4 The first step in getting fio to simulate a desired I/O workload, is writing a
5 job file describing that specific setup. A job file may contain any number of
6 threads and/or files -- the typical contents of the job file is a *global*
7 section defining shared parameters, and one or more job sections describing the
8 jobs involved. When run, fio parses this file and sets everything up as
9 described. If we break down a job from top to bottom, it contains the following
14 Defines the I/O pattern issued to the file(s). We may only be reading
15 sequentially from this file(s), or we may be writing randomly. Or even
16 mixing reads and writes, sequentially or randomly.
17 Should we be doing buffered I/O, or direct/raw I/O?
21 In how large chunks are we issuing I/O? This may be a single value,
22 or it may describe a range of block sizes.
26 How much data are we going to be reading/writing.
30 How do we issue I/O? We could be memory mapping the file, we could be
31 using regular read/write, we could be using splice, async I/O, or even
36 If the I/O engine is async, how large a queuing depth do we want to
42 How many files are we spreading the workload over.
44 `Threads, processes and job synchronization`_
46 How many threads or processes should we spread this workload over.
48 The above are the basic parameters defined for a workload, in addition there's a
49 multitude of parameters that modify other aspects of how this job behaves.
55 .. option:: --debug=type
57 Enable verbose tracing `type` of various fio actions. May be ``all`` for all types
58 or individual types separated by a comma (e.g. ``--debug=file,mem`` will
59 enable file and memory debugging). Currently, additional logging is
63 Dump info related to processes.
65 Dump info related to file actions.
67 Dump info related to I/O queuing.
69 Dump info related to memory allocations.
71 Dump info related to blktrace setup.
73 Dump info related to I/O verification.
75 Enable all debug options.
77 Dump info related to random offset generation.
79 Dump info related to option matching and parsing.
81 Dump info related to disk utilization updates.
83 Dump info only related to job number x.
85 Dump info only related to mutex up/down ops.
87 Dump info related to profile extensions.
89 Dump info related to internal time keeping.
91 Dump info related to networking connections.
93 Dump info related to I/O rate switching.
95 Dump info related to log compress/decompress.
97 Dump info related to steadystate detection.
99 Dump info related to the helper thread.
101 Dump info related to support for zoned block devices.
103 Show available debug options.
105 .. option:: --parse-only
107 Parse options only, don't start any I/O.
109 .. option:: --merge-blktrace-only
111 Merge blktraces only, don't start any I/O.
113 .. option:: --output=filename
115 Write output to file `filename`.
117 .. option:: --output-format=format
119 Set the reporting `format` to `normal`, `terse`, `json`, or `json+`. Multiple
120 formats can be selected, separated by a comma. `terse` is a CSV based
121 format. `json+` is like `json`, except it adds a full dump of the latency
124 .. option:: --bandwidth-log
126 Generate aggregate bandwidth logs.
128 .. option:: --minimal
130 Print statistics in a terse, semicolon-delimited format.
132 .. option:: --append-terse
134 Print statistics in selected mode AND terse, semicolon-delimited format.
135 **Deprecated**, use :option:`--output-format` instead to select multiple
138 .. option:: --terse-version=version
140 Set terse `version` output format (default 3, or 2 or 4 or 5).
142 .. option:: --version
144 Print version information and exit.
148 Print a summary of the command line options and exit.
150 .. option:: --cpuclock-test
152 Perform test and validation of internal CPU clock.
154 .. option:: --crctest=[test]
156 Test the speed of the built-in checksumming functions. If no argument is
157 given, all of them are tested. Alternatively, a comma separated list can
158 be passed, in which case the given ones are tested.
160 .. option:: --cmdhelp=command
162 Print help information for `command`. May be ``all`` for all commands.
164 .. option:: --enghelp=[ioengine[,command]]
166 List all commands defined by `ioengine`, or print help for `command`
167 defined by `ioengine`. If no `ioengine` is given, list all
170 .. option:: --showcmd=jobfile
172 Convert `jobfile` to a set of command-line options.
174 .. option:: --readonly
176 Turn on safety read-only checks, preventing writes and trims. The
177 ``--readonly`` option is an extra safety guard to prevent users from
178 accidentally starting a write or trim workload when that is not desired.
179 Fio will only modify the device under test if
180 `rw=write/randwrite/rw/randrw/trim/randtrim/trimwrite` is given. This
181 safety net can be used as an extra precaution.
183 .. option:: --eta=when
185 Specifies when real-time ETA estimate should be printed. `when` may be
186 `always`, `never` or `auto`. `auto` is the default, it prints ETA
187 when requested if the output is a TTY. `always` disregards the output
188 type, and prints ETA when requested. `never` never prints ETA.
190 .. option:: --eta-interval=time
192 By default, fio requests client ETA status roughly every second. With
193 this option, the interval is configurable. Fio imposes a minimum
194 allowed time to avoid flooding the console, less than 250 msec is
197 .. option:: --eta-newline=time
199 Force a new line for every `time` period passed. When the unit is omitted,
200 the value is interpreted in seconds.
202 .. option:: --status-interval=time
204 Force a full status dump of cumulative (from job start) values at `time`
205 intervals. This option does *not* provide per-period measurements. So
206 values such as bandwidth are running averages. When the time unit is omitted,
207 `time` is interpreted in seconds. Note that using this option with
208 ``--output-format=json`` will yield output that technically isn't valid
209 json, since the output will be collated sets of valid json. It will need
210 to be split into valid sets of json after the run.
212 .. option:: --section=name
214 Only run specified section `name` in job file. Multiple sections can be specified.
215 The ``--section`` option allows one to combine related jobs into one file.
216 E.g. one job file could define light, moderate, and heavy sections. Tell
217 fio to run only the "heavy" section by giving ``--section=heavy``
218 command line option. One can also specify the "write" operations in one
219 section and "verify" operation in another section. The ``--section`` option
220 only applies to job sections. The reserved *global* section is always
223 .. option:: --alloc-size=kb
225 Allocate additional internal smalloc pools of size `kb` in KiB. The
226 ``--alloc-size`` option increases shared memory set aside for use by fio.
227 If running large jobs with randommap enabled, fio can run out of memory.
228 Smalloc is an internal allocator for shared structures from a fixed size
229 memory pool and can grow to 16 pools. The pool size defaults to 16MiB.
231 NOTE: While running :file:`.fio_smalloc.*` backing store files are visible
234 .. option:: --warnings-fatal
236 All fio parser warnings are fatal, causing fio to exit with an
239 .. option:: --max-jobs=nr
241 Set the maximum number of threads/processes to support to `nr`.
242 NOTE: On Linux, it may be necessary to increase the shared-memory
243 limit (:file:`/proc/sys/kernel/shmmax`) if fio runs into errors while
246 .. option:: --server=args
248 Start a backend server, with `args` specifying what to listen to.
249 See `Client/Server`_ section.
251 .. option:: --daemonize=pidfile
253 Background a fio server, writing the pid to the given `pidfile` file.
255 .. option:: --client=hostname
257 Instead of running the jobs locally, send and run them on the given `hostname`
258 or set of `hostname`\s. See `Client/Server`_ section.
260 .. option:: --remote-config=file
262 Tell fio server to load this local `file`.
264 .. option:: --idle-prof=option
266 Report CPU idleness. `option` is one of the following:
269 Run unit work calibration only and exit.
272 Show aggregate system idleness and unit work.
275 As **system** but also show per CPU idleness.
277 .. option:: --inflate-log=log
279 Inflate and output compressed `log`.
281 .. option:: --trigger-file=file
283 Execute trigger command when `file` exists.
285 .. option:: --trigger-timeout=time
287 Execute trigger at this `time`.
289 .. option:: --trigger=command
291 Set this `command` as local trigger.
293 .. option:: --trigger-remote=command
295 Set this `command` as remote trigger.
297 .. option:: --aux-path=path
299 Use the directory specified by `path` for generated state files instead
300 of the current working directory.
302 Any parameters following the options will be assumed to be job files, unless
303 they match a job file parameter. Multiple job files can be listed and each job
304 file will be regarded as a separate group. Fio will :option:`stonewall`
305 execution between each group.
311 As previously described, fio accepts one or more job files describing what it is
312 supposed to do. The job file format is the classic ini file, where the names
313 enclosed in [] brackets define the job name. You are free to use any ASCII name
314 you want, except *global* which has special meaning. Following the job name is
315 a sequence of zero or more parameters, one per line, that define the behavior of
316 the job. If the first character in a line is a ';' or a '#', the entire line is
317 discarded as a comment.
319 A *global* section sets defaults for the jobs described in that file. A job may
320 override a *global* section parameter, and a job file may even have several
321 *global* sections if so desired. A job is only affected by a *global* section
324 The :option:`--cmdhelp` option also lists all options. If used with a `command`
325 argument, :option:`--cmdhelp` will detail the given `command`.
327 See the `examples/` directory for inspiration on how to write job files. Note
328 the copyright and license requirements currently apply to `examples/` files.
330 So let's look at a really simple job file that defines two processes, each
331 randomly reading from a 128MiB file:
335 ; -- start job file --
346 As you can see, the job file sections themselves are empty as all the described
347 parameters are shared. As no :option:`filename` option is given, fio makes up a
348 `filename` for each of the jobs as it sees fit. On the command line, this job
349 would look as follows::
351 $ fio --name=global --rw=randread --size=128m --name=job1 --name=job2
354 Let's look at an example that has a number of processes writing randomly to
359 ; -- start job file --
370 Here we have no *global* section, as we only have one job defined anyway. We
371 want to use async I/O here, with a depth of 4 for each file. We also increased
372 the buffer size used to 32KiB and define numjobs to 4 to fork 4 identical
373 jobs. The result is 4 processes each randomly writing to their own 64MiB
374 file. Instead of using the above job file, you could have given the parameters
375 on the command line. For this case, you would specify::
377 $ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4
379 When fio is utilized as a basis of any reasonably large test suite, it might be
380 desirable to share a set of standardized settings across multiple job files.
381 Instead of copy/pasting such settings, any section may pull in an external
382 :file:`filename.fio` file with *include filename* directive, as in the following
385 ; -- start job file including.fio --
389 include glob-include.fio
396 include test-include.fio
397 ; -- end job file including.fio --
401 ; -- start job file glob-include.fio --
404 ; -- end job file glob-include.fio --
408 ; -- start job file test-include.fio --
411 ; -- end job file test-include.fio --
413 Settings pulled into a section apply to that section only (except *global*
414 section). Include directives may be nested in that any included file may contain
415 further include directive(s). Include files may not contain [] sections.
