, 1000mi, 1000000ki
545 To specify times (units are not case sensitive):
549 * *M* -- means minutes
550 * *s* -- or sec means seconds (default)
551 * *ms* -- or *msec* means milliseconds
552 * *us* -- or *usec* means microseconds
554 If the option accepts an upper and lower range, use a colon ':' or
555 minus '-' to separate such values. See :ref:`irange <irange>`.
556 If the lower value specified happens to be larger than the upper value
557 the two values are swapped.
562 Boolean. Usually parsed as an integer, however only defined for
563 true and false (1 and 0).
568 Integer range with suffix. Allows value range to be given, such as
569 1024-4096. A colon may also be used as the separator, e.g. 1k:4k. If the
570 option allows two sets of ranges, they can be specified with a ',' or '/'
571 delimiter: 1k-4k/8k-32k. Also see :ref:`int <int>`.
574 A list of floating point numbers, separated by a ':' character.
576 With the above in mind, here follows the complete list of fio job parameters.
582 .. option:: kb_base=int
584 Select the interpretation of unit prefixes in input parameters.
587 Inputs comply with IEC 80000-13 and the International
588 System of Units (SI). Use:
590 - power-of-2 values with IEC prefixes (e.g., KiB)
591 - power-of-10 values with SI prefixes (e.g., kB)
594 Compatibility mode (default). To avoid breaking old scripts:
596 - power-of-2 values with SI prefixes
597 - power-of-10 values with IEC prefixes
599 See :option:`bs` for more details on input parameters.
601 Outputs always use correct prefixes. Most outputs include both
604 bw=2383.3kB/s (2327.4KiB/s)
606 If only one value is reported, then kb_base selects the one to use:
608 **1000** -- SI prefixes
610 **1024** -- IEC prefixes
612 .. option:: unit_base=int
614 Base unit for reporting. Allowed values are:
617 Use auto-detection (default).
629 ASCII name of the job. This may be used to override the name printed by fio
630 for this job. Otherwise the job name is used. On the command line this
631 parameter has the special purpose of also signaling the start of a new job.
633 .. option:: description=str
635 Text description of the job. Doesn't do anything except dump this text
636 description when this job is run. It's not parsed.
638 .. option:: loops=int
640 Run the specified number of iterations of this job. Used to repeat the same
641 workload a given number of times. Defaults to 1.
643 .. option:: numjobs=int
645 Create the specified number of clones of this job. Each clone of job
646 is spawned as an independent thread or process. May be used to setup a
647 larger number of threads/processes doing the same thing. Each thread is
648 reported separately; to see statistics for all clones as a whole, use
649 :option:`group_reporting` in conjunction with :option:`new_group`.
650 See :option:`--max-jobs`. Default: 1.
653 Time related parameters
654 ~~~~~~~~~~~~~~~~~~~~~~~
656 .. option:: runtime=time
658 Tell fio to terminate processing after the specified period of time. It
659 can be quite hard to determine for how long a specified job will run, so
660 this parameter is handy to cap the total runtime to a given time. When
661 the unit is omitted, the value is intepreted in seconds.
663 .. option:: time_based
665 If set, fio will run for the duration of the :option:`runtime` specified
666 even if the file(s) are completely read or written. It will simply loop over
667 the same workload as many times as the :option:`runtime` allows.
669 .. option:: startdelay=irange(time)
671 Delay the start of job for the specified amount of time. Can be a single
672 value or a range. When given as a range, each thread will choose a value
673 randomly from within the range. Value is in seconds if a unit is omitted.
675 .. option:: ramp_time=time
677 If set, fio will run the specified workload for this amount of time before
678 logging any performance numbers. Useful for letting performance settle
679 before logging results, thus minimizing the runtime required for stable
680 results. Note that the ``ramp_time`` is considered lead in time for a job,
681 thus it will increase the total runtime if a special timeout or
682 :option:`runtime` is specified. When the unit is omitted, the value is
685 .. option:: clocksource=str
687 Use the given clocksource as the base of timing. The supported options are:
690 :manpage:`gettimeofday(2)`
693 :manpage:`clock_gettime(2)`
696 Internal CPU clock source
698 cpu is the preferred clocksource if it is reliable, as it is very fast (and
699 fio is heavy on time calls). Fio will automatically use this clocksource if
700 it's supported and considered reliable on the system it is running on,
701 unless another clocksource is specifically set. For x86/x86-64 CPUs, this
702 means supporting TSC Invariant.
704 .. option:: gtod_reduce=bool
706 Enable all of the :manpage:`gettimeofday(2)` reducing options
707 (:option:`disable_clat`, :option:`disable_slat`, :option:`disable_bw_measurement`) plus
708 reduce precision of the timeout somewhat to really shrink the
709 :manpage:`gettimeofday(2)` call count. With this option enabled, we only do
710 about 0.4% of the :manpage:`gettimeofday(2)` calls we would have done if all
711 time keeping was enabled.
713 .. option:: gtod_cpu=int
715 Sometimes it's cheaper to dedicate a single thread of execution to just
716 getting the current time. Fio (and databases, for instance) are very
717 intensive on :manpage:`gettimeofday(2)` calls. With this option, you can set
718 one CPU aside for doing nothing but logging current time to a shared memory
719 location. Then the other threads/processes that run I/O workloads need only
720 copy that segment, instead of entering the kernel with a
721 :manpage:`gettimeofday(2)` call. The CPU set aside for doing these time
722 calls will be excluded from other uses. Fio will manually clear it from the
723 CPU mask of other jobs.
729 .. option:: directory=str
731 Prefix filenames with this directory. Used to place files in a different
732 location than :file:`./`. You can specify a number of directories by
733 separating the names with a ':' character. These directories will be
734 assigned equally distributed to job clones created by :option:`numjobs` as
735 long as they are using generated filenames. If specific `filename(s)` are
736 set fio will use the first listed directory, and thereby matching the
737 `filename` semantic which generates a file each clone if not specified, but
738 let all clones use the same if set.
740 See the :option:`filename` option for information on how to escape "``:``" and
741 "``\``" characters within the directory path itself.
743 .. option:: filename=str
745 Fio normally makes up a `filename` based on the job name, thread number, and
746 file number (see :option:`filename_format`). If you want to share files
747 between threads in a job or several
748 jobs with fixed file paths, specify a `filename` for each of them to override
749 the default. If the ioengine is file based, you can specify a number of files
750 by separating the names with a ':' colon. So if you wanted a job to open
751 :file:`/dev/sda` and :file:`/dev/sdb` as the two working files, you would use
752 ``filename=/dev/sda:/dev/sdb``. This also means that whenever this option is
753 specified, :option:`nrfiles` is ignored. The size of regular files specified
754 by this option will be :option:`size` divided by number of files unless an
755 explicit size is specified by :option:`filesize`.
757 Each colon and backslash in the wanted path must be escaped with a ``\``
758 character. For instance, if the path is :file:`/dev/dsk/foo@3,0:c` then you
759 would use ``filename=/dev/dsk/foo@3,0\:c`` and if the path is
760 :file:`F:\\filename` then you would use ``filename=F\:\\filename``.
762 On Windows, disk devices are accessed as :file:`\\\\.\\PhysicalDrive0` for
763 the first device, :file:`\\\\.\\PhysicalDrive1` for the second etc.
764 Note: Windows and FreeBSD prevent write access to areas
765 of the disk containing in-use data (e.g. filesystems).
767 The filename "`-`" is a reserved name, meaning *stdin* or *stdout*. Which
768 of the two depends on the read/write direction set.
770 .. option:: filename_format=str
772 If sharing multiple files between jobs, it is usually necessary to have fio
773 generate the exact names that you want. By default, fio will name a file
774 based on the default file format specification of
775 :file:`jobname.jobnumber.filenumber`. With this option, that can be
776 customized. Fio will recognize and replace the following keywords in this
780 The name of the worker thread or process.
782 The incremental number of the worker thread or process.
784 The incremental number of the file for that worker thread or
787 To have dependent jobs share a set of files, this option can be set to have
788 fio generate filenames that are shared between the two. For instance, if
789 :file:`testfiles.$filenum` is specified, file number 4 for any job will be
790 named :file:`testfiles.4`. The default of :file:`$jobname.$jobnum.$filenum`
791 will be used if no other format specifier is given.
793 .. option:: unique_filename=bool
795 To avoid collisions between networked clients, fio defaults to prefixing any
796 generated filenames (with a directory specified) with the source of the
797 client connecting. To disable this behavior, set this option to 0.
799 .. option:: opendir=str
801 Recursively open any files below directory `str`.
803 .. option:: lockfile=str
805 Fio defaults to not locking any files before it does I/O to them. If a file
806 or file descriptor is shared, fio can serialize I/O to that file to make the
807 end result consistent. This is usual for emulating real workloads that share
808 files. The lock modes are:
811 No locking. The default.
813 Only one thread or process may do I/O at a time, excluding all
816 Read-write locking on the file. Many readers may
817 access the file at the same time, but writes get exclusive access.
819 .. option:: nrfiles=int
821 Number of files to use for this job. Defaults to 1. The size of files
822 will be :option:`size` divided by this unless explicit size is specified by
823 :option:`filesize`. Files are created for each thread separately, and each
824 file will have a file number within its name by default, as explained in
825 :option:`filename` section.
828 .. option:: openfiles=int
830 Number of files to keep open at the same time. Defaults to the same as
831 :option:`nrfiles`, can be set smaller to limit the number simultaneous
834 .. option:: file_service_type=str
836 Defines how fio decides which file from a job to service next. The following
840 Choose a file at random.
843 Round robin over opened files. This is the default.
846 Finish one file before moving on to the next. Multiple files can
847 still be open depending on :option:`openfiles`.
850 Use a *Zipf* distribution to decide what file to access.
853 Use a *Pareto* distribution to decide what file to access.
856 Use a *Gaussian* (normal) distribution to decide what file to
862 For *random*, *roundrobin*, and *sequential*, a postfix can be appended to
863 tell fio how many I/Os to issue before switching to a new file. For example,
864 specifying ``file_service_type=random:8`` would cause fio to issue
865 8 I/Os before selecting a new file at random. For the non-uniform
866 distributions, a floating point postfix can be given to influence how the
867 distribution is skewed. See :option:`random_distribution` for a description
868 of how that would work.
870 .. option:: ioscheduler=str
872 Attempt to switch the device hosting the file to the specified I/O scheduler
875 .. option:: create_serialize=bool
877 If true, serialize the file creation for the jobs. This may be handy to
878 avoid interleaving of data files, which may greatly depend on the filesystem
879 used and even the number of processors in the system. Default: true.
881 .. option:: create_fsync=bool
883 :manpage:`fsync(2)` the data file after creation. This is the default.
885 .. option:: create_on_open=bool
887 If true, don't pre-create files but allow the job's open() to create a file
888 when it's time to do I/O. Default: false -- pre-create all necessary files
891 .. option:: create_only=bool
893 If true, fio will only run the setup phase of the job. If files need to be
894 laid out or updated on disk, only that will be done -- the actual job contents
895 are not executed. Default: false.
897 .. option:: allow_file_create=bool
899 If true, fio is permitted to create files as part of its workload. If this
900 option is false, then fio will error out if
901 the files it needs to use don't already exist. Default: true.
903 .. option:: allow_mounted_write=bool
905 If this isn't set, fio will abort jobs that are destructive (e.g. that write)
906 to what appears to be a mounted device or partition. This should help catch
907 creating inadvertently destructive tests, not realizing that the test will
908 destroy data on the mounted file system. Note that some platforms don't allow
909 writing against a mounted device regardless of this option. Default: false.
911 .. option:: pre_read=bool
913 If this is given, files will be pre-read into memory before starting the
914 given I/O operation. This will also clear the :option:`invalidate` flag,
915 since it is pointless to pre-read and then drop the cache. This will only
916 work for I/O engines that are seek-able, since they allow you to read the
917 same data multiple times. Thus it will not work on non-seekable I/O engines
918 (e.g. network, splice). Default: false.
920 .. option:: unlink=bool
922 Unlink the job files when done. Not the default, as repeated runs of that
923 job would then waste time recreating the file set again and again. Default:
926 .. option:: unlink_each_loop=bool
928 Unlink job files after each iteration or loop. Default: false.
930 .. option:: zonesize=int
932 Divide a file into zones of the specified size. See :option:`zoneskip`.
934 .. option:: zonerange=int
936 Give size of an I/O zone. See :option:`zoneskip`.
938 .. option:: zoneskip=int
940 Skip the specified number of bytes when :option:`zonesize` data has been
941 read. The two zone options can be used to only do I/O on zones of a file.
947 .. option:: direct=bool
949 If value is true, use non-buffered I/O. This is usually O_DIRECT. Note that
950 OpenBSD and ZFS on Solaris don't support direct I/O. On Windows the synchronous
951 ioengines don't support direct I/O. Default: false.
953 .. option:: atomic=bool
955 If value is true, attempt to use atomic direct I/O. Atomic writes are
956 guaranteed to be stable once acknowledged by the operating system. Only
957 Linux supports O_ATOMIC right now.
959 .. option:: buffered=bool
961 If value is true, use buffered I/O. This is the opposite of the
962 :option:`direct` option. Defaults to true.
964 .. option:: readwrite=str, rw=str
966 Type of I/O pattern. Accepted values are:
973 Sequential trims (Linux block devices only).
979 Random trims (Linux block devices only).
981 Sequential mixed reads and writes.
983 Random mixed reads and writes.
985 Sequential trim+write sequences. Blocks will be trimmed first,
986 then the same blocks will be written to.
988 Fio defaults to read if the option is not specified. For the mixed I/O
989 types, the default is to split them 50/50. For certain types of I/O the
990 result may still be skewed a bit, since the speed may be different.