418 Environment variables
419 ~~~~~~~~~~~~~~~~~~~~~
421 Fio also supports environment variable expansion in job files. Any sub-string of
422 the form ``${VARNAME}`` as part of an option value (in other words, on the right
423 of the '='), will be expanded to the value of the environment variable called
424 `VARNAME`. If no such environment variable is defined, or `VARNAME` is the
425 empty string, the empty string will be substituted.
427 As an example, let's look at a sample fio invocation and job file::
429 $ SIZE=64m NUMJOBS=4 fio jobfile.fio
433 ; -- start job file --
440 This will expand to the following equivalent job file at runtime:
444 ; -- start job file --
451 Fio ships with a few example job files, you can also look there for inspiration.
456 Additionally, fio has a set of reserved keywords that will be replaced
457 internally with the appropriate value. Those keywords are:
461 The architecture page size of the running system.
465 Megabytes of total memory in the system.
469 Number of online available CPUs.
471 These can be used on the command line or in the job file, and will be
472 automatically substituted with the current system values when the job is
473 run. Simple math is also supported on these keywords, so you can perform actions
478 and get that properly expanded to 8 times the size of memory in the machine.
484 This section describes in details each parameter associated with a job. Some
485 parameters take an option of a given type, such as an integer or a
486 string. Anywhere a numeric value is required, an arithmetic expression may be
487 used, provided it is surrounded by parentheses. Supported operators are:
496 For time values in expressions, units are microseconds by default. This is
497 different than for time values not in expressions (not enclosed in
498 parentheses). The following types are used:
505 String: A sequence of alphanumeric characters.
508 Integer with possible time suffix. Without a unit value is interpreted as
509 seconds unless otherwise specified. Accepts a suffix of 'd' for days, 'h' for
510 hours, 'm' for minutes, 's' for seconds, 'ms' (or 'msec') for milliseconds and
511 'us' (or 'usec') for microseconds. For example, use 10m for 10 minutes.
516 Integer. A whole number value, which may contain an integer prefix
517 and an integer suffix:
519 [*integer prefix*] **number** [*integer suffix*]
521 The optional *integer prefix* specifies the number's base. The default
522 is decimal. *0x* specifies hexadecimal.
524 The optional *integer suffix* specifies the number's units, and includes an
525 optional unit prefix and an optional unit. For quantities of data, the
526 default unit is bytes. For quantities of time, the default unit is seconds
527 unless otherwise specified.
529 With :option:`kb_base`\=1000, fio follows international standards for unit
530 prefixes. To specify power-of-10 decimal values defined in the
531 International System of Units (SI):
533 * *K* -- means kilo (K) or 1000
534 * *M* -- means mega (M) or 1000**2
535 * *G* -- means giga (G) or 1000**3
536 * *T* -- means tera (T) or 1000**4
537 * *P* -- means peta (P) or 1000**5
539 To specify power-of-2 binary values defined in IEC 80000-13:
541 * *Ki* -- means kibi (Ki) or 1024
542 * *Mi* -- means mebi (Mi) or 1024**2
543 * *Gi* -- means gibi (Gi) or 1024**3
544 * *Ti* -- means tebi (Ti) or 1024**4
545 * *Pi* -- means pebi (Pi) or 1024**5
547 With :option:`kb_base`\=1024 (the default), the unit prefixes are opposite
548 from those specified in the SI and IEC 80000-13 standards to provide
549 compatibility with old scripts. For example, 4k means 4096.
551 For quantities of data, an optional unit of 'B' may be included
552 (e.g., 'kB' is the same as 'k').
554 The *integer suffix* is not case sensitive (e.g., m/mi mean mebi/mega,
555 not milli). 'b' and 'B' both mean byte, not bit.
557 Examples with :option:`kb_base`\=1000:
559 * *4 KiB*: 4096, 4096b, 4096B, 4ki, 4kib, 4kiB, 4Ki, 4KiB
560 * *1 MiB*: 1048576, 1mi, 1024ki
561 * *1 MB*: 1000000, 1m, 1000k
562 * *1 TiB*: 1099511627776, 1ti, 1024gi, 1048576mi
563 * *1 TB*: 1000000000, 1t, 1000m, 1000000k
565 Examples with :option:`kb_base`\=1024 (default):
567 * *4 KiB*: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB
568 * *1 MiB*: 1048576, 1m, 1024k
569 * *1 MB*: 1000000, 1mi, 1000ki
570 * *1 TiB*: 1099511627776, 1t, 1024g, 1048576m
571 * *1 TB*: 1000000000, 1ti, 1000mi, 1000000ki
573 To specify times (units are not case sensitive):
577 * *M* -- means minutes
578 * *s* -- or sec means seconds (default)
579 * *ms* -- or *msec* means milliseconds
580 * *us* -- or *usec* means microseconds
582 If the option accepts an upper and lower range, use a colon ':' or
583 minus '-' to separate such values. See :ref:`irange <irange>`.
584 If the lower value specified happens to be larger than the upper value
585 the two values are swapped.
590 Boolean. Usually parsed as an integer, however only defined for
591 true and false (1 and 0).
596 Integer range with suffix. Allows value range to be given, such as
597 1024-4096. A colon may also be used as the separator, e.g. 1k:4k. If the
598 option allows two sets of ranges, they can be specified with a ',' or '/'
599 delimiter: 1k-4k/8k-32k. Also see :ref:`int <int>`.
602 A list of floating point numbers, separated by a ':' character.
604 With the above in mind, here follows the complete list of fio job parameters.
610 .. option:: kb_base=int
612 Select the interpretation of unit prefixes in input parameters.
615 Inputs comply with IEC 80000-13 and the International
616 System of Units (SI). Use:
618 - power-of-2 values with IEC prefixes (e.g., KiB)
619 - power-of-10 values with SI prefixes (e.g., kB)
622 Compatibility mode (default). To avoid breaking old scripts:
624 - power-of-2 values with SI prefixes
625 - power-of-10 values with IEC prefixes
627 See :option:`bs` for more details on input parameters.
629 Outputs always use correct prefixes. Most outputs include both
632 bw=2383.3kB/s (2327.4KiB/s)
634 If only one value is reported, then kb_base selects the one to use:
636 **1000** -- SI prefixes
638 **1024** -- IEC prefixes
640 .. option:: unit_base=int
642 Base unit for reporting. Allowed values are:
645 Use auto-detection (default).
657 ASCII name of the job. This may be used to override the name printed by fio
658 for this job. Otherwise the job name is used. On the command line this
659 parameter has the special purpose of also signaling the start of a new job.
661 .. option:: description=str
663 Text description of the job. Doesn't do anything except dump this text
664 description when this job is run. It's not parsed.
666 .. option:: loops=int
668 Run the specified number of iterations of this job. Used to repeat the same
669 workload a given number of times. Defaults to 1.
671 .. option:: numjobs=int
673 Create the specified number of clones of this job. Each clone of job
674 is spawned as an independent thread or process. May be used to setup a
675 larger number of threads/processes doing the same thing. Each thread is
676 reported separately; to see statistics for all clones as a whole, use
677 :option:`group_reporting` in conjunction with :option:`new_group`.
678 See :option:`--max-jobs`. Default: 1.
681 Time related parameters
682 ~~~~~~~~~~~~~~~~~~~~~~~
684 .. option:: runtime=time
686 Tell fio to terminate processing after the specified period of time. It
687 can be quite hard to determine for how long a specified job will run, so
688 this parameter is handy to cap the total runtime to a given time. When
689 the unit is omitted, the value is interpreted in seconds.
691 .. option:: time_based
693 If set, fio will run for the duration of the :option:`runtime` specified
694 even if the file(s) are completely read or written. It will simply loop over
695 the same workload as many times as the :option:`runtime` allows.
697 .. option:: startdelay=irange(time)
699 Delay the start of job for the specified amount of time. Can be a single
700 value or a range. When given as a range, each thread will choose a value
701 randomly from within the range. Value is in seconds if a unit is omitted.
703 .. option:: ramp_time=time
705 If set, fio will run the specified workload for this amount of time before
706 logging any performance numbers. Useful for letting performance settle
707 before logging results, thus minimizing the runtime required for stable
708 results. Note that the ``ramp_time`` is considered lead in time for a job,
709 thus it will increase the total runtime if a special timeout or
710 :option:`runtime` is specified. When the unit is omitted, the value is
713 .. option:: clocksource=str
715 Use the given clocksource as the base of timing. The supported options are:
718 :manpage:`gettimeofday(2)`
721 :manpage:`clock_gettime(2)`
724 Internal CPU clock source
726 cpu is the preferred clocksource if it is reliable, as it is very fast (and
727 fio is heavy on time calls). Fio will automatically use this clocksource if
728 it's supported and considered reliable on the system it is running on,
729 unless another clocksource is specifically set. For x86/x86-64 CPUs, this
730 means supporting TSC Invariant.
732 .. option:: gtod_reduce=bool
734 Enable all of the :manpage:`gettimeofday(2)` reducing options
735 (:option:`disable_clat`, :option:`disable_slat`, :option:`disable_bw_measurement`) plus
736 reduce precision of the timeout somewhat to really shrink the
737 :manpage:`gettimeofday(2)` call count. With this option enabled, we only do
738 about 0.4% of the :manpage:`gettimeofday(2)` calls we would have done if all
739 time keeping was enabled.
741 .. option:: gtod_cpu=int
743 Sometimes it's cheaper to dedicate a single thread of execution to just
744 getting the current time. Fio (and databases, for instance) are very
745 intensive on :manpage:`gettimeofday(2)` calls. With this option, you can set
746 one CPU aside for doing nothing but logging current time to a shared memory
747 location. Then the other threads/processes that run I/O workloads need only
748 copy that segment, instead of entering the kernel with a
749 :manpage:`gettimeofday(2)` call. The CPU set aside for doing these time
750 calls will be excluded from other uses. Fio will manually clear it from the
751 CPU mask of other jobs.
757 .. option:: directory=str
759 Prefix filenames with this directory. Used to place files in a different
760 location than :file:`./`. You can specify a number of directories by
761 separating the names with a ':' character. These directories will be
762 assigned equally distributed to job clones created by :option:`numjobs` as
763 long as they are using generated filenames. If specific `filename(s)` are
764 set fio will use the first listed directory, and thereby matching the
765 `filename` semantic (which generates a file for each clone if not
766 specified, but lets all clones use the same file if set).
768 See the :option:`filename` option for information on how to escape "``:``"
769 characters within the directory path itself.
771 Note: To control the directory fio will use for internal state files
772 use :option:`--aux-path`.
774 .. option:: filename=str
776 Fio normally makes up a `filename` based on the job name, thread number, and
777 file number (see :option:`filename_format`). If you want to share files
778 between threads in a job or several
779 jobs with fixed file paths, specify a `filename` for each of them to override
780 the default. If the ioengine is file based, you can specify a number of files
781 by separating the names with a ':' colon. So if you wanted a job to open
782 :file:`/dev/sda` and :file:`/dev/sdb` as the two working files, you would use
783 ``filename=/dev/sda:/dev/sdb``. This also means that whenever this option is
784 specified, :option:`nrfiles` is ignored. The size of regular files specified
785 by this option will be :option:`size` divided by number of files unless an
786 explicit size is specified by :option:`filesize`.
788 Each colon in the wanted path must be escaped with a ``\``
789 character. For instance, if the path is :file:`/dev/dsk/foo@3,0:c` then you
790 would use ``filename=/dev/dsk/foo@3,0\:c`` and if the path is
791 :file:`F:\\filename` then you would use ``filename=F\:\filename``.
793 On Windows, disk devices are accessed as :file:`\\\\.\\PhysicalDrive0` for
794 the first device, :file:`\\\\.\\PhysicalDrive1` for the second etc.
795 Note: Windows and FreeBSD prevent write access to areas
796 of the disk containing in-use data (e.g. filesystems).
798 The filename "`-`" is a reserved name, meaning *stdin* or *stdout*. Which
799 of the two depends on the read/write direction set.
801 .. option:: filename_format=str
803 If sharing multiple files between jobs, it is usually necessary to have fio
804 generate the exact names that you want. By default, fio will name a file
805 based on the default file format specification of
806 :file:`jobname.jobnumber.filenumber`. With this option, that can be
807 customized. Fio will recognize and replace the following keywords in this
811 The name of the worker thread or process.
813 IP of the fio process when using client/server mode.
815 The incremental number of the worker thread or process.
817 The incremental number of the file for that worker thread or
820 To have dependent jobs share a set of files, this option can be set to have
821 fio generate filenames that are shared between the two. For instance, if
822 :file:`testfiles.$filenum` is specified, file number 4 for any job will be
823 named :file:`testfiles.4`. The default of :file:`$jobname.$jobnum.$filenum`
824 will be used if no other format specifier is given.
826 If you specify a path then the directories will be created up to the
827 main directory for the file. So for example if you specify
828 ``filename_format=a/b/c/$jobnum`` then the directories a/b/c will be
829 created before the file setup part of the job. If you specify
830 :option:`directory` then the path will be relative that directory,
831 otherwise it is treated as the absolute path.
833 .. option:: unique_filename=bool
835 To avoid collisions between networked clients, fio defaults to prefixing any
836 generated filenames (with a directory specified) with the source of the
837 client connecting. To disable this behavior, set this option to 0.
839 .. option:: opendir=str
841 Recursively open any files below directory `str`.
843 .. option:: lockfile=str
845 Fio defaults to not locking any files before it does I/O to them. If a file
846 or file descriptor is shared, fio can serialize I/O to that file to make the
847 end result consistent. This is usual for emulating real workloads that share
848 files. The lock modes are:
851 No locking. The default.
853 Only one thread or process may do I/O at a time, excluding all
856 Read-write locking on the file. Many readers may
857 access the file at the same time, but writes get exclusive access.
859 .. option:: nrfiles=int
861 Number of files to use for this job. Defaults to 1. The size of files
862 will be :option:`size` divided by this unless explicit size is specified by
863 :option:`filesize`. Files are created for each thread separately, and each
864 file will have a file number within its name by default, as explained in
865 :option:`filename` section.
868 .. option:: openfiles=int
870 Number of files to keep open at the same time. Defaults to the same as
871 :option:`nrfiles`, can be set smaller to limit the number simultaneous
874 .. option:: file_service_type=str
876 Defines how fio decides which file from a job to service next. The following
880 Choose a file at random.
883 Round robin over opened files. This is the default.
886 Finish one file before moving on to the next. Multiple files can
887 still be open depending on :option:`openfiles`.
890 Use a *Zipf* distribution to decide what file to access.
893 Use a *Pareto* distribution to decide what file to access.
896 Use a *Gaussian* (normal) distribution to decide what file to
902 For *random*, *roundrobin*, and *sequential*, a postfix can be appended to
903 tell fio how many I/Os to issue before switching to a new file. For example,
904 specifying ``file_service_type=random:8`` would cause fio to issue
905 8 I/Os before selecting a new file at random. For the non-uniform
906 distributions, a floating point postfix can be given to influence how the
907 distribution is skewed. See :option:`random_distribution` for a description
908 of how that would work.
910 .. option:: ioscheduler=str
912 Attempt to switch the device hosting the file to the specified I/O scheduler
915 .. option:: create_serialize=bool
917 If true, serialize the file creation for the jobs. This may be handy to
918 avoid interleaving of data files, which may greatly depend on the filesystem
919 used and even the number of processors in the system. Default: true.
921 .. option:: create_fsync=bool
923 :manpage:`fsync(2)` the data file after creation. This is the default.
925 .. option:: create_on_open=bool
927 If true, don't pre-create files but allow the job's open() to create a file
928 when it's time to do I/O. Default: false -- pre-create all necessary files
931 .. option:: create_only=bool
933 If true, fio will only run the setup phase of the job. If files need to be
934 laid out or updated on disk, only that will be done -- the actual job contents
935 are not executed. Default: false.
937 .. option:: allow_file_create=bool
939 If true, fio is permitted to create files as part of its workload. If this
940 option is false, then fio will error out if
941 the files it needs to use don't already exist. Default: true.
943 .. option:: allow_mounted_write=bool
945 If this isn't set, fio will abort jobs that are destructive (e.g. that write)
946 to what appears to be a mounted device or partition. This should help catch
947 creating inadvertently destructive tests, not realizing that the test will
948 destroy data on the mounted file system. Note that some platforms don't allow
949 writing against a mounted device regardless of this option. Default: false.
951 .. option:: pre_read=bool
953 If this is given, files will be pre-read into memory before starting the
954 given I/O operation. This will also clear the :option:`invalidate` flag,
955 since it is pointless to pre-read and then drop the cache. This will only
956 work for I/O engines that are seek-able, since they allow you to read the
957 same data multiple times. Thus it will not work on non-seekable I/O engines
958 (e.g. network, splice). Default: false.
960 .. option:: unlink=bool
962 Unlink the job files when done. Not the default, as repeated runs of that
963 job would then waste time recreating the file set again and again. Default:
966 .. option:: unlink_each_loop=bool
968 Unlink job files after each iteration or loop. Default: false.
970 .. option:: zonemode=str
975 The :option:`zonerange`, :option:`zonesize`,
976 :option `zonecapacity` and option:`zoneskip`
977 parameters are ignored.
979 I/O happens in a single zone until
980 :option:`zonesize` bytes have been transferred.
981 After that number of bytes has been
982 transferred processing of the next zone
983 starts. :option `zonecapacity` is ignored.
985 Zoned block device mode. I/O happens
986 sequentially in each zone, even if random I/O
987 has been selected. Random I/O happens across
988 all zones instead of being restricted to a
989 single zone. The :option:`zoneskip` parameter
990 is ignored. :option:`zonerange` and
991 :option:`zonesize` must be identical.
993 .. option:: zonerange=int
995 Size of a single zone. See also :option:`zonesize` and
998 .. option:: zonesize=int
1000 For :option:`zonemode` =strided, this is the number of bytes to
1001 transfer before skipping :option:`zoneskip` bytes. If this parameter
1002 is smaller than :option:`zonerange` then only a fraction of each zone
1003 with :option:`zonerange` bytes will be accessed. If this parameter is
1004 larger than :option:`zonerange` then each zone will be accessed
1005 multiple times before skipping to the next zone.
1007 For :option:`zonemode` =zbd, this is the size of a single zone. The
1008 :option:`zonerange` parameter is ignored in this mode.
1011 .. option:: zonecapacity=int
1013 For :option:`zonemode` =zbd, this defines the capacity of a single zone,
1014 which is the accessible area starting from the zone start address.
1015 This parameter only applies when using :option:`zonemode` =zbd in
1016 combination with regular block devices. If not specified it defaults to
1017 the zone size. If the target device is a zoned block device, the zone
1018 capacity is obtained from the device information and this option is
1021 .. option:: zoneskip=int
1023 For :option:`zonemode` =strided, the number of bytes to skip after
1024 :option:`zonesize` bytes of data have been transferred. This parameter
1025 must be zero for :option:`zonemode` =zbd.
1027 .. option:: read_beyond_wp=bool
1029 This parameter applies to :option:`zonemode` =zbd only.
1031 Zoned block devices are block devices that consist of multiple zones.
1032 Each zone has a type, e.g. conventional or sequential. A conventional
1033 zone can be written at any offset that is a multiple of the block
1034 size. Sequential zones must be written sequentially. The position at
1035 which a write must occur is called the write pointer. A zoned block
1036 device can be either drive managed, host managed or host aware. For
1037 host managed devices the host must ensure that writes happen
1038 sequentially. Fio recognizes host managed devices and serializes
1039 writes to sequential zones for these devices.
1041 If a read occurs in a sequential zone beyond the write pointer then
1042 the zoned block device will complete the read without reading any data
1043 from the storage medium. Since such reads lead to unrealistically high
1044 bandwidth and IOPS numbers fio only reads beyond the write pointer if
1045 explicitly told to do so. Default: false.
1047 .. option:: max_open_zones=int
1049 When running a random write test across an entire drive many more
1050 zones will be open than in a typical application workload. Hence this
1051 command line option that allows to limit the number of open zones. The
1052 number of open zones is defined as the number of zones to which write
1053 commands are issued.
1055 .. option:: zone_reset_threshold=float
1057 A number between zero and one that indicates the ratio of logical
1058 blocks with data to the total number of logical blocks in the test
1059 above which zones should be reset periodically.