992 It is possible to specify the number of I/Os to do before getting a new
993 offset by appending ``:<nr>`` to the end of the string given. For a
994 random read, it would look like ``rw=randread:8`` for passing in an offset
995 modifier with a value of 8. If the suffix is used with a sequential I/O
996 pattern, then the *<nr>* value specified will be **added** to the generated
997 offset for each I/O turning sequential I/O into sequential I/O with holes.
998 For instance, using ``rw=write:4k`` will skip 4k for every write. Also see
999 the :option:`rw_sequencer` option.
1001 .. option:: rw_sequencer=str
1003 If an offset modifier is given by appending a number to the ``rw=<str>``
1004 line, then this option controls how that number modifies the I/O offset
1005 being generated. Accepted values are:
1008 Generate sequential offset.
1010 Generate the same offset.
1012 ``sequential`` is only useful for random I/O, where fio would normally
1013 generate a new random offset for every I/O. If you append e.g. 8 to randread,
1014 you would get a new random offset for every 8 I/Os. The result would be a
1015 seek for only every 8 I/Os, instead of for every I/O. Use ``rw=randread:8``
1016 to specify that. As sequential I/O is already sequential, setting
1017 ``sequential`` for that would not result in any differences. ``identical``
1018 behaves in a similar fashion, except it sends the same offset 8 number of
1019 times before generating a new offset.
1021 .. option:: unified_rw_reporting=bool
1023 Fio normally reports statistics on a per data direction basis, meaning that
1024 reads, writes, and trims are accounted and reported separately. If this
1025 option is set fio sums the results and report them as "mixed" instead.
1027 .. option:: randrepeat=bool
1029 Seed the random number generator used for random I/O patterns in a
1030 predictable way so the pattern is repeatable across runs. Default: true.
1032 .. option:: allrandrepeat=bool
1034 Seed all random number generators in a predictable way so results are
1035 repeatable across runs. Default: false.
1037 .. option:: randseed=int
1039 Seed the random number generators based on this seed value, to be able to
1040 control what sequence of output is being generated. If not set, the random
1041 sequence depends on the :option:`randrepeat` setting.
1043 .. option:: fallocate=str
1045 Whether pre-allocation is performed when laying down files.
1046 Accepted values are:
1049 Do not pre-allocate space.
1052 Use a platform's native pre-allocation call but fall back to
1053 **none** behavior if it fails/is not implemented.
1056 Pre-allocate via :manpage:`posix_fallocate(3)`.
1059 Pre-allocate via :manpage:`fallocate(2)` with
1060 FALLOC_FL_KEEP_SIZE set.
1063 Backward-compatible alias for **none**.
1066 Backward-compatible alias for **posix**.
1068 May not be available on all supported platforms. **keep** is only available
1069 on Linux. If using ZFS on Solaris this cannot be set to **posix**
1070 because ZFS doesn't support pre-allocation. Default: **native** if any
1071 pre-allocation methods are available, **none** if not.
1073 .. option:: fadvise_hint=str
1075 Use :manpage:`posix_fadvise(2)` to advise the kernel on what I/O patterns
1076 are likely to be issued. Accepted values are:
1079 Backwards-compatible hint for "no hint".
1082 Backwards compatible hint for "advise with fio workload type". This
1083 uses **FADV_RANDOM** for a random workload, and **FADV_SEQUENTIAL**
1084 for a sequential workload.
1087 Advise using **FADV_SEQUENTIAL**.
1090 Advise using **FADV_RANDOM**.
1092 .. option:: write_hint=str
1094 Use :manpage:`fcntl(2)` to advise the kernel what life time to expect
1095 from a write. Only supported on Linux, as of version 4.13. Accepted
1099 No particular life time associated with this file.
1102 Data written to this file has a short life time.
1105 Data written to this file has a medium life time.
1108 Data written to this file has a long life time.
1111 Data written to this file has a very long life time.
1113 The values are all relative to each other, and no absolute meaning
1114 should be associated with them.
1116 .. option:: offset=int
1118 Start I/O at the provided offset in the file, given as either a fixed size in
1119 bytes or a percentage. If a percentage is given, the next ``blockalign``-ed
1120 offset will be used. Data before the given offset will not be touched. This
1121 effectively caps the file size at `real_size - offset`. Can be combined with
1122 :option:`size` to constrain the start and end range of the I/O workload.
1123 A percentage can be specified by a number between 1 and 100 followed by '%',
1124 for example, ``offset=20%`` to specify 20%.
1126 .. option:: offset_increment=int
1128 If this is provided, then the real offset becomes `offset + offset_increment
1129 * thread_number`, where the thread number is a counter that starts at 0 and
1130 is incremented for each sub-job (i.e. when :option:`numjobs` option is
1131 specified). This option is useful if there are several jobs which are
1132 intended to operate on a file in parallel disjoint segments, with even
1133 spacing between the starting points.
1135 .. option:: number_ios=int
1137 Fio will normally perform I/Os until it has exhausted the size of the region
1138 set by :option:`size`, or if it exhaust the allocated time (or hits an error
1139 condition). With this setting, the range/size can be set independently of
1140 the number of I/Os to perform. When fio reaches this number, it will exit
1141 normally and report status. Note that this does not extend the amount of I/O
1142 that will be done, it will only stop fio if this condition is met before
1143 other end-of-job criteria.
1145 .. option:: fsync=int
1147 If writing to a file, issue an :manpage:`fsync(2)` (or its equivalent) of
1148 the dirty data for every number of blocks given. For example, if you give 32
1149 as a parameter, fio will sync the file after every 32 writes issued. If fio is
1150 using non-buffered I/O, we may not sync the file. The exception is the sg
1151 I/O engine, which synchronizes the disk cache anyway. Defaults to 0, which
1152 means fio does not periodically issue and wait for a sync to complete. Also
1153 see :option:`end_fsync` and :option:`fsync_on_close`.
1155 .. option:: fdatasync=int
1157 Like :option:`fsync` but uses :manpage:`fdatasync(2)` to only sync data and
1158 not metadata blocks. In Windows, FreeBSD, and DragonFlyBSD there is no
1159 :manpage:`fdatasync(2)` so this falls back to using :manpage:`fsync(2)`.
1160 Defaults to 0, which means fio does not periodically issue and wait for a
1161 data-only sync to complete.
1163 .. option:: write_barrier=int
1165 Make every `N-th` write a barrier write.
1167 .. option:: sync_file_range=str:int
1169 Use :manpage:`sync_file_range(2)` for every `int` number of write
1170 operations. Fio will track range of writes that have happened since the last
1171 :manpage:`sync_file_range(2)` call. `str` can currently be one or more of:
1174 SYNC_FILE_RANGE_WAIT_BEFORE
1176 SYNC_FILE_RANGE_WRITE
1178 SYNC_FILE_RANGE_WAIT_AFTER
1180 So if you do ``sync_file_range=wait_before,write:8``, fio would use
1181 ``SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE`` for every 8
1182 writes. Also see the :manpage:`sync_file_range(2)` man page. This option is
1185 .. option:: overwrite=bool
1187 If true, writes to a file will always overwrite existing data. If the file
1188 doesn't already exist, it will be created before the write phase begins. If
1189 the file exists and is large enough for the specified write phase, nothing
1190 will be done. Default: false.
1192 .. option:: end_fsync=bool
1194 If true, :manpage:`fsync(2)` file contents when a write stage has completed.
1197 .. option:: fsync_on_close=bool
1199 If true, fio will :manpage:`fsync(2)` a dirty file on close. This differs
1200 from :option:`end_fsync` in that it will happen on every file close, not
1201 just at the end of the job. Default: false.
1203 .. option:: rwmixread=int
1205 Percentage of a mixed workload that should be reads. Default: 50.
1207 .. option:: rwmixwrite=int
1209 Percentage of a mixed workload that should be writes. If both
1210 :option:`rwmixread` and :option:`rwmixwrite` is given and the values do not
1211 add up to 100%, the latter of the two will be used to override the
1212 first. This may interfere with a given rate setting, if fio is asked to
1213 limit reads or writes to a certain rate. If that is the case, then the
1214 distribution may be skewed. Default: 50.
1216 .. option:: random_distribution=str:float[,str:float][,str:float]
1218 By default, fio will use a completely uniform random distribution when asked
1219 to perform random I/O. Sometimes it is useful to skew the distribution in
1220 specific ways, ensuring that some parts of the data is more hot than others.
1221 fio includes the following distribution models:
1224 Uniform random distribution
1233 Normal (Gaussian) distribution
1236 Zoned random distribution
1238 When using a **zipf** or **pareto** distribution, an input value is also
1239 needed to define the access pattern. For **zipf**, this is the `Zipf
1240 theta`. For **pareto**, it's the `Pareto power`. Fio includes a test
1241 program, :command:`fio-genzipf`, that can be used visualize what the given input
1242 values will yield in terms of hit rates. If you wanted to use **zipf** with
1243 a `theta` of 1.2, you would use ``random_distribution=zipf:1.2`` as the
1244 option. If a non-uniform model is used, fio will disable use of the random
1245 map. For the **normal** distribution, a normal (Gaussian) deviation is
1246 supplied as a value between 0 and 100.
1248 For a **zoned** distribution, fio supports specifying percentages of I/O
1249 access that should fall within what range of the file or device. For
1250 example, given a criteria of:
1252 * 60% of accesses should be to the first 10%
1253 * 30% of accesses should be to the next 20%
1254 * 8% of accesses should be to the next 30%
1255 * 2% of accesses should be to the next 40%
1257 we can define that through zoning of the random accesses. For the above
1258 example, the user would do::
1260 random_distribution=zoned:60/10:30/20:8/30:2/40
1262 similarly to how :option:`bssplit` works for setting ranges and percentages
1263 of block sizes. Like :option:`bssplit`, it's possible to specify separate
1264 zones for reads, writes, and trims. If just one set is given, it'll apply to
1267 .. option:: percentage_random=int[,int][,int]
1269 For a random workload, set how big a percentage should be random. This
1270 defaults to 100%, in which case the workload is fully random. It can be set
1271 from anywhere from 0 to 100. Setting it to 0 would make the workload fully
1272 sequential. Any setting in between will result in a random mix of sequential
1273 and random I/O, at the given percentages. Comma-separated values may be
1274 specified for reads, writes, and trims as described in :option:`blocksize`.
1276 .. option:: norandommap
1278 Normally fio will cover every block of the file when doing random I/O. If
1279 this option is given, fio will just get a new random offset without looking
1280 at past I/O history. This means that some blocks may not be read or written,
1281 and that some blocks may be read/written more than once. If this option is
1282 used with :option:`verify` and multiple blocksizes (via :option:`bsrange`),
1283 only intact blocks are verified, i.e., partially-overwritten blocks are
1286 .. option:: softrandommap=bool
1288 See :option:`norandommap`. If fio runs with the random block map enabled and
1289 it fails to allocate the map, if this option is set it will continue without
1290 a random block map. As coverage will not be as complete as with random maps,
1291 this option is disabled by default.
1293 .. option:: random_generator=str
1295 Fio supports the following engines for generating I/O offsets for random I/O:
1298 Strong 2^88 cycle random number generator.
1300 Linear feedback shift register generator.
1302 Strong 64-bit 2^258 cycle random number generator.
1304 **tausworthe** is a strong random number generator, but it requires tracking
1305 on the side if we want to ensure that blocks are only read or written
1306 once. **lfsr** guarantees that we never generate the same offset twice, and
1307 it's also less computationally expensive. It's not a true random generator,
1308 however, though for I/O purposes it's typically good enough. **lfsr** only
1309 works with single block sizes, not with workloads that use multiple block
1310 sizes. If used with such a workload, fio may read or write some blocks
1311 multiple times. The default value is **tausworthe**, unless the required
1312 space exceeds 2^32 blocks. If it does, then **tausworthe64** is
1313 selected automatically.
1319 .. option:: blocksize=int[,int][,int], bs=int[,int][,int]
1321 The block size in bytes used for I/O units. Default: 4096. A single value
1322 applies to reads, writes, and trims. Comma-separated values may be
1323 specified for reads, writes, and trims. A value not terminated in a comma
1324 applies to subsequent types.
1329 means 256k for reads, writes and trims.
1332 means 8k for reads, 32k for writes and trims.
1335 means 8k for reads, 32k for writes, and default for trims.
1338 means default for reads, 8k for writes and trims.
1341 means default for reads, 8k for writes, and default for trims.
1343 .. option:: blocksize_range=irange[,irange][,irange], bsrange=irange[,irange][,irange]
1345 A range of block sizes in bytes for I/O units. The issued I/O unit will
1346 always be a multiple of the minimum size, unless
1347 :option:`blocksize_unaligned` is set.
1349 Comma-separated ranges may be specified for reads, writes, and trims as
1350 described in :option:`blocksize`.
1352 Example: ``bsrange=1k-4k,2k-8k``.
1354 .. option:: bssplit=str[,str][,str]
1356 Sometimes you want even finer grained control of the block sizes issued, not
1357 just an even split between them. This option allows you to weight various
1358 block sizes, so that you are able to define a specific amount of block sizes
1359 issued. The format for this option is::
1361 bssplit=blocksize/percentage:blocksize/percentage
1363 for as many block sizes as needed. So if you want to define a workload that
1364 has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would write::
1366 bssplit=4k/10:64k/50:32k/40
1368 Ordering does not matter. If the percentage is left blank, fio will fill in
1369 the remaining values evenly. So a bssplit option like this one::
1371 bssplit=4k/50:1k/:32k/
1373 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always add up
1374 to 100, if bssplit is given a range that adds up to more, it will error out.
1376 Comma-separated values may be specified for reads, writes, and trims as
1377 described in :option:`blocksize`.