1061 .. option:: zone_reset_frequency=float
1063 A number between zero and one that indicates how often a zone reset
1064 should be issued if the zone reset threshold has been exceeded. A zone
1065 reset is submitted after each (1 / zone_reset_frequency) write
1066 requests. This and the previous parameter can be used to simulate
1067 garbage collection activity.
1073 .. option:: direct=bool
1075 If value is true, use non-buffered I/O. This is usually O_DIRECT. Note that
1076 OpenBSD and ZFS on Solaris don't support direct I/O. On Windows the synchronous
1077 ioengines don't support direct I/O. Default: false.
1079 .. option:: atomic=bool
1081 If value is true, attempt to use atomic direct I/O. Atomic writes are
1082 guaranteed to be stable once acknowledged by the operating system. Only
1083 Linux supports O_ATOMIC right now.
1085 .. option:: buffered=bool
1087 If value is true, use buffered I/O. This is the opposite of the
1088 :option:`direct` option. Defaults to true.
1090 .. option:: readwrite=str, rw=str
1092 Type of I/O pattern. Accepted values are:
1099 Sequential trims (Linux block devices and SCSI
1100 character devices only).
1106 Random trims (Linux block devices and SCSI
1107 character devices only).
1109 Sequential mixed reads and writes.
1111 Random mixed reads and writes.
1113 Sequential trim+write sequences. Blocks will be trimmed first,
1114 then the same blocks will be written to.
1116 Fio defaults to read if the option is not specified. For the mixed I/O
1117 types, the default is to split them 50/50. For certain types of I/O the
1118 result may still be skewed a bit, since the speed may be different.
1120 It is possible to specify the number of I/Os to do before getting a new
1121 offset by appending ``:<nr>`` to the end of the string given. For a
1122 random read, it would look like ``rw=randread:8`` for passing in an offset
1123 modifier with a value of 8. If the suffix is used with a sequential I/O
1124 pattern, then the *<nr>* value specified will be **added** to the generated
1125 offset for each I/O turning sequential I/O into sequential I/O with holes.
1126 For instance, using ``rw=write:4k`` will skip 4k for every write. Also see
1127 the :option:`rw_sequencer` option.
1129 .. option:: rw_sequencer=str
1131 If an offset modifier is given by appending a number to the ``rw=<str>``
1132 line, then this option controls how that number modifies the I/O offset
1133 being generated. Accepted values are:
1136 Generate sequential offset.
1138 Generate the same offset.
1140 ``sequential`` is only useful for random I/O, where fio would normally
1141 generate a new random offset for every I/O. If you append e.g. 8 to randread,
1142 you would get a new random offset for every 8 I/Os. The result would be a
1143 seek for only every 8 I/Os, instead of for every I/O. Use ``rw=randread:8``
1144 to specify that. As sequential I/O is already sequential, setting
1145 ``sequential`` for that would not result in any differences. ``identical``
1146 behaves in a similar fashion, except it sends the same offset 8 number of
1147 times before generating a new offset.
1149 .. option:: unified_rw_reporting=str
1151 Fio normally reports statistics on a per data direction basis, meaning that
1152 reads, writes, and trims are accounted and reported separately. This option
1153 determines whether fio reports the results normally, summed together, or as
1155 Accepted values are:
1158 Normal statistics reporting.
1161 Statistics are summed per data direction and reported together.
1164 Statistics are reported normally, followed by the mixed statistics.
1167 Backward-compatible alias for **none**.
1170 Backward-compatible alias for **mixed**.
1175 .. option:: randrepeat=bool
1177 Seed the random number generator used for random I/O patterns in a
1178 predictable way so the pattern is repeatable across runs. Default: true.
1180 .. option:: allrandrepeat=bool
1182 Seed all random number generators in a predictable way so results are
1183 repeatable across runs. Default: false.
1185 .. option:: randseed=int
1187 Seed the random number generators based on this seed value, to be able to
1188 control what sequence of output is being generated. If not set, the random
1189 sequence depends on the :option:`randrepeat` setting.
1191 .. option:: fallocate=str
1193 Whether pre-allocation is performed when laying down files.
1194 Accepted values are:
1197 Do not pre-allocate space.
1200 Use a platform's native pre-allocation call but fall back to
1201 **none** behavior if it fails/is not implemented.
1204 Pre-allocate via :manpage:`posix_fallocate(3)`.
1207 Pre-allocate via :manpage:`fallocate(2)` with
1208 FALLOC_FL_KEEP_SIZE set.
1211 Extend file to final size via :manpage:`ftruncate(2)`
1212 instead of allocating.
1215 Backward-compatible alias for **none**.
1218 Backward-compatible alias for **posix**.
1220 May not be available on all supported platforms. **keep** is only available
1221 on Linux. If using ZFS on Solaris this cannot be set to **posix**
1222 because ZFS doesn't support pre-allocation. Default: **native** if any
1223 pre-allocation methods except **truncate** are available, **none** if not.
1225 Note that using **truncate** on Windows will interact surprisingly
1226 with non-sequential write patterns. When writing to a file that has
1227 been extended by setting the end-of-file information, Windows will
1228 backfill the unwritten portion of the file up to that offset with
1229 zeroes before issuing the new write. This means that a single small
1230 write to the end of an extended file will stall until the entire
1231 file has been filled with zeroes.
1233 .. option:: fadvise_hint=str
1235 Use :manpage:`posix_fadvise(2)` or :manpage:`posix_fadvise(2)` to
1236 advise the kernel on what I/O patterns are likely to be issued.
1237 Accepted values are:
1240 Backwards-compatible hint for "no hint".
1243 Backwards compatible hint for "advise with fio workload type". This
1244 uses **FADV_RANDOM** for a random workload, and **FADV_SEQUENTIAL**
1245 for a sequential workload.
1248 Advise using **FADV_SEQUENTIAL**.
1251 Advise using **FADV_RANDOM**.
1253 .. option:: write_hint=str
1255 Use :manpage:`fcntl(2)` to advise the kernel what life time to expect
1256 from a write. Only supported on Linux, as of version 4.13. Accepted
1260 No particular life time associated with this file.
1263 Data written to this file has a short life time.
1266 Data written to this file has a medium life time.
1269 Data written to this file has a long life time.
1272 Data written to this file has a very long life time.
1274 The values are all relative to each other, and no absolute meaning
1275 should be associated with them.
1277 .. option:: offset=int
1279 Start I/O at the provided offset in the file, given as either a fixed size in
1280 bytes or a percentage. If a percentage is given, the generated offset will be
1281 aligned to the minimum ``blocksize`` or to the value of ``offset_align`` if
1282 provided. Data before the given offset will not be touched. This
1283 effectively caps the file size at `real_size - offset`. Can be combined with
1284 :option:`size` to constrain the start and end range of the I/O workload.
1285 A percentage can be specified by a number between 1 and 100 followed by '%',
1286 for example, ``offset=20%`` to specify 20%.
1288 .. option:: offset_align=int
1290 If set to non-zero value, the byte offset generated by a percentage ``offset``
1291 is aligned upwards to this value. Defaults to 0 meaning that a percentage
1292 offset is aligned to the minimum block size.
1294 .. option:: offset_increment=int
1296 If this is provided, then the real offset becomes `offset + offset_increment
1297 * thread_number`, where the thread number is a counter that starts at 0 and
1298 is incremented for each sub-job (i.e. when :option:`numjobs` option is
1299 specified). This option is useful if there are several jobs which are
1300 intended to operate on a file in parallel disjoint segments, with even
1301 spacing between the starting points. Percentages can be used for this option.
1302 If a percentage is given, the generated offset will be aligned to the minimum
1303 ``blocksize`` or to the value of ``offset_align`` if provided.
1305 .. option:: number_ios=int
1307 Fio will normally perform I/Os until it has exhausted the size of the region
1308 set by :option:`size`, or if it exhaust the allocated time (or hits an error
1309 condition). With this setting, the range/size can be set independently of
1310 the number of I/Os to perform. When fio reaches this number, it will exit
1311 normally and report status. Note that this does not extend the amount of I/O
1312 that will be done, it will only stop fio if this condition is met before
1313 other end-of-job criteria.
1315 .. option:: fsync=int
1317 If writing to a file, issue an :manpage:`fsync(2)` (or its equivalent) of
1318 the dirty data for every number of blocks given. For example, if you give 32
1319 as a parameter, fio will sync the file after every 32 writes issued. If fio is
1320 using non-buffered I/O, we may not sync the file. The exception is the sg
1321 I/O engine, which synchronizes the disk cache anyway. Defaults to 0, which
1322 means fio does not periodically issue and wait for a sync to complete. Also
1323 see :option:`end_fsync` and :option:`fsync_on_close`.
1325 .. option:: fdatasync=int
1327 Like :option:`fsync` but uses :manpage:`fdatasync(2)` to only sync data and
1328 not metadata blocks. In Windows, FreeBSD, DragonFlyBSD or OSX there is no
1329 :manpage:`fdatasync(2)` so this falls back to using :manpage:`fsync(2)`.
1330 Defaults to 0, which means fio does not periodically issue and wait for a
1331 data-only sync to complete.
1333 .. option:: write_barrier=int
1335 Make every `N-th` write a barrier write.
1337 .. option:: sync_file_range=str:int
1339 Use :manpage:`sync_file_range(2)` for every `int` number of write
1340 operations. Fio will track range of writes that have happened since the last
1341 :manpage:`sync_file_range(2)` call. `str` can currently be one or more of:
1344 SYNC_FILE_RANGE_WAIT_BEFORE
1346 SYNC_FILE_RANGE_WRITE
1348 SYNC_FILE_RANGE_WAIT_AFTER
1350 So if you do ``sync_file_range=wait_before,write:8``, fio would use
1351 ``SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE`` for every 8
1352 writes. Also see the :manpage:`sync_file_range(2)` man page. This option is
1355 .. option:: overwrite=bool
1357 If true, writes to a file will always overwrite existing data. If the file
1358 doesn't already exist, it will be created before the write phase begins. If
1359 the file exists and is large enough for the specified write phase, nothing
1360 will be done. Default: false.
1362 .. option:: end_fsync=bool
1364 If true, :manpage:`fsync(2)` file contents when a write stage has completed.
1367 .. option:: fsync_on_close=bool
1369 If true, fio will :manpage:`fsync(2)` a dirty file on close. This differs
1370 from :option:`end_fsync` in that it will happen on every file close, not
1371 just at the end of the job. Default: false.