1379 If you want a workload that has 50% 2k reads and 50% 4k reads, while having
1380 90% 4k writes and 10% 8k writes, you would specify::
1382 bssplit=2k/50:4k/50,4k/90,8k/10
1384 .. option:: blocksize_unaligned, bs_unaligned
1386 If set, fio will issue I/O units with any size within
1387 :option:`blocksize_range`, not just multiples of the minimum size. This
1388 typically won't work with direct I/O, as that normally requires sector
1391 .. option:: bs_is_seq_rand=bool
1393 If this option is set, fio will use the normal read,write blocksize settings
1394 as sequential,random blocksize settings instead. Any random read or write
1395 will use the WRITE blocksize settings, and any sequential read or write will
1396 use the READ blocksize settings.
1398 .. option:: blockalign=int[,int][,int], ba=int[,int][,int]
1400 Boundary to which fio will align random I/O units. Default:
1401 :option:`blocksize`. Minimum alignment is typically 512b for using direct
1402 I/O, though it usually depends on the hardware block size. This option is
1403 mutually exclusive with using a random map for files, so it will turn off
1404 that option. Comma-separated values may be specified for reads, writes, and
1405 trims as described in :option:`blocksize`.
1411 .. option:: zero_buffers
1413 Initialize buffers with all zeros. Default: fill buffers with random data.
1415 .. option:: refill_buffers
1417 If this option is given, fio will refill the I/O buffers on every
1418 submit. The default is to only fill it at init time and reuse that
1419 data. Only makes sense if zero_buffers isn't specified, naturally. If data
1420 verification is enabled, `refill_buffers` is also automatically enabled.
1422 .. option:: scramble_buffers=bool
1424 If :option:`refill_buffers` is too costly and the target is using data
1425 deduplication, then setting this option will slightly modify the I/O buffer
1426 contents to defeat normal de-dupe attempts. This is not enough to defeat
1427 more clever block compression attempts, but it will stop naive dedupe of
1428 blocks. Default: true.
1430 .. option:: buffer_compress_percentage=int
1432 If this is set, then fio will attempt to provide I/O buffer content (on
1433 WRITEs) that compresses to the specified level. Fio does this by providing a
1434 mix of random data and a fixed pattern. The fixed pattern is either zeros,
1435 or the pattern specified by :option:`buffer_pattern`. If the pattern option
1436 is used, it might skew the compression ratio slightly. Note that this is per
1437 block size unit, for file/disk wide compression level that matches this
1438 setting, you'll also want to set :option:`refill_buffers`.
1440 .. option:: buffer_compress_chunk=int
1442 See :option:`buffer_compress_percentage`. This setting allows fio to manage
1443 how big the ranges of random data and zeroed data is. Without this set, fio
1444 will provide :option:`buffer_compress_percentage` of blocksize random data,
1445 followed by the remaining zeroed. With this set to some chunk size smaller
1446 than the block size, fio can alternate random and zeroed data throughout the
1449 .. option:: buffer_pattern=str
1451 If set, fio will fill the I/O buffers with this pattern or with the contents
1452 of a file. If not set, the contents of I/O buffers are defined by the other
1453 options related to buffer contents. The setting can be any pattern of bytes,
1454 and can be prefixed with 0x for hex values. It may also be a string, where
1455 the string must then be wrapped with ``""``. Or it may also be a filename,
1456 where the filename must be wrapped with ``''`` in which case the file is
1457 opened and read. Note that not all the file contents will be read if that
1458 would cause the buffers to overflow. So, for example::
1460 buffer_pattern='filename'
1464 buffer_pattern="abcd"
1472 buffer_pattern=0xdeadface
1474 Also you can combine everything together in any order::
1476 buffer_pattern=0xdeadface"abcd"-12'filename'
1478 .. option:: dedupe_percentage=int
1480 If set, fio will generate this percentage of identical buffers when
1481 writing. These buffers will be naturally dedupable. The contents of the
1482 buffers depend on what other buffer compression settings have been set. It's
1483 possible to have the individual buffers either fully compressible, or not at
1484 all. This option only controls the distribution of unique buffers.
1486 .. option:: invalidate=bool
1488 Invalidate the buffer/page cache parts of the files to be used prior to
1489 starting I/O if the platform and file type support it. Defaults to true.
1490 This will be ignored if :option:`pre_read` is also specified for the
1493 .. option:: sync=bool
1495 Use synchronous I/O for buffered writes. For the majority of I/O engines,
1496 this means using O_SYNC. Default: false.
1498 .. option:: iomem=str, mem=str
1500 Fio can use various types of memory as the I/O unit buffer. The allowed
1504 Use memory from :manpage:`malloc(3)` as the buffers. Default memory
1508 Use shared memory as the buffers. Allocated through
1509 :manpage:`shmget(2)`.
1512 Same as shm, but use huge pages as backing.
1515 Use :manpage:`mmap(2)` to allocate buffers. May either be anonymous memory, or can
1516 be file backed if a filename is given after the option. The format
1517 is `mem=mmap:/path/to/file`.
1520 Use a memory mapped huge file as the buffer backing. Append filename
1521 after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file`.
1524 Same as mmap, but use a MMAP_SHARED mapping.
1527 Use GPU memory as the buffers for GPUDirect RDMA benchmark.
1528 The :option:`ioengine` must be `rdma`.
1530 The area allocated is a function of the maximum allowed bs size for the job,
1531 multiplied by the I/O depth given. Note that for **shmhuge** and
1532 **mmaphuge** to work, the system must have free huge pages allocated. This
1533 can normally be checked and set by reading/writing
1534 :file:`/proc/sys/vm/nr_hugepages` on a Linux system. Fio assumes a huge page
1535 is 4MiB in size. So to calculate the number of huge pages you need for a
1536 given job file, add up the I/O depth of all jobs (normally one unless
1537 :option:`iodepth` is used) and multiply by the maximum bs set. Then divide
1538 that number by the huge page size. You can see the size of the huge pages in
1539 :file:`/proc/meminfo`. If no huge pages are allocated by having a non-zero
1540 number in `nr_hugepages`, using **mmaphuge** or **shmhuge** will fail. Also
1541 see :option:`hugepage-size`.
1543 **mmaphuge** also needs to have hugetlbfs mounted and the file location
1544 should point there. So if it's mounted in :file:`/huge`, you would use
1545 `mem=mmaphuge:/huge/somefile`.
1547 .. option:: iomem_align=int, mem_align=int
1549 This indicates the memory alignment of the I/O memory buffers. Note that
1550 the given alignment is applied to the first I/O unit buffer, if using
1551 :option:`iodepth` the alignment of the following buffers are given by the
1552 :option:`bs` used. In other words, if using a :option:`bs` that is a
1553 multiple of the page sized in the system, all buffers will be aligned to
1554 this value. If using a :option:`bs` that is not page aligned, the alignment
1555 of subsequent I/O memory buffers is the sum of the :option:`iomem_align` and
1558 .. option:: hugepage-size=int
1560 Defines the size of a huge page. Must at least be equal to the system
1561 setting, see :file:`/proc/meminfo`. Defaults to 4MiB. Should probably
1562 always be a multiple of megabytes, so using ``hugepage-size=Xm`` is the
1563 preferred way to set this to avoid setting a non-pow-2 bad value.
1565 .. option:: lockmem=int
1567 Pin the specified amount of memory with :manpage:`mlock(2)`. Can be used to
1568 simulate a smaller amount of memory. The amount specified is per worker.
1574 .. option:: size=int
1576 The total size of file I/O for each thread of this job. Fio will run until
1577 this many bytes has been transferred, unless runtime is limited by other options
1578 (such as :option:`runtime`, for instance, or increased/decreased by :option:`io_size`).
1579 Fio will divide this size between the available files determined by options
1580 such as :option:`nrfiles`, :option:`filename`, unless :option:`filesize` is
1581 specified by the job. If the result of division happens to be 0, the size is
1582 set to the physical size of the given files or devices if they exist.
1583 If this option is not specified, fio will use the full size of the given
1584 files or devices. If the files do not exist, size must be given. It is also
1585 possible to give size as a percentage between 1 and 100. If ``size=20%`` is
1586 given, fio will use 20% of the full size of the given files or devices.
1587 Can be combined with :option:`offset` to constrain the start and end range
1588 that I/O will be done within.
1590 .. option:: io_size=int, io_limit=int
1592 Normally fio operates within the region set by :option:`size`, which means
1593 that the :option:`size` option sets both the region and size of I/O to be
1594 performed. Sometimes that is not what you want. With this option, it is
1595 possible to define just the amount of I/O that fio should do. For instance,
1596 if :option:`size` is set to 20GiB and :option:`io_size` is set to 5GiB, fio
1597 will perform I/O within the first 20GiB but exit when 5GiB have been
1598 done. The opposite is also possible -- if :option:`size` is set to 20GiB,
1599 and :option:`io_size` is set to 40GiB, then fio will do 40GiB of I/O within
1600 the 0..20GiB region.
1602 .. option:: filesize=irange(int)
1604 Individual file sizes. May be a range, in which case fio will select sizes
1605 for files at random within the given range and limited to :option:`size` in
1606 total (if that is given). If not given, each created file is the same size.
1607 This option overrides :option:`size` in terms of file size, which means
1608 this value is used as a fixed size or possible range of each file.
1610 .. option:: file_append=bool
1612 Perform I/O after the end of the file. Normally fio will operate within the
1613 size of a file. If this option is set, then fio will append to the file
1614 instead. This has identical behavior to setting :option:`offset` to the size
1615 of a file. This option is ignored on non-regular files.
1617 .. option:: fill_device=bool, fill_fs=bool
1619 Sets size to something really large and waits for ENOSPC (no space left on
1620 device) as the terminating condition. Only makes sense with sequential
1621 write. For a read workload, the mount point will be filled first then I/O
1622 started on the result. This option doesn't make sense if operating on a raw
1623 device node, since the size of that is already known by the file system.
1624 Additionally, writing beyond end-of-device will not return ENOSPC there.
1630 .. option:: ioengine=str
1632 Defines how the job issues I/O to the file. The following types are defined:
1635 Basic :manpage:`read(2)` or :manpage:`write(2)`
1636 I/O. :manpage:`lseek(2)` is used to position the I/O location.
1637 See :option:`fsync` and :option:`fdatasync` for syncing write I/Os.
1640 Basic :manpage:`pread(2)` or :manpage:`pwrite(2)` I/O. Default on
1641 all supported operating systems except for Windows.
1644 Basic :manpage:`readv(2)` or :manpage:`writev(2)` I/O. Will emulate
1645 queuing by coalescing adjacent I/Os into a single submission.
1648 Basic :manpage:`preadv(2)` or :manpage:`pwritev(2)` I/O.
1651 Basic :manpage:`preadv2(2)` or :manpage:`pwritev2(2)` I/O.
1654 Linux native asynchronous I/O. Note that Linux may only support
1655 queued behavior with non-buffered I/O (set ``direct=1`` or
1657 This engine defines engine specific options.
1660 POSIX asynchronous I/O using :manpage:`aio_read(3)` and
1661 :manpage:`aio_write(3)`.
1664 Solaris native asynchronous I/O.
1667 Windows native asynchronous I/O. Default on Windows.
1670 File is memory mapped with :manpage:`mmap(2)` and data copied
1671 to/from using :manpage:`memcpy(3)`.
1674 :manpage:`splice(2)` is used to transfer the data and
1675 :manpage:`vmsplice(2)` to transfer data from user space to the
1679 SCSI generic sg v3 I/O. May either be synchronous using the SG_IO
1680 ioctl, or if the target is an sg character device we use
1681 :manpage:`read(2)` and :manpage:`write(2)` for asynchronous
1682 I/O. Requires :option:`filename` option to specify either block or
1686 Doesn't transfer any data, just pretends to. This is mainly used to
1687 exercise fio itself and for debugging/testing purposes.
1690 Transfer over the network to given ``host:port``. Depending on the
1691 :option:`protocol` used, the :option:`hostname`, :option:`port`,
1692 :option:`listen` and :option:`filename` options are used to specify
1693 what sort of connection to make, while the :option:`protocol` option
1694 determines which protocol will be used. This engine defines engine
1698 Like **net**, but uses :manpage:`splice(2)` and
1699 :manpage:`vmsplice(2)` to map data and send/receive.
1700 This engine defines engine specific options.
1703 Doesn't transfer any data, but burns CPU cycles according to the
1704 :option:`cpuload` and :option:`cpuchunks` options. Setting
1705 :option:`cpuload`\=85 will cause that job to do nothing but burn 85%
1706 of the CPU. In case of SMP machines, use :option:`numjobs`=<nr_of_cpu>
1707 to get desired CPU usage, as the cpuload only loads a
1708 single CPU at the desired rate. A job never finishes unless there is
1709 at least one non-cpuio job.
1712 The GUASI I/O engine is the Generic Userspace Asyncronous Syscall
1713 Interface approach to async I/O. See
1715 http://www.xmailserver.org/guasi-lib.html
1717 for more info on GUASI.
1720 The RDMA I/O engine supports both RDMA memory semantics
1721 (RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the
1722 InfiniBand, RoCE and iWARP protocols.
1725 I/O engine that does regular fallocate to simulate data transfer as
1729 does fallocate(,mode = FALLOC_FL_KEEP_SIZE,).
1732 does fallocate(,mode = 0).
1735 does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE).
1738 I/O engine that sends :manpage:`ftruncate(2)` operations in response
1739 to write (DDIR_WRITE) events. Each ftruncate issued sets the file's
1740 size to the current block offset. :option:`blocksize` is ignored.
1743 I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate
1744 defragment activity in request to DDIR_WRITE event.
1747 I/O engine supporting direct access to Ceph Rados Block Devices
1748 (RBD) via librbd without the need to use the kernel rbd driver. This
1749 ioengine defines engine specific options.