1373 .. option:: rwmixread=int
1375 Percentage of a mixed workload that should be reads. Default: 50.
1377 .. option:: rwmixwrite=int
1379 Percentage of a mixed workload that should be writes. If both
1380 :option:`rwmixread` and :option:`rwmixwrite` is given and the values do not
1381 add up to 100%, the latter of the two will be used to override the
1382 first. This may interfere with a given rate setting, if fio is asked to
1383 limit reads or writes to a certain rate. If that is the case, then the
1384 distribution may be skewed. Default: 50.
1386 .. option:: random_distribution=str:float[:float][,str:float][,str:float]
1388 By default, fio will use a completely uniform random distribution when asked
1389 to perform random I/O. Sometimes it is useful to skew the distribution in
1390 specific ways, ensuring that some parts of the data is more hot than others.
1391 fio includes the following distribution models:
1394 Uniform random distribution
1403 Normal (Gaussian) distribution
1406 Zoned random distribution
1409 Zone absolute random distribution
1411 When using a **zipf** or **pareto** distribution, an input value is also
1412 needed to define the access pattern. For **zipf**, this is the `Zipf
1413 theta`. For **pareto**, it's the `Pareto power`. Fio includes a test
1414 program, :command:`fio-genzipf`, that can be used visualize what the given input
1415 values will yield in terms of hit rates. If you wanted to use **zipf** with
1416 a `theta` of 1.2, you would use ``random_distribution=zipf:1.2`` as the
1417 option. If a non-uniform model is used, fio will disable use of the random
1418 map. For the **normal** distribution, a normal (Gaussian) deviation is
1419 supplied as a value between 0 and 100.
1421 The second, optional float is allowed for **pareto**, **zipf** and **normal** distributions.
1422 It allows to set base of distribution in non-default place, giving more control
1423 over most probable outcome. This value is in range [0-1] which maps linearly to
1424 range of possible random values.
1425 Defaults are: random for **pareto** and **zipf**, and 0.5 for **normal**.
1426 If you wanted to use **zipf** with a `theta` of 1.2 centered on 1/4 of allowed value range,
1427 you would use ``random_distibution=zipf:1.2:0.25``.
1429 For a **zoned** distribution, fio supports specifying percentages of I/O
1430 access that should fall within what range of the file or device. For
1431 example, given a criteria of:
1433 * 60% of accesses should be to the first 10%
1434 * 30% of accesses should be to the next 20%
1435 * 8% of accesses should be to the next 30%
1436 * 2% of accesses should be to the next 40%
1438 we can define that through zoning of the random accesses. For the above
1439 example, the user would do::
1441 random_distribution=zoned:60/10:30/20:8/30:2/40
1443 A **zoned_abs** distribution works exactly like the **zoned**, except
1444 that it takes absolute sizes. For example, let's say you wanted to
1445 define access according to the following criteria:
1447 * 60% of accesses should be to the first 20G
1448 * 30% of accesses should be to the next 100G
1449 * 10% of accesses should be to the next 500G
1451 we can define an absolute zoning distribution with:
1453 random_distribution=zoned_abs=60/20G:30/100G:10/500g
1455 For both **zoned** and **zoned_abs**, fio supports defining up to
1458 Similarly to how :option:`bssplit` works for setting ranges and
1459 percentages of block sizes. Like :option:`bssplit`, it's possible to
1460 specify separate zones for reads, writes, and trims. If just one set
1461 is given, it'll apply to all of them. This goes for both **zoned**
1462 **zoned_abs** distributions.
1464 .. option:: percentage_random=int[,int][,int]
1466 For a random workload, set how big a percentage should be random. This
1467 defaults to 100%, in which case the workload is fully random. It can be set
1468 from anywhere from 0 to 100. Setting it to 0 would make the workload fully
1469 sequential. Any setting in between will result in a random mix of sequential
1470 and random I/O, at the given percentages. Comma-separated values may be
1471 specified for reads, writes, and trims as described in :option:`blocksize`.
1473 .. option:: norandommap
1475 Normally fio will cover every block of the file when doing random I/O. If
1476 this option is given, fio will just get a new random offset without looking
1477 at past I/O history. This means that some blocks may not be read or written,
1478 and that some blocks may be read/written more than once. If this option is
1479 used with :option:`verify` and multiple blocksizes (via :option:`bsrange`),
1480 only intact blocks are verified, i.e., partially-overwritten blocks are
1481 ignored. With an async I/O engine and an I/O depth > 1, it is possible for
1482 the same block to be overwritten, which can cause verification errors. Either
1483 do not use norandommap in this case, or also use the lfsr random generator.
1485 .. option:: softrandommap=bool
1487 See :option:`norandommap`. If fio runs with the random block map enabled and
1488 it fails to allocate the map, if this option is set it will continue without
1489 a random block map. As coverage will not be as complete as with random maps,
1490 this option is disabled by default.
1492 .. option:: random_generator=str
1494 Fio supports the following engines for generating I/O offsets for random I/O:
1497 Strong 2^88 cycle random number generator.
1499 Linear feedback shift register generator.
1501 Strong 64-bit 2^258 cycle random number generator.
1503 **tausworthe** is a strong random number generator, but it requires tracking
1504 on the side if we want to ensure that blocks are only read or written
1505 once. **lfsr** guarantees that we never generate the same offset twice, and
1506 it's also less computationally expensive. It's not a true random generator,
1507 however, though for I/O purposes it's typically good enough. **lfsr** only
1508 works with single block sizes, not with workloads that use multiple block
1509 sizes. If used with such a workload, fio may read or write some blocks
1510 multiple times. The default value is **tausworthe**, unless the required
1511 space exceeds 2^32 blocks. If it does, then **tausworthe64** is
1512 selected automatically.
1518 .. option:: blocksize=int[,int][,int], bs=int[,int][,int]
1520 The block size in bytes used for I/O units. Default: 4096. A single value
1521 applies to reads, writes, and trims. Comma-separated values may be
1522 specified for reads, writes, and trims. A value not terminated in a comma
1523 applies to subsequent types.
1528 means 256k for reads, writes and trims.
1531 means 8k for reads, 32k for writes and trims.
1534 means 8k for reads, 32k for writes, and default for trims.
1537 means default for reads, 8k for writes and trims.
1540 means default for reads, 8k for writes, and default for trims.
1542 .. option:: blocksize_range=irange[,irange][,irange], bsrange=irange[,irange][,irange]
1544 A range of block sizes in bytes for I/O units. The issued I/O unit will
1545 always be a multiple of the minimum size, unless
1546 :option:`blocksize_unaligned` is set.
1548 Comma-separated ranges may be specified for reads, writes, and trims as
1549 described in :option:`blocksize`.
1551 Example: ``bsrange=1k-4k,2k-8k``.
1553 .. option:: bssplit=str[,str][,str]
1555 Sometimes you want even finer grained control of the block sizes
1556 issued, not just an even split between them. This option allows you to
1557 weight various block sizes, so that you are able to define a specific
1558 amount of block sizes issued. The format for this option is::
1560 bssplit=blocksize/percentage:blocksize/percentage
1562 for as many block sizes as needed. So if you want to define a workload
1563 that has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would
1566 bssplit=4k/10:64k/50:32k/40
1568 Ordering does not matter. If the percentage is left blank, fio will
1569 fill in the remaining values evenly. So a bssplit option like this one::
1571 bssplit=4k/50:1k/:32k/
1573 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always
1574 add up to 100, if bssplit is given a range that adds up to more, it
1577 Comma-separated values may be specified for reads, writes, and trims as
1578 described in :option:`blocksize`.
1580 If you want a workload that has 50% 2k reads and 50% 4k reads, while
1581 having 90% 4k writes and 10% 8k writes, you would specify::
1583 bssplit=2k/50:4k/50,4k/90:8k/10
1585 Fio supports defining up to 64 different weights for each data
1588 .. option:: blocksize_unaligned, bs_unaligned
1590 If set, fio will issue I/O units with any size within
1591 :option:`blocksize_range`, not just multiples of the minimum size. This
1592 typically won't work with direct I/O, as that normally requires sector
1595 .. option:: bs_is_seq_rand=bool
1597 If this option is set, fio will use the normal read,write blocksize settings
1598 as sequential,random blocksize settings instead. Any random read or write
1599 will use the WRITE blocksize settings, and any sequential read or write will
1600 use the READ blocksize settings.
1602 .. option:: blockalign=int[,int][,int], ba=int[,int][,int]
1604 Boundary to which fio will align random I/O units. Default:
1605 :option:`blocksize`. Minimum alignment is typically 512b for using direct
1606 I/O, though it usually depends on the hardware block size. This option is
1607 mutually exclusive with using a random map for files, so it will turn off
1608 that option. Comma-separated values may be specified for reads, writes, and
1609 trims as described in :option:`blocksize`.
1615 .. option:: zero_buffers
1617 Initialize buffers with all zeros. Default: fill buffers with random data.
1619 .. option:: refill_buffers
1621 If this option is given, fio will refill the I/O buffers on every
1622 submit. Only makes sense if :option:`zero_buffers` isn't specified,
1623 naturally. Defaults to being unset i.e., the buffer is only filled at
1624 init time and the data in it is reused when possible but if any of
1625 :option:`verify`, :option:`buffer_compress_percentage` or
1626 :option:`dedupe_percentage` are enabled then `refill_buffers` is also
1627 automatically enabled.
1629 .. option:: scramble_buffers=bool
1631 If :option:`refill_buffers` is too costly and the target is using data
1632 deduplication, then setting this option will slightly modify the I/O buffer
1633 contents to defeat normal de-dupe attempts. This is not enough to defeat
1634 more clever block compression attempts, but it will stop naive dedupe of
1635 blocks. Default: true.
1637 .. option:: buffer_compress_percentage=int
1639 If this is set, then fio will attempt to provide I/O buffer content
1640 (on WRITEs) that compresses to the specified level. Fio does this by
1641 providing a mix of random data followed by fixed pattern data. The
1642 fixed pattern is either zeros, or the pattern specified by
1643 :option:`buffer_pattern`. If the `buffer_pattern` option is used, it
1644 might skew the compression ratio slightly. Setting
1645 `buffer_compress_percentage` to a value other than 100 will also
1646 enable :option:`refill_buffers` in order to reduce the likelihood that
1647 adjacent blocks are so similar that they over compress when seen
1648 together. See :option:`buffer_compress_chunk` for how to set a finer or
1649 coarser granularity for the random/fixed data region. Defaults to unset
1650 i.e., buffer data will not adhere to any compression level.