1752 Using GlusterFS libgfapi sync interface to direct access to
1753 GlusterFS volumes without having to go through FUSE. This ioengine
1754 defines engine specific options.
1757 Using GlusterFS libgfapi async interface to direct access to
1758 GlusterFS volumes without having to go through FUSE. This ioengine
1759 defines engine specific options.
1762 Read and write through Hadoop (HDFS). The :option:`filename` option
1763 is used to specify host,port of the hdfs name-node to connect. This
1764 engine interprets offsets a little differently. In HDFS, files once
1765 created cannot be modified so random writes are not possible. To
1766 imitate this the libhdfs engine expects a bunch of small files to be
1767 created over HDFS and will randomly pick a file from them
1768 based on the offset generated by fio backend (see the example
1769 job file to create such files, use ``rw=write`` option). Please
1770 note, it may be necessary to set environment variables to work
1771 with HDFS/libhdfs properly. Each job uses its own connection to
1775 Read, write and erase an MTD character device (e.g.,
1776 :file:`/dev/mtd0`). Discards are treated as erases. Depending on the
1777 underlying device type, the I/O may have to go in a certain pattern,
1778 e.g., on NAND, writing sequentially to erase blocks and discarding
1779 before overwriting. The `trimwrite` mode works well for this
1783 Read and write using filesystem DAX to a file on a filesystem
1784 mounted with DAX on a persistent memory device through the NVML
1788 Read and write using device DAX to a persistent memory device (e.g.,
1789 /dev/dax0.0) through the NVML libpmem library.
1792 Prefix to specify loading an external I/O engine object file. Append
1793 the engine filename, e.g. ``ioengine=external:/tmp/foo.o`` to load
1794 ioengine :file:`foo.o` in :file:`/tmp`. The path can be either
1795 absolute or relative. See :file:`engines/skeleton_external.c` for
1796 details of writing an external I/O engine.
1799 I/O engine specific parameters
1800 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1802 In addition, there are some parameters which are only valid when a specific
1803 :option:`ioengine` is in use. These are used identically to normal parameters,
1804 with the caveat that when used on the command line, they must come after the
1805 :option:`ioengine` that defines them is selected.
1807 .. option:: userspace_reap : [libaio]
1809 Normally, with the libaio engine in use, fio will use the
1810 :manpage:`io_getevents(2)` system call to reap newly returned events. With
1811 this flag turned on, the AIO ring will be read directly from user-space to
1812 reap events. The reaping mode is only enabled when polling for a minimum of
1813 0 events (e.g. when :option:`iodepth_batch_complete` `=0`).
1815 .. option:: hipri : [pvsync2]
1817 Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority
1820 .. option:: hipri_percentage : [pvsync2]
1822 When hipri is set this determines the probability of a pvsync2 I/O being high
1823 priority. The default is 100%.
1825 .. option:: cpuload=int : [cpuio]
1827 Attempt to use the specified percentage of CPU cycles. This is a mandatory
1828 option when using cpuio I/O engine.
1830 .. option:: cpuchunks=int : [cpuio]
1832 Split the load into cycles of the given time. In microseconds.
1834 .. option:: exit_on_io_done=bool : [cpuio]
1836 Detect when I/O threads are done, then exit.
1838 .. option:: namenode=str : [libhdfs]
1840 The hostname or IP address of a HDFS cluster namenode to contact.
1842 .. option:: port=int
1846 The listening port of the HFDS cluster namenode.
1850 The TCP or UDP port to bind to or connect to. If this is used with
1851 :option:`numjobs` to spawn multiple instances of the same job type, then
1852 this will be the starting port number since fio will use a range of
1855 .. option:: hostname=str : [netsplice] [net]
1857 The hostname or IP address to use for TCP or UDP based I/O. If the job is
1858 a TCP listener or UDP reader, the hostname is not used and must be omitted
1859 unless it is a valid UDP multicast address.
1861 .. option:: interface=str : [netsplice] [net]
1863 The IP address of the network interface used to send or receive UDP
1866 .. option:: ttl=int : [netsplice] [net]
1868 Time-to-live value for outgoing UDP multicast packets. Default: 1.
1870 .. option:: nodelay=bool : [netsplice] [net]
1872 Set TCP_NODELAY on TCP connections.
1874 .. option:: protocol=str, proto=str : [netsplice] [net]
1876 The network protocol to use. Accepted values are:
1879 Transmission control protocol.
1881 Transmission control protocol V6.
1883 User datagram protocol.
1885 User datagram protocol V6.
1889 When the protocol is TCP or UDP, the port must also be given, as well as the
1890 hostname if the job is a TCP listener or UDP reader. For unix sockets, the
1891 normal :option:`filename` option should be used and the port is invalid.
1893 .. option:: listen : [netsplice] [net]
1895 For TCP network connections, tell fio to listen for incoming connections
1896 rather than initiating an outgoing connection. The :option:`hostname` must
1897 be omitted if this option is used.
1899 .. option:: pingpong : [netsplice] [net]
1901 Normally a network writer will just continue writing data, and a network
1902 reader will just consume packages. If ``pingpong=1`` is set, a writer will
1903 send its normal payload to the reader, then wait for the reader to send the
1904 same payload back. This allows fio to measure network latencies. The
1905 submission and completion latencies then measure local time spent sending or
1906 receiving, and the completion latency measures how long it took for the
1907 other end to receive and send back. For UDP multicast traffic
1908 ``pingpong=1`` should only be set for a single reader when multiple readers
1909 are listening to the same address.
1911 .. option:: window_size : [netsplice] [net]
1913 Set the desired socket buffer size for the connection.
1915 .. option:: mss : [netsplice] [net]
1917 Set the TCP maximum segment size (TCP_MAXSEG).
1919 .. option:: donorname=str : [e4defrag]
1921 File will be used as a block donor (swap extents between files).
1923 .. option:: inplace=int : [e4defrag]
1925 Configure donor file blocks allocation strategy:
1928 Default. Preallocate donor's file on init.
1930 Allocate space immediately inside defragment event, and free right
1933 .. option:: clustername=str : [rbd]
1935 Specifies the name of the Ceph cluster.
1937 .. option:: rbdname=str : [rbd]
1939 Specifies the name of the RBD.
1941 .. option:: pool=str : [rbd]
1943 Specifies the name of the Ceph pool containing RBD.
1945 .. option:: clientname=str : [rbd]
1947 Specifies the username (without the 'client.' prefix) used to access the
1948 Ceph cluster. If the *clustername* is specified, the *clientname* shall be
1949 the full *type.id* string. If no type. prefix is given, fio will add
1950 'client.' by default.
1952 .. option:: skip_bad=bool : [mtd]
1954 Skip operations against known bad blocks.
1956 .. option:: hdfsdirectory : [libhdfs]
1958 libhdfs will create chunk in this HDFS directory.
1960 .. option:: chunk_size : [libhdfs]
1962 The size of the chunk to use for each file.
1968 .. option:: iodepth=int
1970 Number of I/O units to keep in flight against the file. Note that
1971 increasing *iodepth* beyond 1 will not affect synchronous ioengines (except
1972 for small degrees when :option:`verify_async` is in use). Even async
1973 engines may impose OS restrictions causing the desired depth not to be
1974 achieved. This may happen on Linux when using libaio and not setting
1975 :option:`direct`\=1, since buffered I/O is not async on that OS. Keep an
1976 eye on the I/O depth distribution in the fio output to verify that the
1977 achieved depth is as expected. Default: 1.
1979 .. option:: iodepth_batch_submit=int, iodepth_batch=int
1981 This defines how many pieces of I/O to submit at once. It defaults to 1
1982 which means that we submit each I/O as soon as it is available, but can be
1983 raised to submit bigger batches of I/O at the time. If it is set to 0 the
1984 :option:`iodepth` value will be used.
1986 .. option:: iodepth_batch_complete_min=int, iodepth_batch_complete=int
1988 This defines how many pieces of I/O to retrieve at once. It defaults to 1
1989 which means that we'll ask for a minimum of 1 I/O in the retrieval process
1990 from the kernel. The I/O retrieval will go on until we hit the limit set by
1991 :option:`iodepth_low`. If this variable is set to 0, then fio will always
1992 check for completed events before queuing more I/O. This helps reduce I/O
1993 latency, at the cost of more retrieval system calls.
1995 .. option:: iodepth_batch_complete_max=int
1997 This defines maximum pieces of I/O to retrieve at once. This variable should
1998 be used along with :option:`iodepth_batch_complete_min`\=int variable,
1999 specifying the range of min and max amount of I/O which should be
2000 retrieved. By default it is equal to the :option:`iodepth_batch_complete_min`
2005 iodepth_batch_complete_min=1
2006 iodepth_batch_complete_max=<iodepth>
2008 which means that we will retrieve at least 1 I/O and up to the whole
2009 submitted queue depth. If none of I/O has been completed yet, we will wait.
2013 iodepth_batch_complete_min=0
2014 iodepth_batch_complete_max=<iodepth>
2016 which means that we can retrieve up to the whole submitted queue depth, but
2017 if none of I/O has been completed yet, we will NOT wait and immediately exit
2018 the system call. In this example we simply do polling.
2020 .. option:: iodepth_low=int
2022 The low water mark indicating when to start filling the queue
2023 again. Defaults to the same as :option:`iodepth`, meaning that fio will
2024 attempt to keep the queue full at all times. If :option:`iodepth` is set to
2025 e.g. 16 and *iodepth_low* is set to 4, then after fio has filled the queue of
2026 16 requests, it will let the depth drain down to 4 before starting to fill
2029 .. option:: serialize_overlap=bool
2031 Serialize in-flight I/Os that might otherwise cause or suffer from data races.
2032 When two or more I/Os are submitted simultaneously, there is no guarantee that
2033 the I/Os will be processed or completed in the submitted order. Further, if
2034 two or more of those I/Os are writes, any overlapping region between them can
2035 become indeterminate/undefined on certain storage. These issues can cause
2036 verification to fail erratically when at least one of the racing I/Os is
2037 changing data and the overlapping region has a non-zero size. Setting
2038 ``serialize_overlap`` tells fio to avoid provoking this behavior by explicitly
2039 serializing in-flight I/Os that have a non-zero overlap. Note that setting
2040 this option can reduce both performance and the `:option:iodepth` achieved.
2041 Additionally this option does not work when :option:`io_submit_mode` is set to
2042 offload. Default: false.
2044 .. option:: io_submit_mode=str
2046 This option controls how fio submits the I/O to the I/O engine. The default
2047 is `inline`, which means that the fio job threads submit and reap I/O
2048 directly. If set to `offload`, the job threads will offload I/O submission
2049 to a dedicated pool of I/O threads. This requires some coordination and thus
2050 has a bit of extra overhead, especially for lower queue depth I/O where it
2051 can increase latencies. The benefit is that fio can manage submission rates
2052 independently of the device completion rates. This avoids skewed latency
2053 reporting if I/O gets backed up on the device side (the coordinated omission
2060 .. option:: thinktime=time
2062 Stall the job for the specified period of time after an I/O has completed before issuing the
2063 next. May be used to simulate processing being done by an application.
2064 When the unit is omitted, the value is interpreted in microseconds. See
2065 :option:`thinktime_blocks` and :option:`thinktime_spin`.
2067 .. option:: thinktime_spin=time
2069 Only valid if :option:`thinktime` is set - pretend to spend CPU time doing
2070 something with the data received, before falling back to sleeping for the
2071 rest of the period specified by :option:`thinktime`. When the unit is
2072 omitted, the value is interpreted in microseconds.
2074 .. option:: thinktime_blocks=int
2076 Only valid if :option:`thinktime` is set - control how many blocks to issue,
2077 before waiting :option:`thinktime` usecs. If not set, defaults to 1 which will make
2078 fio wait :option:`thinktime` usecs after every block. This effectively makes any
2079 queue depth setting redundant, since no more than 1 I/O will be queued
2080 before we have to complete it and do our :option:`thinktime`. In other words, this
2081 setting effectively caps the queue depth if the latter is larger.
2083 .. option:: rate=int[,int][,int]
2085 Cap the bandwidth used by this job. The number is in bytes/sec, the normal
2086 suffix rules apply. Comma-separated values may be specified for reads,
2087 writes, and trims as described in :option:`blocksize`.
2089 For example, using `rate=1m,500k` would limit reads to 1MiB/sec and writes to
2090 500KiB/sec. Capping only reads or writes can be done with `rate=,500k` or
2091 `rate=500k,` where the former will only limit writes (to 500KiB/sec) and the
2092 latter will only limit reads.
2094 .. option:: rate_min=int[,int][,int]
2096 Tell fio to do whatever it can to maintain at least this bandwidth. Failing
2097 to meet this requirement will cause the job to exit. Comma-separated values
2098 may be specified for reads, writes, and trims as described in
2099 :option:`blocksize`.
2101 .. option:: rate_iops=int[,int][,int]
2103 Cap the bandwidth to this number of IOPS. Basically the same as
2104 :option:`rate`, just specified independently of bandwidth. If the job is
2105 given a block size range instead of a fixed value, the smallest block size
2106 is used as the metric. Comma-separated values may be specified for reads,
2107 writes, and trims as described in :option:`blocksize`.
2109 .. option:: rate_iops_min=int[,int][,int]
2111 If fio doesn't meet this rate of I/O, it will cause the job to exit.
2112 Comma-separated values may be specified for reads, writes, and trims as
2113 described in :option:`blocksize`.
2115 .. option:: rate_process=str
2117 This option controls how fio manages rated I/O submissions. The default is
2118 `linear`, which submits I/O in a linear fashion with fixed delays between
2119 I/Os that gets adjusted based on I/O completion rates. If this is set to
2120 `poisson`, fio will submit I/O based on a more real world random request
2121 flow, known as the Poisson process
2122 (https://en.wikipedia.org/wiki/Poisson_point_process). The lambda will be
2123 10^6 / IOPS for the given workload.
2129 .. option:: latency_target=time
2131 If set, fio will attempt to find the max performance point that the given
2132 workload will run at while maintaining a latency below this target. When
2133 the unit is omitted, the value is interpreted in microseconds. See
2134 :option:`latency_window` and :option:`latency_percentile`.
2136 .. option:: latency_window=time
2138 Used with :option:`latency_target` to specify the sample window that the job
2139 is run at varying queue depths to test the performance. When the unit is
2140 omitted, the value is interpreted in microseconds.
2142 .. option:: latency_percentile=float
2144 The percentage of I/Os that must fall within the criteria specified by
2145 :option:`latency_target` and :option:`latency_window`. If not set, this
2146 defaults to 100.0, meaning that all I/Os must be equal or below to the value
2147 set by :option:`latency_target`.
2149 .. option:: max_latency=time
2151 If set, fio will exit the job with an ETIMEDOUT error if it exceeds this
2152 maximum latency. When the unit is omitted, the value is interpreted in
2155 .. option:: rate_cycle=int
2157 Average bandwidth for :option:`rate` and :option:`rate_min` over this number
2158 of milliseconds. Defaults to 1000.
2164 .. option:: write_iolog=str
2166 Write the issued I/O patterns to the specified file. See
2167 :option:`read_iolog`. Specify a separate file for each job, otherwise the
2168 iologs will be interspersed and the file may be corrupt.
2170 .. option:: read_iolog=str
2172 Open an iolog with the specified filename and replay the I/O patterns it
2173 contains. This can be used to store a workload and replay it sometime
2174 later. The iolog given may also be a blktrace binary file, which allows fio
2175 to replay a workload captured by :command:`blktrace`. See
2176 :manpage:`blktrace(8)` for how to capture such logging data. For blktrace
2177 replay, the file needs to be turned into a blkparse binary data file first
2178 (``blkparse <device> -o /dev/null -d file_for_fio.bin``).
2180 .. option:: replay_no_stall=bool
2182 When replaying I/O with :option:`read_iolog` the default behavior is to
2183 attempt to respect the timestamps within the log and replay them with the
2184 appropriate delay between IOPS. By setting this variable fio will not
2185 respect the timestamps and attempt to replay them as fast as possible while
2186 still respecting ordering. The result is the same I/O pattern to a given
2187 device, but different timings.
2189 .. option:: replay_redirect=str
2191 While replaying I/O patterns using :option:`read_iolog` the default behavior
2192 is to replay the IOPS onto the major/minor device that each IOP was recorded
2193 from. This is sometimes undesirable because on a different machine those
2194 major/minor numbers can map to a different device. Changing hardware on the
2195 same system can also result in a different major/minor mapping.
2196 ``replay_redirect`` causes all I/Os to be replayed onto the single specified
2197 device regardless of the device it was recorded
2198 from. i.e. :option:`replay_redirect`\= :file:`/dev/sdc` would cause all I/O
2199 in the blktrace or iolog to be replayed onto :file:`/dev/sdc`. This means
2200 multiple devices will be replayed onto a single device, if the trace
2201 contains multiple devices. If you want multiple devices to be replayed
2202 concurrently to multiple redirected devices you must blkparse your trace
2203 into separate traces and replay them with independent fio invocations.
2204 Unfortunately this also breaks the strict time ordering between multiple
2207 .. option:: replay_align=int
2209 Force alignment of I/O offsets and lengths in a trace to this power of 2
2212 .. option:: replay_scale=int
2214 Scale sector offsets down by this factor when replaying traces.
2217 Threads, processes and job synchronization
2218 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2222 Fio defaults to creating jobs by using fork, however if this option is
2223 given, fio will create jobs by using POSIX Threads' function
2224 :manpage:`pthread_create(3)` to create threads instead.
2226 .. option:: wait_for=str
2228 If set, the current job won't be started until all workers of the specified
2229 waitee job are done.
2231 ``wait_for`` operates on the job name basis, so there are a few
2232 limitations. First, the waitee must be defined prior to the waiter job
2233 (meaning no forward references). Second, if a job is being referenced as a
2234 waitee, it must have a unique name (no duplicate waitees).
2236 .. option:: nice=int
2238 Run the job with the given nice value. See man :manpage:`nice(2)`.
2240 On Windows, values less than -15 set the process class to "High"; -1 through
2241 -15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle"
2244 .. option:: prio=int
2246 Set the I/O priority value of this job. Linux limits us to a positive value
2247 between 0 and 7, with 0 being the highest. See man
2248 :manpage:`ionice(1)`. Refer to an appropriate manpage for other operating
2249 systems since meaning of priority may differ.
2251 .. option:: prioclass=int
2253 Set the I/O priority class. See man :manpage:`ionice(1)`.
2255 .. option:: cpumask=int
2257 Set the CPU affinity of this job. The parameter given is a bit mask of
2258 allowed CPUs the job may run on. So if you want the allowed CPUs to be 1
2259 and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man
2260 :manpage:`sched_setaffinity(2)`. This may not work on all supported
2261 operating systems or kernel versions. This option doesn't work well for a
2262 higher CPU count than what you can store in an integer mask, so it can only
2263 control cpus 1-32. For boxes with larger CPU counts, use
2264 :option:`cpus_allowed`.
2266 .. option:: cpus_allowed=str
2268 Controls the same options as :option:`cpumask`, but accepts a textual
2269 specification of the permitted CPUs instead. So to use CPUs 1 and 5 you
2270 would specify ``cpus_allowed=1,5``. This option also allows a range of CPUs
2271 to be specified -- say you wanted a binding to CPUs 1, 5, and 8 to 15, you
2272 would set ``cpus_allowed=1,5,8-15``.
2274 .. option:: cpus_allowed_policy=str
2276 Set the policy of how fio distributes the CPUs specified by
2277 :option:`cpus_allowed` or :option:`cpumask`. Two policies are supported:
2280 All jobs will share the CPU set specified.
2282 Each job will get a unique CPU from the CPU set.
2284 **shared** is the default behavior, if the option isn't specified. If
2285 **split** is specified, then fio will will assign one cpu per job. If not
2286 enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs
2289 .. option:: numa_cpu_nodes=str
2291 Set this job running on specified NUMA nodes' CPUs. The arguments allow
2292 comma delimited list of cpu numbers, A-B ranges, or `all`. Note, to enable
2293 NUMA options support, fio must be built on a system with libnuma-dev(el)
2296 .. option:: numa_mem_policy=str
2298 Set this job's memory policy and corresponding NUMA nodes. Format of the
2303 ``mode`` is one of the following memory poicies: ``default``, ``prefer``,
2304 ``bind``, ``interleave`` or ``local``. For ``default`` and ``local`` memory
2305 policies, no node needs to be specified. For ``prefer``, only one node is
2306 allowed. For ``bind`` and ``interleave`` the ``nodelist`` may be as
2307 follows: a comma delimited list of numbers, A-B ranges, or `all`.
2309 .. option:: cgroup=str
2311 Add job to this control group. If it doesn't exist, it will be created. The
2312 system must have a mounted cgroup blkio mount point for this to work. If
2313 your system doesn't have it mounted, you can do so with::
2315 # mount -t cgroup -o blkio none /cgroup
2317 .. option:: cgroup_weight=int
2319 Set the weight of the cgroup to this value. See the documentation that comes
2320 with the kernel, allowed values are in the range of 100..1000.
2322 .. option:: cgroup_nodelete=bool
2324 Normally fio will delete the cgroups it has created after the job
2325 completion. To override this behavior and to leave cgroups around after the
2326 job completion, set ``cgroup_nodelete=1``. This can be useful if one wants
2327 to inspect various cgroup files after job completion. Default: false.
2329 .. option:: flow_id=int
2331 The ID of the flow. If not specified, it defaults to being a global
2332 flow. See :option:`flow`.
2334 .. option:: flow=int
2336 Weight in token-based flow control. If this value is used, then there is a
2337 'flow counter' which is used to regulate the proportion of activity between
2338 two or more jobs. Fio attempts to keep this flow counter near zero. The
2339 ``flow`` parameter stands for how much should be added or subtracted to the
2340 flow counter on each iteration of the main I/O loop. That is, if one job has
2341 ``flow=8`` and another job has ``flow=-1``, then there will be a roughly 1:8
2342 ratio in how much one runs vs the other.
2344 .. option:: flow_watermark=int
2346 The maximum value that the absolute value of the flow counter is allowed to
2347 reach before the job must wait for a lower value of the counter.
2349 .. option:: flow_sleep=int
2351 The period of time, in microseconds, to wait after the flow watermark has
2352 been exceeded before retrying operations.
2354 .. option:: stonewall, wait_for_previous
2356 Wait for preceding jobs in the job file to exit, before starting this
2357 one. Can be used to insert serialization points in the job file. A stone
2358 wall also implies starting a new reporting group, see
2359 :option:`group_reporting`.
2363 By default, fio will continue running all other jobs when one job finishes
2364 but sometimes this is not the desired action. Setting ``exitall`` will
2365 instead make fio terminate all other jobs when one job finishes.
2367 .. option:: exec_prerun=str
2369 Before running this job, issue the command specified through
2370 :manpage:`system(3)`. Output is redirected in a file called
2371 :file:`jobname.prerun.txt`.
2373 .. option:: exec_postrun=str
2375 After the job completes, issue the command specified though
2376 :manpage:`system(3)`. Output is redirected in a file called
2377 :file:`jobname.postrun.txt`.
2381 Instead of running as the invoking user, set the user ID to this value
2382 before the thread/process does any work.
2386 Set group ID, see :option:`uid`.
2392 .. option:: verify_only
2394 Do not perform specified workload, only verify data still matches previous
2395 invocation of this workload. This option allows one to check data multiple
2396 times at a later date without overwriting it. This option makes sense only
2397 for workloads that write data, and does not support workloads with the
2398 :option:`time_based` option set.
2400 .. option:: do_verify=bool
2402 Run the verify phase after a write phase. Only valid if :option:`verify` is
2405 .. option:: verify=str
2407 If writing to a file, fio can verify the file contents after each iteration
2408 of the job. Each verification method also implies verification of special
2409 header, which is written to the beginning of each block. This header also
2410 includes meta information, like offset of the block, block number, timestamp
2411 when block was written, etc. :option:`verify` can be combined with
2412 :option:`verify_pattern` option. The allowed values are:
2415 Use an md5 sum of the data area and store it in the header of
2419 Use an experimental crc64 sum of the data area and store it in the
2420 header of each block.
2423 Use a crc32c sum of the data area and store it in the header of
2424 each block. This will automatically use hardware acceleration
2425 (e.g. SSE4.2 on an x86 or CRC crypto extensions on ARM64) but will
2426 fall back to software crc32c if none is found. Generally the
2427 fatest checksum fio supports when hardware accelerated.
2433 Use a crc32 sum of the data area and store it in the header of each
2437 Use a crc16 sum of the data area and store it in the header of each
2441 Use a crc7 sum of the data area and store it in the header of each
2445 Use xxhash as the checksum function. Generally the fastest software
2446 checksum that fio supports.
2449 Use sha512 as the checksum function.
2452 Use sha256 as the checksum function.
2455 Use optimized sha1 as the checksum function.
2458 Use optimized sha3-224 as the checksum function.
2461 Use optimized sha3-256 as the checksum function.
2464 Use optimized sha3-384 as the checksum function.
2467 Use optimized sha3-512 as the checksum function.
2470 This option is deprecated, since now meta information is included in
2471 generic verification header and meta verification happens by
2472 default. For detailed information see the description of the
2473 :option:`verify` setting. This option is kept because of
2474 compatibility's sake with old configurations. Do not use it.
2477 Verify a strict pattern. Normally fio includes a header with some
2478 basic information and checksumming, but if this option is set, only
2479 the specific pattern set with :option:`verify_pattern` is verified.
2482 Only pretend to verify. Useful for testing internals with
2483 :option:`ioengine`\=null, not for much else.
2485 This option can be used for repeated burn-in tests of a system to make sure
2486 that the written data is also correctly read back. If the data direction
2487 given is a read or random read, fio will assume that it should verify a
2488 previously written file. If the data direction includes any form of write,
2489 the verify will be of the newly written data.
2491 .. option:: verifysort=bool
2493 If true, fio will sort written verify blocks when it deems it faster to read
2494 them back in a sorted manner. This is often the case when overwriting an
2495 existing file, since the blocks are already laid out in the file system. You
2496 can ignore this option unless doing huge amounts of really fast I/O where
2497 the red-black tree sorting CPU time becomes significant. Default: true.
2499 .. option:: verifysort_nr=int
2501 Pre-load and sort verify blocks for a read workload.
2503 .. option:: verify_offset=int
2505 Swap the verification header with data somewhere else in the block before
2506 writing. It is swapped back before verifying.
2508 .. option:: verify_interval=int
2510 Write the verification header at a finer granularity than the
2511 :option:`blocksize`. It will be written for chunks the size of
2512 ``verify_interval``. :option:`blocksize` should divide this evenly.
2514 .. option:: verify_pattern=str
2516 If set, fio will fill the I/O buffers with this pattern. Fio defaults to
2517 filling with totally random bytes, but sometimes it's interesting to fill
2518 with a known pattern for I/O verification purposes. Depending on the width
2519 of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time (it can
2520 be either a decimal or a hex number). The ``verify_pattern`` if larger than
2521 a 32-bit quantity has to be a hex number that starts with either "0x" or
2522 "0X". Use with :option:`verify`. Also, ``verify_pattern`` supports %o
2523 format, which means that for each block offset will be written and then
2524 verified back, e.g.::
2528 Or use combination of everything::
2530 verify_pattern=0xff%o"abcd"-12
2532 .. option:: verify_fatal=bool
2534 Normally fio will keep checking the entire contents before quitting on a
2535 block verification failure. If this option is set, fio will exit the job on
2536 the first observed failure. Default: false.