1652 .. option:: buffer_compress_chunk=int
1654 This setting allows fio to manage how big the random/fixed data region
1655 is when using :option:`buffer_compress_percentage`. When
1656 `buffer_compress_chunk` is set to some non-zero value smaller than the
1657 block size, fio can repeat the random/fixed region throughout the I/O
1658 buffer at the specified interval (which particularly useful when
1659 bigger block sizes are used for a job). When set to 0, fio will use a
1660 chunk size that matches the block size resulting in a single
1661 random/fixed region within the I/O buffer. Defaults to 512. When the
1662 unit is omitted, the value is interpreted in bytes.
1664 .. option:: buffer_pattern=str
1666 If set, fio will fill the I/O buffers with this pattern or with the contents
1667 of a file. If not set, the contents of I/O buffers are defined by the other
1668 options related to buffer contents. The setting can be any pattern of bytes,
1669 and can be prefixed with 0x for hex values. It may also be a string, where
1670 the string must then be wrapped with ``""``. Or it may also be a filename,
1671 where the filename must be wrapped with ``''`` in which case the file is
1672 opened and read. Note that not all the file contents will be read if that
1673 would cause the buffers to overflow. So, for example::
1675 buffer_pattern='filename'
1679 buffer_pattern="abcd"
1687 buffer_pattern=0xdeadface
1689 Also you can combine everything together in any order::
1691 buffer_pattern=0xdeadface"abcd"-12'filename'
1693 .. option:: dedupe_percentage=int
1695 If set, fio will generate this percentage of identical buffers when
1696 writing. These buffers will be naturally dedupable. The contents of the
1697 buffers depend on what other buffer compression settings have been set. It's
1698 possible to have the individual buffers either fully compressible, or not at
1699 all -- this option only controls the distribution of unique buffers. Setting
1700 this option will also enable :option:`refill_buffers` to prevent every buffer
1703 .. option:: invalidate=bool
1705 Invalidate the buffer/page cache parts of the files to be used prior to
1706 starting I/O if the platform and file type support it. Defaults to true.
1707 This will be ignored if :option:`pre_read` is also specified for the
1710 .. option:: sync=str
1712 Whether, and what type, of synchronous I/O to use for writes. The allowed
1716 Do not use synchronous IO, the default.
1722 Use synchronous file IO. For the majority of I/O engines,
1723 this means using O_SYNC.
1729 Use synchronous data IO. For the majority of I/O engines,
1730 this means using O_DSYNC.
1733 .. option:: iomem=str, mem=str
1735 Fio can use various types of memory as the I/O unit buffer. The allowed
1739 Use memory from :manpage:`malloc(3)` as the buffers. Default memory
1743 Use shared memory as the buffers. Allocated through
1744 :manpage:`shmget(2)`.
1747 Same as shm, but use huge pages as backing.
1750 Use :manpage:`mmap(2)` to allocate buffers. May either be anonymous memory, or can
1751 be file backed if a filename is given after the option. The format
1752 is `mem=mmap:/path/to/file`.
1755 Use a memory mapped huge file as the buffer backing. Append filename
1756 after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file`.
1759 Same as mmap, but use a MMAP_SHARED mapping.
1762 Use GPU memory as the buffers for GPUDirect RDMA benchmark.
1763 The :option:`ioengine` must be `rdma`.
1765 The area allocated is a function of the maximum allowed bs size for the job,
1766 multiplied by the I/O depth given. Note that for **shmhuge** and
1767 **mmaphuge** to work, the system must have free huge pages allocated. This
1768 can normally be checked and set by reading/writing
1769 :file:`/proc/sys/vm/nr_hugepages` on a Linux system. Fio assumes a huge page
1770 is 4MiB in size. So to calculate the number of huge pages you need for a
1771 given job file, add up the I/O depth of all jobs (normally one unless
1772 :option:`iodepth` is used) and multiply by the maximum bs set. Then divide
1773 that number by the huge page size. You can see the size of the huge pages in
1774 :file:`/proc/meminfo`. If no huge pages are allocated by having a non-zero
1775 number in `nr_hugepages`, using **mmaphuge** or **shmhuge** will fail. Also
1776 see :option:`hugepage-size`.
1778 **mmaphuge** also needs to have hugetlbfs mounted and the file location
1779 should point there. So if it's mounted in :file:`/huge`, you would use
1780 `mem=mmaphuge:/huge/somefile`.
1782 .. option:: iomem_align=int, mem_align=int
1784 This indicates the memory alignment of the I/O memory buffers. Note that
1785 the given alignment is applied to the first I/O unit buffer, if using
1786 :option:`iodepth` the alignment of the following buffers are given by the
1787 :option:`bs` used. In other words, if using a :option:`bs` that is a
1788 multiple of the page sized in the system, all buffers will be aligned to
1789 this value. If using a :option:`bs` that is not page aligned, the alignment
1790 of subsequent I/O memory buffers is the sum of the :option:`iomem_align` and
1793 .. option:: hugepage-size=int
1795 Defines the size of a huge page. Must at least be equal to the system
1796 setting, see :file:`/proc/meminfo`. Defaults to 4MiB. Should probably
1797 always be a multiple of megabytes, so using ``hugepage-size=Xm`` is the
1798 preferred way to set this to avoid setting a non-pow-2 bad value.
1800 .. option:: lockmem=int
1802 Pin the specified amount of memory with :manpage:`mlock(2)`. Can be used to
1803 simulate a smaller amount of memory. The amount specified is per worker.
1809 .. option:: size=int
1811 The total size of file I/O for each thread of this job. Fio will run until
1812 this many bytes has been transferred, unless runtime is limited by other options
1813 (such as :option:`runtime`, for instance, or increased/decreased by :option:`io_size`).
1814 Fio will divide this size between the available files determined by options
1815 such as :option:`nrfiles`, :option:`filename`, unless :option:`filesize` is
1816 specified by the job. If the result of division happens to be 0, the size is
1817 set to the physical size of the given files or devices if they exist.
1818 If this option is not specified, fio will use the full size of the given
1819 files or devices. If the files do not exist, size must be given. It is also
1820 possible to give size as a percentage between 1 and 100. If ``size=20%`` is
1821 given, fio will use 20% of the full size of the given files or devices.
1822 Can be combined with :option:`offset` to constrain the start and end range
1823 that I/O will be done within.
1825 .. option:: io_size=int, io_limit=int
1827 Normally fio operates within the region set by :option:`size`, which means
1828 that the :option:`size` option sets both the region and size of I/O to be
1829 performed. Sometimes that is not what you want. With this option, it is
1830 possible to define just the amount of I/O that fio should do. For instance,
1831 if :option:`size` is set to 20GiB and :option:`io_size` is set to 5GiB, fio
1832 will perform I/O within the first 20GiB but exit when 5GiB have been
1833 done. The opposite is also possible -- if :option:`size` is set to 20GiB,
1834 and :option:`io_size` is set to 40GiB, then fio will do 40GiB of I/O within
1835 the 0..20GiB region.
1837 .. option:: filesize=irange(int)
1839 Individual file sizes. May be a range, in which case fio will select sizes
1840 for files at random within the given range and limited to :option:`size` in
1841 total (if that is given). If not given, each created file is the same size.
1842 This option overrides :option:`size` in terms of file size, which means
1843 this value is used as a fixed size or possible range of each file.
1845 .. option:: file_append=bool
1847 Perform I/O after the end of the file. Normally fio will operate within the
1848 size of a file. If this option is set, then fio will append to the file
1849 instead. This has identical behavior to setting :option:`offset` to the size
1850 of a file. This option is ignored on non-regular files.
1852 .. option:: fill_device=bool, fill_fs=bool
1854 Sets size to something really large and waits for ENOSPC (no space left on
1855 device) as the terminating condition. Only makes sense with sequential
1856 write. For a read workload, the mount point will be filled first then I/O
1857 started on the result. This option doesn't make sense if operating on a raw
1858 device node, since the size of that is already known by the file system.
1859 Additionally, writing beyond end-of-device will not return ENOSPC there.
1865 .. option:: ioengine=str
1867 Defines how the job issues I/O to the file. The following types are defined:
1870 Basic :manpage:`read(2)` or :manpage:`write(2)`
1871 I/O. :manpage:`lseek(2)` is used to position the I/O location.
1872 See :option:`fsync` and :option:`fdatasync` for syncing write I/Os.
1875 Basic :manpage:`pread(2)` or :manpage:`pwrite(2)` I/O. Default on
1876 all supported operating systems except for Windows.
1879 Basic :manpage:`readv(2)` or :manpage:`writev(2)` I/O. Will emulate
1880 queuing by coalescing adjacent I/Os into a single submission.
1883 Basic :manpage:`preadv(2)` or :manpage:`pwritev(2)` I/O.
1886 Basic :manpage:`preadv2(2)` or :manpage:`pwritev2(2)` I/O.
1889 Fast Linux native asynchronous I/O. Supports async IO
1890 for both direct and buffered IO.
1891 This engine defines engine specific options.
1894 Linux native asynchronous I/O. Note that Linux may only support
1895 queued behavior with non-buffered I/O (set ``direct=1`` or
1897 This engine defines engine specific options.
1900 POSIX asynchronous I/O using :manpage:`aio_read(3)` and
1901 :manpage:`aio_write(3)`.
1904 Solaris native asynchronous I/O.
1907 Windows native asynchronous I/O. Default on Windows.
1910 File is memory mapped with :manpage:`mmap(2)` and data copied
1911 to/from using :manpage:`memcpy(3)`.
1914 :manpage:`splice(2)` is used to transfer the data and
1915 :manpage:`vmsplice(2)` to transfer data from user space to the
1919 SCSI generic sg v3 I/O. May either be synchronous using the SG_IO
1920 ioctl, or if the target is an sg character device we use
1921 :manpage:`read(2)` and :manpage:`write(2)` for asynchronous
1922 I/O. Requires :option:`filename` option to specify either block or
1923 character devices. This engine supports trim operations.
1924 The sg engine includes engine specific options.
1927 Doesn't transfer any data, just pretends to. This is mainly used to
1928 exercise fio itself and for debugging/testing purposes.