2538 .. option:: verify_dump=bool
2540 If set, dump the contents of both the original data block and the data block
2541 we read off disk to files. This allows later analysis to inspect just what
2542 kind of data corruption occurred. Off by default.
2544 .. option:: verify_async=int
2546 Fio will normally verify I/O inline from the submitting thread. This option
2547 takes an integer describing how many async offload threads to create for I/O
2548 verification instead, causing fio to offload the duty of verifying I/O
2549 contents to one or more separate threads. If using this offload option, even
2550 sync I/O engines can benefit from using an :option:`iodepth` setting higher
2551 than 1, as it allows them to have I/O in flight while verifies are running.
2552 Defaults to 0 async threads, i.e. verification is not asynchronous.
2554 .. option:: verify_async_cpus=str
2556 Tell fio to set the given CPU affinity on the async I/O verification
2557 threads. See :option:`cpus_allowed` for the format used.
2559 .. option:: verify_backlog=int
2561 Fio will normally verify the written contents of a job that utilizes verify
2562 once that job has completed. In other words, everything is written then
2563 everything is read back and verified. You may want to verify continually
2564 instead for a variety of reasons. Fio stores the meta data associated with
2565 an I/O block in memory, so for large verify workloads, quite a bit of memory
2566 would be used up holding this meta data. If this option is enabled, fio will
2567 write only N blocks before verifying these blocks.
2569 .. option:: verify_backlog_batch=int
2571 Control how many blocks fio will verify if :option:`verify_backlog` is
2572 set. If not set, will default to the value of :option:`verify_backlog`
2573 (meaning the entire queue is read back and verified). If
2574 ``verify_backlog_batch`` is less than :option:`verify_backlog` then not all
2575 blocks will be verified, if ``verify_backlog_batch`` is larger than
2576 :option:`verify_backlog`, some blocks will be verified more than once.
2578 .. option:: verify_state_save=bool
2580 When a job exits during the write phase of a verify workload, save its
2581 current state. This allows fio to replay up until that point, if the verify
2582 state is loaded for the verify read phase. The format of the filename is,
2585 <type>-<jobname>-<jobindex>-verify.state.
2587 <type> is "local" for a local run, "sock" for a client/server socket
2588 connection, and "ip" (192.168.0.1, for instance) for a networked
2589 client/server connection. Defaults to true.
2591 .. option:: verify_state_load=bool
2593 If a verify termination trigger was used, fio stores the current write state
2594 of each thread. This can be used at verification time so that fio knows how
2595 far it should verify. Without this information, fio will run a full
2596 verification pass, according to the settings in the job file used. Default
2599 .. option:: trim_percentage=int
2601 Number of verify blocks to discard/trim.
2603 .. option:: trim_verify_zero=bool
2605 Verify that trim/discarded blocks are returned as zeros.
2607 .. option:: trim_backlog=int
2609 Trim after this number of blocks are written.
2611 .. option:: trim_backlog_batch=int
2613 Trim this number of I/O blocks.
2615 .. option:: experimental_verify=bool
2617 Enable experimental verification.
2622 .. option:: steadystate=str:float, ss=str:float
2624 Define the criterion and limit for assessing steady state performance. The
2625 first parameter designates the criterion whereas the second parameter sets
2626 the threshold. When the criterion falls below the threshold for the
2627 specified duration, the job will stop. For example, `iops_slope:0.1%` will
2628 direct fio to terminate the job when the least squares regression slope
2629 falls below 0.1% of the mean IOPS. If :option:`group_reporting` is enabled
2630 this will apply to all jobs in the group. Below is the list of available
2631 steady state assessment criteria. All assessments are carried out using only
2632 data from the rolling collection window. Threshold limits can be expressed
2633 as a fixed value or as a percentage of the mean in the collection window.
2636 Collect IOPS data. Stop the job if all individual IOPS measurements
2637 are within the specified limit of the mean IOPS (e.g., ``iops:2``
2638 means that all individual IOPS values must be within 2 of the mean,
2639 whereas ``iops:0.2%`` means that all individual IOPS values must be
2640 within 0.2% of the mean IOPS to terminate the job).
2643 Collect IOPS data and calculate the least squares regression
2644 slope. Stop the job if the slope falls below the specified limit.
2647 Collect bandwidth data. Stop the job if all individual bandwidth
2648 measurements are within the specified limit of the mean bandwidth.
2651 Collect bandwidth data and calculate the least squares regression
2652 slope. Stop the job if the slope falls below the specified limit.
2654 .. option:: steadystate_duration=time, ss_dur=time
2656 A rolling window of this duration will be used to judge whether steady state
2657 has been reached. Data will be collected once per second. The default is 0
2658 which disables steady state detection. When the unit is omitted, the
2659 value is interpreted in seconds.
2661 .. option:: steadystate_ramp_time=time, ss_ramp=time
2663 Allow the job to run for the specified duration before beginning data
2664 collection for checking the steady state job termination criterion. The
2665 default is 0. When the unit is omitted, the value is interpreted in seconds.
2668 Measurements and reporting
2669 ~~~~~~~~~~~~~~~~~~~~~~~~~~
2671 .. option:: per_job_logs=bool
2673 If set, this generates bw/clat/iops log with per file private filenames. If
2674 not set, jobs with identical names will share the log filename. Default:
2677 .. option:: group_reporting
2679 It may sometimes be interesting to display statistics for groups of jobs as
2680 a whole instead of for each individual job. This is especially true if
2681 :option:`numjobs` is used; looking at individual thread/process output
2682 quickly becomes unwieldy. To see the final report per-group instead of
2683 per-job, use :option:`group_reporting`. Jobs in a file will be part of the
2684 same reporting group, unless if separated by a :option:`stonewall`, or by
2685 using :option:`new_group`.
2687 .. option:: new_group
2689 Start a new reporting group. See: :option:`group_reporting`. If not given,
2690 all jobs in a file will be part of the same reporting group, unless
2691 separated by a :option:`stonewall`.
2693 .. option:: stats=bool
2695 By default, fio collects and shows final output results for all jobs
2696 that run. If this option is set to 0, then fio will ignore it in
2697 the final stat output.
2699 .. option:: write_bw_log=str
2701 If given, write a bandwidth log for this job. Can be used to store data of
2702 the bandwidth of the jobs in their lifetime. The included
2703 :command:`fio_generate_plots` script uses :command:`gnuplot` to turn these
2704 text files into nice graphs. See :option:`write_lat_log` for behavior of
2705 given filename. For this option, the postfix is :file:`_bw.x.log`, where `x`
2706 is the index of the job (`1..N`, where `N` is the number of jobs). If
2707 :option:`per_job_logs` is false, then the filename will not include the job
2708 index. See `Log File Formats`_.
2710 .. option:: write_lat_log=str
2712 Same as :option:`write_bw_log`, except that this option stores I/O
2713 submission, completion, and total latencies instead. If no filename is given
2714 with this option, the default filename of :file:`jobname_type.log` is
2715 used. Even if the filename is given, fio will still append the type of
2716 log. So if one specifies::
2720 The actual log names will be :file:`foo_slat.x.log`, :file:`foo_clat.x.log`,
2721 and :file:`foo_lat.x.log`, where `x` is the index of the job (`1..N`, where `N`
2722 is the number of jobs). This helps :command:`fio_generate_plots` find the
2723 logs automatically. If :option:`per_job_logs` is false, then the filename
2724 will not include the job index. See `Log File Formats`_.
2726 .. option:: write_hist_log=str
2728 Same as :option:`write_lat_log`, but writes I/O completion latency
2729 histograms. If no filename is given with this option, the default filename
2730 of :file:`jobname_clat_hist.x.log` is used, where `x` is the index of the
2731 job (`1..N`, where `N` is the number of jobs). Even if the filename is given,
2732 fio will still append the type of log. If :option:`per_job_logs` is false,
2733 then the filename will not include the job index. See `Log File Formats`_.
2735 .. option:: write_iops_log=str
2737 Same as :option:`write_bw_log`, but writes IOPS. If no filename is given
2738 with this option, the default filename of :file:`jobname_type.x.log` is
2739 used, where `x` is the index of the job (`1..N`, where `N` is the number of
2740 jobs). Even if the filename is given, fio will still append the type of
2741 log. If :option:`per_job_logs` is false, then the filename will not include
2742 the job index. See `Log File Formats`_.
2744 .. option:: log_avg_msec=int
2746 By default, fio will log an entry in the iops, latency, or bw log for every
2747 I/O that completes. When writing to the disk log, that can quickly grow to a
2748 very large size. Setting this option makes fio average the each log entry
2749 over the specified period of time, reducing the resolution of the log. See
2750 :option:`log_max_value` as well. Defaults to 0, logging all entries.
2751 Also see `Log File Formats`_.
2753 .. option:: log_hist_msec=int
2755 Same as :option:`log_avg_msec`, but logs entries for completion latency
2756 histograms. Computing latency percentiles from averages of intervals using
2757 :option:`log_avg_msec` is inaccurate. Setting this option makes fio log
2758 histogram entries over the specified period of time, reducing log sizes for
2759 high IOPS devices while retaining percentile accuracy. See
2760 :option:`log_hist_coarseness` as well. Defaults to 0, meaning histogram
2761 logging is disabled.
2763 .. option:: log_hist_coarseness=int
2765 Integer ranging from 0 to 6, defining the coarseness of the resolution of
2766 the histogram logs enabled with :option:`log_hist_msec`. For each increment
2767 in coarseness, fio outputs half as many bins. Defaults to 0, for which
2768 histogram logs contain 1216 latency bins. See `Log File Formats`_.
2770 .. option:: log_max_value=bool
2772 If :option:`log_avg_msec` is set, fio logs the average over that window. If
2773 you instead want to log the maximum value, set this option to 1. Defaults to
2774 0, meaning that averaged values are logged.
2776 .. option:: log_offset=bool
2778 If this is set, the iolog options will include the byte offset for the I/O
2779 entry as well as the other data values. Defaults to 0 meaning that
2780 offsets are not present in logs. Also see `Log File Formats`_.
2782 .. option:: log_compression=int
2784 If this is set, fio will compress the I/O logs as it goes, to keep the
2785 memory footprint lower. When a log reaches the specified size, that chunk is
2786 removed and compressed in the background. Given that I/O logs are fairly
2787 highly compressible, this yields a nice memory savings for longer runs. The
2788 downside is that the compression will consume some background CPU cycles, so
2789 it may impact the run. This, however, is also true if the logging ends up
2790 consuming most of the system memory. So pick your poison. The I/O logs are
2791 saved normally at the end of a run, by decompressing the chunks and storing
2792 them in the specified log file. This feature depends on the availability of
2795 .. option:: log_compression_cpus=str
2797 Define the set of CPUs that are allowed to handle online log compression for
2798 the I/O jobs. This can provide better isolation between performance
2799 sensitive jobs, and background compression work.
2801 .. option:: log_store_compressed=bool
2803 If set, fio will store the log files in a compressed format. They can be
2804 decompressed with fio, using the :option:`--inflate-log` command line
2805 parameter. The files will be stored with a :file:`.fz` suffix.
2807 .. option:: log_unix_epoch=bool
2809 If set, fio will log Unix timestamps to the log files produced by enabling
2810 write_type_log for each log type, instead of the default zero-based
2813 .. option:: block_error_percentiles=bool
2815 If set, record errors in trim block-sized units from writes and trims and
2816 output a histogram of how many trims it took to get to errors, and what kind
2817 of error was encountered.
2819 .. option:: bwavgtime=int
2821 Average the calculated bandwidth over the given time. Value is specified in
2822 milliseconds. If the job also does bandwidth logging through
2823 :option:`write_bw_log`, then the minimum of this option and
2824 :option:`log_avg_msec` will be used. Default: 500ms.
2826 .. option:: iopsavgtime=int
2828 Average the calculated IOPS over the given time. Value is specified in
2829 milliseconds. If the job also does IOPS logging through
2830 :option:`write_iops_log`, then the minimum of this option and
2831 :option:`log_avg_msec` will be used. Default: 500ms.
2833 .. option:: disk_util=bool
2835 Generate disk utilization statistics, if the platform supports it.
2838 .. option:: disable_lat=bool
2840 Disable measurements of total latency numbers. Useful only for cutting back
2841 the number of calls to :manpage:`gettimeofday(2)`, as that does impact
2842 performance at really high IOPS rates. Note that to really get rid of a
2843 large amount of these calls, this option must be used with
2844 :option:`disable_slat` and :option:`disable_bw_measurement` as well.
2846 .. option:: disable_clat=bool
2848 Disable measurements of completion latency numbers. See
2849 :option:`disable_lat`.
2851 .. option:: disable_slat=bool
2853 Disable measurements of submission latency numbers. See
2854 :option:`disable_lat`.
2856 .. option:: disable_bw_measurement=bool, disable_bw=bool
2858 Disable measurements of throughput/bandwidth numbers. See
2859 :option:`disable_lat`.
2861 .. option:: clat_percentiles=bool
2863 Enable the reporting of percentiles of completion latencies.
2865 .. option:: percentile_list=float_list
2867 Overwrite the default list of percentiles for completion latencies and the
2868 block error histogram. Each number is a floating number in the range
2869 (0,100], and the maximum length of the list is 20. Use ``:`` to separate the
2870 numbers, and list the numbers in ascending order. For example,
2871 ``--percentile_list=99.5:99.9`` will cause fio to report the values of
2872 completion latency below which 99.5% and 99.9% of the observed latencies
2879 .. option:: exitall_on_error
2881 When one job finishes in error, terminate the rest. The default is to wait
2882 for each job to finish.
2884 .. option:: continue_on_error=str
2886 Normally fio will exit the job on the first observed failure. If this option
2887 is set, fio will continue the job when there is a 'non-fatal error' (EIO or
2888 EILSEQ) until the runtime is exceeded or the I/O size specified is
2889 completed. If this option is used, there are two more stats that are
2890 appended, the total error count and the first error. The error field given
2891 in the stats is the first error that was hit during the run.
2893 The allowed values are:
2896 Exit on any I/O or verify errors.
2899 Continue on read errors, exit on all others.
2902 Continue on write errors, exit on all others.
2905 Continue on any I/O error, exit on all others.
2908 Continue on verify errors, exit on all others.
2911 Continue on all errors.
2914 Backward-compatible alias for 'none'.
2917 Backward-compatible alias for 'all'.
2919 .. option:: ignore_error=str
2921 Sometimes you want to ignore some errors during test in that case you can
2922 specify error list for each error type, instead of only being able to
2923 ignore the default 'non-fatal error' using :option:`continue_on_error`.
2924 ``ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST`` errors for
2925 given error type is separated with ':'. Error may be symbol ('ENOSPC',
2926 'ENOMEM') or integer. Example::
2928 ignore_error=EAGAIN,ENOSPC:122
2930 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from
2931 WRITE. This option works by overriding :option:`continue_on_error` with
2932 the list of errors for each error type if any.
2934 .. option:: error_dump=bool
2936 If set dump every error even if it is non fatal, true by default. If
2937 disabled only fatal error will be dumped.
2939 Running predefined workloads
2940 ----------------------------
2942 Fio includes predefined profiles that mimic the I/O workloads generated by
2945 .. option:: profile=str
2947 The predefined workload to run. Current profiles are:
2950 Threaded I/O bench (tiotest/tiobench) like workload.
2953 Aerospike Certification Tool (ACT) like workload.
2955 To view a profile's additional options use :option:`--cmdhelp` after specifying
2956 the profile. For example::
2958 $ fio --profile=act --cmdhelp
2963 .. option:: device-names=str
2968 .. option:: load=int
2971 ACT load multiplier. Default: 1.
2973 .. option:: test-duration=time
2976 How long the entire test takes to run. When the unit is omitted, the value
2977 is given in seconds. Default: 24h.
2979 .. option:: threads-per-queue=int
2982 Number of read I/O threads per device. Default: 8.
2984 .. option:: read-req-num-512-blocks=int
2987 Number of 512B blocks to read at the time. Default: 3.
2989 .. option:: large-block-op-kbytes=int
2992 Size of large block ops in KiB (writes). Default: 131072.
2997 Set to run ACT prep phase.
2999 Tiobench profile options
3000 ~~~~~~~~~~~~~~~~~~~~~~~~
3002 .. option:: size=str
3007 .. option:: block=int
3010 Block size in bytes. Default: 4096.
3012 .. option:: numruns=int
3022 .. option:: threads=int
3027 Interpreting the output
3028 -----------------------
3031 Example output was based on the following:
3032 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --time_based \
3033 --rate=1256k --bs=14K --name=quick --runtime=1s --name=mixed \
3034 --runtime=2m --rw=rw
3036 Fio spits out a lot of output. While running, fio will display the status of the
3037 jobs created. An example of that would be::
3039 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]
3041 The characters inside the first set of square brackets denote the current status of
3042 each thread. The first character is the first job defined in the job file, and so
3043 forth. The possible values (in typical life cycle order) are:
3045 +------+-----+-----------------------------------------------------------+
3047 +======+=====+===========================================================+
3048 | P | | Thread setup, but not started. |
3049 +------+-----+-----------------------------------------------------------+
3050 | C | | Thread created. |
3051 +------+-----+-----------------------------------------------------------+
3052 | I | | Thread initialized, waiting or generating necessary data. |
3053 +------+-----+-----------------------------------------------------------+
3054 | | p | Thread running pre-reading file(s). |
3055 +------+-----+-----------------------------------------------------------+
3056 | | / | Thread is in ramp period. |
3057 +------+-----+-----------------------------------------------------------+
3058 | | R | Running, doing sequential reads. |
3059 +------+-----+-----------------------------------------------------------+
3060 | | r | Running, doing random reads. |
3061 +------+-----+-----------------------------------------------------------+
3062 | | W | Running, doing sequential writes. |
3063 +------+-----+-----------------------------------------------------------+
3064 | | w | Running, doing random writes. |
3065 +------+-----+-----------------------------------------------------------+
3066 | | M | Running, doing mixed sequential reads/writes. |
3067 +------+-----+-----------------------------------------------------------+
3068 | | m | Running, doing mixed random reads/writes. |
3069 +------+-----+-----------------------------------------------------------+
3070 | | D | Running, doing sequential trims. |
3071 +------+-----+-----------------------------------------------------------+
3072 | | d | Running, doing random trims. |
3073 +------+-----+-----------------------------------------------------------+
3074 | | F | Running, currently waiting for :manpage:`fsync(2)`. |
3075 +------+-----+-----------------------------------------------------------+
3076 | | V | Running, doing verification of written data. |
3077 +------+-----+-----------------------------------------------------------+
3078 | f | | Thread finishing. |
3079 +------+-----+-----------------------------------------------------------+
3080 | E | | Thread exited, not reaped by main thread yet. |
3081 +------+-----+-----------------------------------------------------------+
3082 | _ | | Thread reaped. |
3083 +------+-----+-----------------------------------------------------------+
3084 | X | | Thread reaped, exited with an error. |
3085 +------+-----+-----------------------------------------------------------+
3086 | K | | Thread reaped, exited due to signal. |
3087 +------+-----+-----------------------------------------------------------+
3090 Example output was based on the following:
3091 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --runtime=58m \
3092 --time_based --rate=2512k --bs=256K --numjobs=10 \
3093 --name=readers --rw=read --name=writers --rw=write
3095 Fio will condense the thread string as not to take up more space on the command
3096 line than needed. For instance, if you have 10 readers and 10 writers running,
3097 the output would look like this::
3099 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]
3101 Note that the status string is displayed in order, so it's possible to tell which of
3102 the jobs are currently doing what. In the example above this means that jobs 1--10
3103 are readers and 11--20 are writers.
3105 The other values are fairly self explanatory -- number of threads currently
3106 running and doing I/O, the number of currently open files (f=), the estimated
3107 completion percentage, the rate of I/O since last check (read speed listed first,
3108 then write speed and optionally trim speed) in terms of bandwidth and IOPS,
3109 and time to completion for the current running group. It's impossible to estimate
3110 runtime of the following groups (if any).
3113 Example output was based on the following:
3114 TZ=UTC fio --iodepth=16 --ioengine=posixaio --filename=/tmp/fiofile \
3115 --direct=1 --size=100M --time_based --runtime=50s --rate_iops=89 \
3116 --bs=7K --name=Client1 --rw=write
3118 When fio is done (or interrupted by :kbd:`Ctrl-C`), it will show the data for
3119 each thread, group of threads, and disks in that order. For each overall thread (or
3120 group) the output looks like::
3122 Client1: (groupid=0, jobs=1): err= 0: pid=16109: Sat Jun 24 12:07:54 2017
3123 write: IOPS=88, BW=623KiB/s (638kB/s)(30.4MiB/50032msec)
3124 slat (nsec): min=500, max=145500, avg=8318.00, stdev=4781.50
3125 clat (usec): min=170, max=78367, avg=4019.02, stdev=8293.31
3126 lat (usec): min=174, max=78375, avg=4027.34, stdev=8291.79
3127 clat percentiles (usec):
3128 | 1.00th=[ 302], 5.00th=[ 326], 10.00th=[ 343], 20.00th=[ 363],
3129 | 30.00th=[ 392], 40.00th=[ 404], 50.00th=[ 416], 60.00th=[ 445],
3130 | 70.00th=[ 816], 80.00th=[ 6718], 90.00th=[12911], 95.00th=[21627],
3131 | 99.00th=[43779], 99.50th=[51643], 99.90th=[68682], 99.95th=[72877],
3133 bw ( KiB/s): min= 532, max= 686, per=0.10%, avg=622.87, stdev=24.82, samples= 100
3134 iops : min= 76, max= 98, avg=88.98, stdev= 3.54, samples= 100
3135 lat (usec) : 250=0.04%, 500=64.11%, 750=4.81%, 1000=2.79%
3136 lat (msec) : 2=4.16%, 4=1.84%, 10=4.90%, 20=11.33%, 50=5.37%
3137 lat (msec) : 100=0.65%
3138 cpu : usr=0.27%, sys=0.18%, ctx=12072, majf=0, minf=21
3139 IO depths : 1=85.0%, 2=13.1%, 4=1.8%, 8=0.1%, 16=0.0%, 32=0.0%, >=64=0.0%
3140 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
3141 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
3142 issued rwt: total=0,4450,0, short=0,0,0, dropped=0,0,0
3143 latency : target=0, window=0, percentile=100.00%, depth=8
3145 The job name (or first job's name when using :option:`group_reporting`) is printed,
3146 along with the group id, count of jobs being aggregated, last error id seen (which
3147 is 0 when there are no errors), pid/tid of that thread and the time the job/group
3148 completed. Below are the I/O statistics for each data direction performed (showing
3149 writes in the example above). In the order listed, they denote:
3152 The string before the colon shows the I/O direction the statistics
3153 are for. **IOPS** is the average I/Os performed per second. **BW**
3154 is the average bandwidth rate shown as: value in power of 2 format
3155 (value in power of 10 format). The last two values show: (**total
3156 I/O performed** in power of 2 format / **runtime** of that thread).
3159 Submission latency (**min** being the minimum, **max** being the
3160 maximum, **avg** being the average, **stdev** being the standard
3161 deviation). This is the time it took to submit the I/O. For
3162 sync I/O this row is not displayed as the slat is really the
3163 completion latency (since queue/complete is one operation there).
3164 This value can be in nanoseconds, microseconds or milliseconds ---
3165 fio will choose the most appropriate base and print that (in the
3166 example above nanoseconds was the best scale). Note: in :option:`--minimal` mode
3167 latencies are always expressed in microseconds.
3170 Completion latency. Same names as slat, this denotes the time from
3171 submission to completion of the I/O pieces. For sync I/O, clat will
3172 usually be equal (or very close) to 0, as the time from submit to
3173 complete is basically just CPU time (I/O has already been done, see slat
3177 Total latency. Same names as slat and clat, this denotes the time from
3178 when fio created the I/O unit to completion of the I/O operation.
3181 Bandwidth statistics based on samples. Same names as the xlat stats,
3182 but also includes the number of samples taken (**samples**) and an
3183 approximate percentage of total aggregate bandwidth this thread
3184 received in its group (**per**). This last value is only really
3185 useful if the threads in this group are on the same disk, since they
3186 are then competing for disk access.
3189 IOPS statistics based on samples. Same names as bw.
3191 **lat (nsec/usec/msec)**
3192 The distribution of I/O completion latencies. This is the time from when
3193 I/O leaves fio and when it gets completed. Unlike the separate
3194 read/write/trim sections above, the data here and in the remaining
3195 sections apply to all I/Os for the reporting group. 250=0.04% means that
3196 0.04% of the I/Os completed in under 250us. 500=64.11% means that 64.11%
3197 of the I/Os required 250 to 499us for completion.
3200 CPU usage. User and system time, along with the number of context
3201 switches this thread went through, usage of system and user time, and
3202 finally the number of major and minor page faults. The CPU utilization
3203 numbers are averages for the jobs in that reporting group, while the
3204 context and fault counters are summed.
3207 The distribution of I/O depths over the job lifetime. The numbers are
3208 divided into powers of 2 and each entry covers depths from that value
3209 up to those that are lower than the next entry -- e.g., 16= covers
3210 depths from 16 to 31. Note that the range covered by a depth
3211 distribution entry can be different to the range covered by the
3212 equivalent submit/complete distribution entry.
3215 How many pieces of I/O were submitting in a single submit call. Each
3216 entry denotes that amount and below, until the previous entry -- e.g.,
3217 16=100% means that we submitted anywhere between 9 to 16 I/Os per submit
3218 call. Note that the range covered by a submit distribution entry can
3219 be different to the range covered by the equivalent depth distribution
3223 Like the above submit number, but for completions instead.
3226 The number of read/write/trim requests issued, and how many of them were
3230 These values are for `--latency-target` and related options. When
3231 these options are engaged, this section describes the I/O depth required
3232 to meet the specified latency target.
3235 Example output was based on the following:
3236 TZ=UTC fio --ioengine=null --iodepth=2 --size=100M --numjobs=2 \
3237 --rate_process=poisson --io_limit=32M --name=read --bs=128k \
3238 --rate=11M --name=write --rw=write --bs=2k --rate=700k
3240 After each client has been listed, the group statistics are printed. They
3241 will look like this::
3243 Run status group 0 (all jobs):
3244 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
3245 WRITE: bw=1231KiB/s (1261kB/s), 616KiB/s-621KiB/s (630kB/s-636kB/s), io=64.0MiB (67.1MB), run=52747-53223msec
3247 For each data direction it prints:
3250 Aggregate bandwidth of threads in this group followed by the
3251 minimum and maximum bandwidth of all the threads in this group.
3252 Values outside of brackets are power-of-2 format and those
3253 within are the equivalent value in a power-of-10 format.
3255 Aggregate I/O performed of all threads in this group. The
3256 format is the same as bw.
3258 The smallest and longest runtimes of the threads in this group.
3260 And finally, the disk statistics are printed. This is Linux specific. They will look like this::
3262 Disk stats (read/write):
3263 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
3265 Each value is printed for both reads and writes, with reads first. The
3269 Number of I/Os performed by all groups.
3271 Number of merges performed by the I/O scheduler.
3273 Number of ticks we kept the disk busy.
3275 Total time spent in the disk queue.
3277 The disk utilization. A value of 100% means we kept the disk
3278 busy constantly, 50% would be a disk idling half of the time.