1931 Transfer over the network to given ``host:port``. Depending on the
1932 :option:`protocol` used, the :option:`hostname`, :option:`port`,
1933 :option:`listen` and :option:`filename` options are used to specify
1934 what sort of connection to make, while the :option:`protocol` option
1935 determines which protocol will be used. This engine defines engine
1939 Like **net**, but uses :manpage:`splice(2)` and
1940 :manpage:`vmsplice(2)` to map data and send/receive.
1941 This engine defines engine specific options.
1944 Doesn't transfer any data, but burns CPU cycles according to the
1945 :option:`cpuload`, :option:`cpuchunks` and :option:`cpumode` options.
1946 Setting :option:`cpuload`\=85 will cause that job to do nothing but burn 85%
1947 of the CPU. In case of SMP machines, use :option:`numjobs`\=<nr_of_cpu>
1948 to get desired CPU usage, as the cpuload only loads a
1949 single CPU at the desired rate. A job never finishes unless there is
1950 at least one non-cpuio job.
1951 Setting :option:`cpumode`\=qsort replace the default noop instructions loop
1952 by a qsort algorithm to consume more energy.
1955 The RDMA I/O engine supports both RDMA memory semantics
1956 (RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the
1957 InfiniBand, RoCE and iWARP protocols. This engine defines engine
1961 I/O engine that does regular fallocate to simulate data transfer as
1965 does fallocate(,mode = FALLOC_FL_KEEP_SIZE,).
1968 does fallocate(,mode = 0).
1971 does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE).
1974 I/O engine that sends :manpage:`ftruncate(2)` operations in response
1975 to write (DDIR_WRITE) events. Each ftruncate issued sets the file's
1976 size to the current block offset. :option:`blocksize` is ignored.
1979 I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate
1980 defragment activity in request to DDIR_WRITE event.
1983 I/O engine supporting direct access to Ceph Reliable Autonomic
1984 Distributed Object Store (RADOS) via librados. This ioengine
1985 defines engine specific options.
1988 I/O engine supporting direct access to Ceph Rados Block Devices
1989 (RBD) via librbd without the need to use the kernel rbd driver. This
1990 ioengine defines engine specific options.
1993 I/O engine supporting GET/PUT requests over HTTP(S) with libcurl to
1994 a WebDAV or S3 endpoint. This ioengine defines engine specific options.
1996 This engine only supports direct IO of iodepth=1; you need to scale this
1997 via numjobs. blocksize defines the size of the objects to be created.
1999 TRIM is translated to object deletion.
2002 Using GlusterFS libgfapi sync interface to direct access to
2003 GlusterFS volumes without having to go through FUSE. This ioengine
2004 defines engine specific options.
2007 Using GlusterFS libgfapi async interface to direct access to
2008 GlusterFS volumes without having to go through FUSE. This ioengine
2009 defines engine specific options.
2012 Read and write through Hadoop (HDFS). The :option:`filename` option
2013 is used to specify host,port of the hdfs name-node to connect. This
2014 engine interprets offsets a little differently. In HDFS, files once
2015 created cannot be modified so random writes are not possible. To
2016 imitate this the libhdfs engine expects a bunch of small files to be
2017 created over HDFS and will randomly pick a file from them
2018 based on the offset generated by fio backend (see the example
2019 job file to create such files, use ``rw=write`` option). Please
2020 note, it may be necessary to set environment variables to work
2021 with HDFS/libhdfs properly. Each job uses its own connection to
2025 Read, write and erase an MTD character device (e.g.,
2026 :file:`/dev/mtd0`). Discards are treated as erases. Depending on the
2027 underlying device type, the I/O may have to go in a certain pattern,
2028 e.g., on NAND, writing sequentially to erase blocks and discarding
2029 before overwriting. The `trimwrite` mode works well for this
2033 Read and write using filesystem DAX to a file on a filesystem
2034 mounted with DAX on a persistent memory device through the PMDK
2038 Read and write using device DAX to a persistent memory device (e.g.,
2039 /dev/dax0.0) through the PMDK libpmem library.
2042 Prefix to specify loading an external I/O engine object file. Append
2043 the engine filename, e.g. ``ioengine=external:/tmp/foo.o`` to load
2044 ioengine :file:`foo.o` in :file:`/tmp`. The path can be either
2045 absolute or relative. See :file:`engines/skeleton_external.c` for
2046 details of writing an external I/O engine.
2049 Simply create the files and do no I/O to them. You still need to
2050 set `filesize` so that all the accounting still occurs, but no
2051 actual I/O will be done other than creating the file.
2054 Simply do stat() and do no I/O to the file. You need to set 'filesize'
2055 and 'nrfiles', so that files will be created.
2056 This engine is to measure file lookup and meta data access.
2059 Simply delete the files by unlink() and do no I/O to them. You need to set 'filesize'
2060 and 'nrfiles', so that the files will be created.
2061 This engine is to measure file delete.
2064 Read and write using mmap I/O to a file on a filesystem
2065 mounted with DAX on a persistent memory device through the PMDK
2069 Synchronous read and write using DDN's Infinite Memory Engine (IME).
2070 This engine is very basic and issues calls to IME whenever an IO is
2074 Synchronous read and write using DDN's Infinite Memory Engine (IME).
2075 This engine uses iovecs and will try to stack as much IOs as possible
2076 (if the IOs are "contiguous" and the IO depth is not exceeded)
2077 before issuing a call to IME.
2080 Asynchronous read and write using DDN's Infinite Memory Engine (IME).
2081 This engine will try to stack as much IOs as possible by creating
2082 requests for IME. FIO will then decide when to commit these requests.
2084 Read and write iscsi lun with libiscsi.
2086 Read and write a Network Block Device (NBD).
2089 I/O engine supporting libcufile synchronous access to nvidia-fs and a
2090 GPUDirect Storage-supported filesystem. This engine performs
2091 I/O without transferring buffers between user-space and the kernel,
2092 unless :option:`verify` is set or :option:`cuda_io` is `posix`.
2093 :option:`iomem` must not be `cudamalloc`. This ioengine defines
2094 engine specific options.
2096 I/O engine supporting asynchronous read and write operations to the
2097 DAOS File System (DFS) via libdfs.
2099 I/O engine specific parameters
2100 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2102 In addition, there are some parameters which are only valid when a specific
2103 :option:`ioengine` is in use. These are used identically to normal parameters,
2104 with the caveat that when used on the command line, they must come after the
2105 :option:`ioengine` that defines them is selected.
2107 .. option:: cmdprio_percentage=int : [io_uring] [libaio]
2109 Set the percentage of I/O that will be issued with higher priority by setting
2110 the priority bit. Non-read I/O is likely unaffected by ``cmdprio_percentage``.
2111 This option cannot be used with the `prio` or `prioclass` options. For this
2112 option to set the priority bit properly, NCQ priority must be supported and
2113 enabled and :option:`direct`\=1 option must be used. fio must also be run as
2116 .. option:: fixedbufs : [io_uring]
2118 If fio is asked to do direct IO, then Linux will map pages for each
2119 IO call, and release them when IO is done. If this option is set, the
2120 pages are pre-mapped before IO is started. This eliminates the need to
2121 map and release for each IO. This is more efficient, and reduces the
2124 .. option:: hipri : [io_uring]
2126 If this option is set, fio will attempt to use polled IO completions.
2127 Normal IO completions generate interrupts to signal the completion of
2128 IO, polled completions do not. Hence they are require active reaping
2129 by the application. The benefits are more efficient IO for high IOPS
2130 scenarios, and lower latencies for low queue depth IO.
2132 .. option:: registerfiles : [io_uring]
2134 With this option, fio registers the set of files being used with the
2135 kernel. This avoids the overhead of managing file counts in the kernel,
2136 making the submission and completion part more lightweight. Required
2137 for the below :option:`sqthread_poll` option.
2139 .. option:: sqthread_poll : [io_uring]
2141 Normally fio will submit IO by issuing a system call to notify the
2142 kernel of available items in the SQ ring. If this option is set, the
2143 act of submitting IO will be done by a polling thread in the kernel.
2144 This frees up cycles for fio, at the cost of using more CPU in the
2147 .. option:: sqthread_poll_cpu : [io_uring]
2149 When :option:`sqthread_poll` is set, this option provides a way to
2150 define which CPU should be used for the polling thread.
2152 .. option:: userspace_reap : [libaio]
2154 Normally, with the libaio engine in use, fio will use the
2155 :manpage:`io_getevents(2)` system call to reap newly returned events. With
2156 this flag turned on, the AIO ring will be read directly from user-space to
2157 reap events. The reaping mode is only enabled when polling for a minimum of
2158 0 events (e.g. when :option:`iodepth_batch_complete` `=0`).
2160 .. option:: hipri : [pvsync2]
2162 Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority
2165 .. option:: hipri_percentage : [pvsync2]
2167 When hipri is set this determines the probability of a pvsync2 I/O being high
2168 priority. The default is 100%.
2170 .. option:: nowait : [pvsync2] [libaio] [io_uring]
2172 By default if a request cannot be executed immediately (e.g. resource starvation,
2173 waiting on locks) it is queued and the initiating process will be blocked until
2174 the required resource becomes free.
2176 This option sets the RWF_NOWAIT flag (supported from the 4.14 Linux kernel) and
2177 the call will return instantly with EAGAIN or a partial result rather than waiting.
2179 It is useful to also use ignore_error=EAGAIN when using this option.
2181 Note: glibc 2.27, 2.28 have a bug in syscall wrappers preadv2, pwritev2.
2182 They return EOPNOTSUP instead of EAGAIN.
2184 For cached I/O, using this option usually means a request operates only with
2185 cached data. Currently the RWF_NOWAIT flag does not supported for cached write.
2187 For direct I/O, requests will only succeed if cache invalidation isn't required,
2188 file blocks are fully allocated and the disk request could be issued immediately.
2190 .. option:: cpuload=int : [cpuio]
2192 Attempt to use the specified percentage of CPU cycles. This is a mandatory
2193 option when using cpuio I/O engine.
2195 .. option:: cpuchunks=int : [cpuio]
2197 Split the load into cycles of the given time. In microseconds.
2199 .. option:: exit_on_io_done=bool : [cpuio]
2201 Detect when I/O threads are done, then exit.
2203 .. option:: namenode=str : [libhdfs]
2205 The hostname or IP address of a HDFS cluster namenode to contact.
2207 .. option:: port=int
2211 The listening port of the HFDS cluster namenode.
2215 The TCP or UDP port to bind to or connect to. If this is used with
2216 :option:`numjobs` to spawn multiple instances of the same job type, then
2217 this will be the starting port number since fio will use a range of
2222 The port to use for RDMA-CM communication. This should be the same value
2223 on the client and the server side.