3280 It is also possible to get fio to dump the current output while it is running,
3281 without terminating the job. To do that, send fio the **USR1** signal. You can
3282 also get regularly timed dumps by using the :option:`--status-interval`
3283 parameter, or by creating a file in :file:`/tmp` named
3284 :file:`fio-dump-status`. If fio sees this file, it will unlink it and dump the
3285 current output status.
3291 For scripted usage where you typically want to generate tables or graphs of the
3292 results, fio can output the results in a semicolon separated format. The format
3293 is one long line of values, such as::
3295 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%
3296 A description of this job goes here.
3298 The job description (if provided) follows on a second line.
3300 To enable terse output, use the :option:`--minimal` or
3301 :option:`--output-format`\=terse command line options. The
3302 first value is the version of the terse output format. If the output has to be
3303 changed for some reason, this number will be incremented by 1 to signify that
3306 Split up, the format is as follows (comments in brackets denote when a
3307 field was introduced or whether it's specific to some terse version):
3311 terse version, fio version [v3], jobname, groupid, error
3315 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3316 Submission latency: min, max, mean, stdev (usec)
3317 Completion latency: min, max, mean, stdev (usec)
3318 Completion latency percentiles: 20 fields (see below)
3319 Total latency: min, max, mean, stdev (usec)
3320 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
3321 IOPS [v5]: min, max, mean, stdev, number of samples
3327 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3328 Submission latency: min, max, mean, stdev (usec)
3329 Completion latency: min, max, mean, stdev (usec)
3330 Completion latency percentiles: 20 fields (see below)
3331 Total latency: min, max, mean, stdev (usec)
3332 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
3333 IOPS [v5]: min, max, mean, stdev, number of samples
3335 TRIM status [all but version 3]:
3337 Fields are similar to READ/WRITE status.
3341 user, system, context switches, major faults, minor faults
3345 <=1, 2, 4, 8, 16, 32, >=64
3347 I/O latencies microseconds::
3349 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
3351 I/O latencies milliseconds::
3353 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
3355 Disk utilization [v3]::
3357 disk name, read ios, write ios, read merges, write merges, read ticks, write ticks,
3358 time spent in queue, disk utilization percentage
3360 Additional Info (dependent on continue_on_error, default off)::
3362 total # errors, first error code
3364 Additional Info (dependent on description being set)::
3368 Completion latency percentiles can be a grouping of up to 20 sets, so for the
3369 terse output fio writes all of them. Each field will look like this::
3373 which is the Xth percentile, and the `usec` latency associated with it.
3375 For `Disk utilization`, all disks used by fio are shown. So for each disk there
3376 will be a disk utilization section.
3378 Below is a single line containing short names for each of the fields in the
3379 minimal output v3, separated by semicolons::
3381 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
3387 The `json+` output format is identical to the `json` output format except that it
3388 adds a full dump of the completion latency bins. Each `bins` object contains a
3389 set of (key, value) pairs where keys are latency durations and values count how
3390 many I/Os had completion latencies of the corresponding duration. For example,
3393 "bins" : { "87552" : 1, "89600" : 1, "94720" : 1, "96768" : 1, "97792" : 1, "99840" : 1, "100864" : 2, "103936" : 6, "104960" : 534, "105984" : 5995, "107008" : 7529, ... }
3395 This data indicates that one I/O required 87,552ns to complete, two I/Os required
3396 100,864ns to complete, and 7529 I/Os required 107,008ns to complete.
3398 Also included with fio is a Python script `fio_jsonplus_clat2csv` that takes
3399 json+ output and generates CSV-formatted latency data suitable for plotting.
3401 The latency durations actually represent the midpoints of latency intervals.
3402 For details refer to :file:`stat.h`.
3408 There are two trace file format that you can encounter. The older (v1) format is
3409 unsupported since version 1.20-rc3 (March 2008). It will still be described
3410 below in case that you get an old trace and want to understand it.
3412 In any case the trace is a simple text file with a single action per line.
3415 Trace file format v1
3416 ~~~~~~~~~~~~~~~~~~~~
3418 Each line represents a single I/O action in the following format::
3422 where `rw=0/1` for read/write, and the `offset` and `length` entries being in bytes.
3424 This format is not supported in fio versions >= 1.20-rc3.
3427 Trace file format v2
3428 ~~~~~~~~~~~~~~~~~~~~
3430 The second version of the trace file format was added in fio version 1.17. It
3431 allows to access more then one file per trace and has a bigger set of possible
3434 The first line of the trace file has to be::
3438 Following this can be lines in two different formats, which are described below.
3440 The file management format::
3444 The `filename` is given as an absolute path. The `action` can be one of these:
3447 Add the given `filename` to the trace.
3449 Open the file with the given `filename`. The `filename` has to have
3450 been added with the **add** action before.
3452 Close the file with the given `filename`. The file has to have been
3456 The file I/O action format::
3458 filename action offset length
3460 The `filename` is given as an absolute path, and has to have been added and
3461 opened before it can be used with this format. The `offset` and `length` are
3462 given in bytes. The `action` can be one of these:
3465 Wait for `offset` microseconds. Everything below 100 is discarded.
3466 The time is relative to the previous `wait` statement.
3468 Read `length` bytes beginning from `offset`.
3470 Write `length` bytes beginning from `offset`.
3472 :manpage:`fsync(2)` the file.
3474 :manpage:`fdatasync(2)` the file.
3476 Trim the given file from the given `offset` for `length` bytes.
3478 CPU idleness profiling
3479 ----------------------
3481 In some cases, we want to understand CPU overhead in a test. For example, we
3482 test patches for the specific goodness of whether they reduce CPU usage.
3483 Fio implements a balloon approach to create a thread per CPU that runs at idle
3484 priority, meaning that it only runs when nobody else needs the cpu.
3485 By measuring the amount of work completed by the thread, idleness of each CPU
3486 can be derived accordingly.
3488 An unit work is defined as touching a full page of unsigned characters. Mean and
3489 standard deviation of time to complete an unit work is reported in "unit work"
3490 section. Options can be chosen to report detailed percpu idleness or overall
3491 system idleness by aggregating percpu stats.
3494 Verification and triggers
3495 -------------------------
3497 Fio is usually run in one of two ways, when data verification is done. The first
3498 is a normal write job of some sort with verify enabled. When the write phase has
3499 completed, fio switches to reads and verifies everything it wrote. The second
3500 model is running just the write phase, and then later on running the same job
3501 (but with reads instead of writes) to repeat the same I/O patterns and verify
3502 the contents. Both of these methods depend on the write phase being completed,
3503 as fio otherwise has no idea how much data was written.
3505 With verification triggers, fio supports dumping the current write state to
3506 local files. Then a subsequent read verify workload can load this state and know
3507 exactly where to stop. This is useful for testing cases where power is cut to a
3508 server in a managed fashion, for instance.
3510 A verification trigger consists of two things:
3512 1) Storing the write state of each job.
3513 2) Executing a trigger command.
3515 The write state is relatively small, on the order of hundreds of bytes to single
3516 kilobytes. It contains information on the number of completions done, the last X
3519 A trigger is invoked either through creation ('touch') of a specified file in
3520 the system, or through a timeout setting. If fio is run with
3521 :option:`--trigger-file`\= :file:`/tmp/trigger-file`, then it will continually
3522 check for the existence of :file:`/tmp/trigger-file`. When it sees this file, it
3523 will fire off the trigger (thus saving state, and executing the trigger
3526 For client/server runs, there's both a local and remote trigger. If fio is
3527 running as a server backend, it will send the job states back to the client for
3528 safe storage, then execute the remote trigger, if specified. If a local trigger
3529 is specified, the server will still send back the write state, but the client
3530 will then execute the trigger.
3532 Verification trigger example
3533 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3535 Let's say we want to run a powercut test on the remote Linux machine 'server'.
3536 Our write workload is in :file:`write-test.fio`. We want to cut power to 'server' at
3537 some point during the run, and we'll run this test from the safety or our local
3538 machine, 'localbox'. On the server, we'll start the fio backend normally::
3540 server# fio --server
3542 and on the client, we'll fire off the workload::
3544 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger-remote="bash -c \"echo b > /proc/sysrq-triger\""
3546 We set :file:`/tmp/my-trigger` as the trigger file, and we tell fio to execute::
3548 echo b > /proc/sysrq-trigger
3550 on the server once it has received the trigger and sent us the write state. This
3551 will work, but it's not **really** cutting power to the server, it's merely
3552 abruptly rebooting it. If we have a remote way of cutting power to the server
3553 through IPMI or similar, we could do that through a local trigger command
3554 instead. Let's assume we have a script that does IPMI reboot of a given hostname,
3555 ipmi-reboot. On localbox, we could then have run fio with a local trigger
3558 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger="ipmi-reboot server"
3560 For this case, fio would wait for the server to send us the write state, then
3561 execute ``ipmi-reboot server`` when that happened.
3563 Loading verify state
3564 ~~~~~~~~~~~~~~~~~~~~
3566 To load stored write state, a read verification job file must contain the
3567 :option:`verify_state_load` option. If that is set, fio will load the previously
3568 stored state. For a local fio run this is done by loading the files directly,
3569 and on a client/server run, the server backend will ask the client to send the
3570 files over and load them from there.
3576 Fio supports a variety of log file formats, for logging latencies, bandwidth,
3577 and IOPS. The logs share a common format, which looks like this:
3579 *time* (`msec`), *value*, *data direction*, *block size* (`bytes`),
3582 *Time* for the log entry is always in milliseconds. The *value* logged depends
3583 on the type of log, it will be one of the following:
3586 Value is latency in nsecs
3592 *Data direction* is one of the following:
3601 The entry's *block size* is always in bytes. The *offset* is the offset, in bytes,
3602 from the start of the file, for that particular I/O. The logging of the offset can be
3603 toggled with :option:`log_offset`.
3605 Fio defaults to logging every individual I/O. When IOPS are logged for individual
3606 I/Os the *value* entry will always be 1. If windowed logging is enabled through
3607 :option:`log_avg_msec`, fio logs the average values over the specified period of time.
3608 If windowed logging is enabled and :option:`log_max_value` is set, then fio logs
3609 maximum values in that window instead of averages. Since *data direction*, *block
3610 size* and *offset* are per-I/O values, if windowed logging is enabled they
3611 aren't applicable and will be 0.
3616 Normally fio is invoked as a stand-alone application on the machine where the
3617 I/O workload should be generated. However, the backend and frontend of fio can
3618 be run separately i.e., the fio server can generate an I/O workload on the "Device
3619 Under Test" while being controlled by a client on another machine.
3621 Start the server on the machine which has access to the storage DUT::
3625 where `args` defines what fio listens to. The arguments are of the form
3626 ``type,hostname`` or ``IP,port``. *type* is either ``ip`` (or ip4) for TCP/IP
3627 v4, ``ip6`` for TCP/IP v6, or ``sock`` for a local unix domain socket.
3628 *hostname* is either a hostname or IP address, and *port* is the port to listen
3629 to (only valid for TCP/IP, not a local socket). Some examples:
3633 Start a fio server, listening on all interfaces on the default port (8765).
3635 2) ``fio --server=ip:hostname,4444``
3637 Start a fio server, listening on IP belonging to hostname and on port 4444.
3639 3) ``fio --server=ip6:::1,4444``
3641 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
3643 4) ``fio --server=,4444``
3645 Start a fio server, listening on all interfaces on port 4444.
3647 5) ``fio --server=1.2.3.4``
3649 Start a fio server, listening on IP 1.2.3.4 on the default port.
3651 6) ``fio --server=sock:/tmp/fio.sock``
3653 Start a fio server, listening on the local socket :file:`/tmp/fio.sock`.
3655 Once a server is running, a "client" can connect to the fio server with::
3657 fio <local-args> --client=<server> <remote-args> <job file(s)>
3659 where `local-args` are arguments for the client where it is running, `server`
3660 is the connect string, and `remote-args` and `job file(s)` are sent to the
3661 server. The `server` string follows the same format as it does on the server
3662 side, to allow IP/hostname/socket and port strings.
3664 Fio can connect to multiple servers this way::
3666 fio --client=<server1> <job file(s)> --client=<server2> <job file(s)>
3668 If the job file is located on the fio server, then you can tell the server to
3669 load a local file as well. This is done by using :option:`--remote-config` ::
3671 fio --client=server --remote-config /path/to/file.fio
3673 Then fio will open this local (to the server) job file instead of being passed
3674 one from the client.
3676 If you have many servers (example: 100 VMs/containers), you can input a pathname
3677 of a file containing host IPs/names as the parameter value for the
3678 :option:`--client` option. For example, here is an example :file:`host.list`
3679 file containing 2 hostnames::
3681 host1.your.dns.domain
3682 host2.your.dns.domain
3684 The fio command would then be::
3686 fio --client=host.list <job file(s)>
3688 In this mode, you cannot input server-specific parameters or job files -- all
3689 servers receive the same job file.
3691 In order to let ``fio --client`` runs use a shared filesystem from multiple
3692 hosts, ``fio --client`` now prepends the IP address of the server to the
3693 filename. For example, if fio is using the directory :file:`/mnt/nfs/fio` and is
3694 writing filename :file:`fileio.tmp`, with a :option:`--client` `hostfile`
3695 containing two hostnames ``h1`` and ``h2`` with IP addresses 192.168.10.120 and
3696 192.168.10.121, then fio will create two files::
3698 /mnt/nfs/fio/192.168.10.120.fileio.tmp
3699 /mnt/nfs/fio/192.168.10.121.fileio.tmp