2225 .. option:: hostname=str : [netsplice] [net] [rdma]
2227 The hostname or IP address to use for TCP, UDP or RDMA-CM based I/O. If the job
2228 is a TCP listener or UDP reader, the hostname is not used and must be omitted
2229 unless it is a valid UDP multicast address.
2231 .. option:: serverip=str : [librpma_*]
2233 The IP address to be used for RDMA-CM based I/O.
2235 .. option:: direct_write_to_pmem=bool : [librpma_*]
2237 Set to 1 only when Direct Write to PMem from the remote host is possible.
2238 Otherwise, set to 0.
2240 .. option:: busy_wait_polling=bool : [librpma_*_server]
2242 Set to 0 to wait for completion instead of busy-wait polling completion.
2245 .. option:: interface=str : [netsplice] [net]
2247 The IP address of the network interface used to send or receive UDP
2250 .. option:: ttl=int : [netsplice] [net]
2252 Time-to-live value for outgoing UDP multicast packets. Default: 1.
2254 .. option:: nodelay=bool : [netsplice] [net]
2256 Set TCP_NODELAY on TCP connections.
2258 .. option:: protocol=str, proto=str : [netsplice] [net]
2260 The network protocol to use. Accepted values are:
2263 Transmission control protocol.
2265 Transmission control protocol V6.
2267 User datagram protocol.
2269 User datagram protocol V6.
2273 When the protocol is TCP or UDP, the port must also be given, as well as the
2274 hostname if the job is a TCP listener or UDP reader. For unix sockets, the
2275 normal :option:`filename` option should be used and the port is invalid.
2277 .. option:: listen : [netsplice] [net]
2279 For TCP network connections, tell fio to listen for incoming connections
2280 rather than initiating an outgoing connection. The :option:`hostname` must
2281 be omitted if this option is used.
2283 .. option:: pingpong : [netsplice] [net]
2285 Normally a network writer will just continue writing data, and a network
2286 reader will just consume packages. If ``pingpong=1`` is set, a writer will
2287 send its normal payload to the reader, then wait for the reader to send the
2288 same payload back. This allows fio to measure network latencies. The
2289 submission and completion latencies then measure local time spent sending or
2290 receiving, and the completion latency measures how long it took for the
2291 other end to receive and send back. For UDP multicast traffic
2292 ``pingpong=1`` should only be set for a single reader when multiple readers
2293 are listening to the same address.
2295 .. option:: window_size : [netsplice] [net]
2297 Set the desired socket buffer size for the connection.
2299 .. option:: mss : [netsplice] [net]
2301 Set the TCP maximum segment size (TCP_MAXSEG).
2303 .. option:: donorname=str : [e4defrag]
2305 File will be used as a block donor (swap extents between files).
2307 .. option:: inplace=int : [e4defrag]
2309 Configure donor file blocks allocation strategy:
2312 Default. Preallocate donor's file on init.
2314 Allocate space immediately inside defragment event, and free right
2317 .. option:: clustername=str : [rbd,rados]
2319 Specifies the name of the Ceph cluster.
2321 .. option:: rbdname=str : [rbd]
2323 Specifies the name of the RBD.
2325 .. option:: pool=str : [rbd,rados]
2327 Specifies the name of the Ceph pool containing RBD or RADOS data.
2329 .. option:: clientname=str : [rbd,rados]
2331 Specifies the username (without the 'client.' prefix) used to access the
2332 Ceph cluster. If the *clustername* is specified, the *clientname* shall be
2333 the full *type.id* string. If no type. prefix is given, fio will add
2334 'client.' by default.
2336 .. option:: busy_poll=bool : [rbd,rados]
2338 Poll store instead of waiting for completion. Usually this provides better
2339 throughput at cost of higher(up to 100%) CPU utilization.
2341 .. option:: touch_objects=bool : [rados]
2343 During initialization, touch (create if do not exist) all objects (files).
2344 Touching all objects affects ceph caches and likely impacts test results.
2347 .. option:: skip_bad=bool : [mtd]
2349 Skip operations against known bad blocks.
2351 .. option:: hdfsdirectory : [libhdfs]
2353 libhdfs will create chunk in this HDFS directory.
2355 .. option:: chunk_size : [libhdfs]
2357 The size of the chunk to use for each file.
2359 .. option:: verb=str : [rdma]
2361 The RDMA verb to use on this side of the RDMA ioengine connection. Valid
2362 values are write, read, send and recv. These correspond to the equivalent
2363 RDMA verbs (e.g. write = rdma_write etc.). Note that this only needs to be
2364 specified on the client side of the connection. See the examples folder.
2366 .. option:: bindname=str : [rdma]
2368 The name to use to bind the local RDMA-CM connection to a local RDMA device.
2369 This could be a hostname or an IPv4 or IPv6 address. On the server side this
2370 will be passed into the rdma_bind_addr() function and on the client site it
2371 will be used in the rdma_resolve_add() function. This can be useful when
2372 multiple paths exist between the client and the server or in certain loopback
2375 .. option:: stat_type=str : [filestat]
2377 Specify stat system call type to measure lookup/getattr performance.
2378 Default is **stat** for :manpage:`stat(2)`.
2380 .. option:: readfua=bool : [sg]
2382 With readfua option set to 1, read operations include
2383 the force unit access (fua) flag. Default is 0.
2385 .. option:: writefua=bool : [sg]
2387 With writefua option set to 1, write operations include
2388 the force unit access (fua) flag. Default is 0.
2390 .. option:: sg_write_mode=str : [sg]
2392 Specify the type of write commands to issue. This option can take three values:
2395 This is the default where write opcodes are issued as usual.
2397 Issue WRITE AND VERIFY commands. The BYTCHK bit is set to 0. This
2398 directs the device to carry out a medium verification with no data
2399 comparison. The writefua option is ignored with this selection.
2401 Issue WRITE SAME commands. This transfers a single block to the device
2402 and writes this same block of data to a contiguous sequence of LBAs
2403 beginning at the specified offset. fio's block size parameter specifies
2404 the amount of data written with each command. However, the amount of data
2405 actually transferred to the device is equal to the device's block
2406 (sector) size. For a device with 512 byte sectors, blocksize=8k will
2407 write 16 sectors with each command. fio will still generate 8k of data
2408 for each command but only the first 512 bytes will be used and
2409 transferred to the device. The writefua option is ignored with this
2412 .. option:: hipri : [sg]
2414 If this option is set, fio will attempt to use polled IO completions.
2415 This will have a similar effect as (io_uring)hipri. Only SCSI READ and
2416 WRITE commands will have the SGV4_FLAG_HIPRI set (not UNMAP (trim) nor
2417 VERIFY). Older versions of the Linux sg driver that do not support
2418 hipri will simply ignore this flag and do normal IO. The Linux SCSI
2419 Low Level Driver (LLD) that "owns" the device also needs to support
2420 hipri (also known as iopoll and mq_poll). The MegaRAID driver is an
2421 example of a SCSI LLD. Default: clear (0) which does normal
2422 (interrupted based) IO.
2424 .. option:: http_host=str : [http]
2426 Hostname to connect to. For S3, this could be the bucket hostname.
2427 Default is **localhost**
2429 .. option:: http_user=str : [http]
2431 Username for HTTP authentication.
2433 .. option:: http_pass=str : [http]
2435 Password for HTTP authentication.
2437 .. option:: https=str : [http]
2439 Enable HTTPS instead of http. *on* enables HTTPS; *insecure*
2440 will enable HTTPS, but disable SSL peer verification (use with
2441 caution!). Default is **off**
2443 .. option:: http_mode=str : [http]
2445 Which HTTP access mode to use: *webdav*, *swift*, or *s3*.
2446 Default is **webdav**
2448 .. option:: http_s3_region=str : [http]
2450 The S3 region/zone string.
2451 Default is **us-east-1**
2453 .. option:: http_s3_key=str : [http]
2457 .. option:: http_s3_keyid=str : [http]
2459 The S3 key/access id.
2461 .. option:: http_swift_auth_token=str : [http]
2463 The Swift auth token. See the example configuration file on how
2466 .. option:: http_verbose=int : [http]
2468 Enable verbose requests from libcurl. Useful for debugging. 1
2469 turns on verbose logging from libcurl, 2 additionally enables
2470 HTTP IO tracing. Default is **0**
2472 .. option:: uri=str : [nbd]
2474 Specify the NBD URI of the server to test. The string
2475 is a standard NBD URI
2476 (see https://github.com/NetworkBlockDevice/nbd/tree/master/doc).
2477 Example URIs: nbd://localhost:10809
2478 nbd+unix:///?socket=/tmp/socket
2479 nbds://tlshost/exportname
2481 .. option:: gpu_dev_ids=str : [libcufile]
2483 Specify the GPU IDs to use with CUDA. This is a colon-separated list of
2484 int. GPUs are assigned to workers roundrobin. Default is 0.
2486 .. option:: cuda_io=str : [libcufile]
2488 Specify the type of I/O to use with CUDA. Default is **cufile**.
2491 Use libcufile and nvidia-fs. This option performs I/O directly
2492 between a GPUDirect Storage filesystem and GPU buffers,
2493 avoiding use of a bounce buffer. If :option:`verify` is set,
2494 cudaMemcpy is used to copy verificaton data between RAM and GPU.
2495 Verification data is copied from RAM to GPU before a write
2496 and from GPU to RAM after a read. :option:`direct` must be 1.
2498 Use POSIX to perform I/O with a RAM buffer, and use cudaMemcpy
2499 to transfer data between RAM and the GPUs. Data is copied from
2500 GPU to RAM before a write and copied from RAM to GPU after a
2501 read. :option:`verify` does not affect use of cudaMemcpy.
2503 .. option:: pool=str : [dfs]
2505 Specify the UUID of the DAOS pool to connect to.
2507 .. option:: cont=str : [dfs]
2509 Specify the UUID of the DAOS container to open.
2511 .. option:: chunk_size=int : [dfs]
2513 Specificy a different chunk size (in bytes) for the dfs file.
2514 Use DAOS container's chunk size by default.
2516 .. option:: object_class=str : [dfs]
2518 Specificy a different object class for the dfs file.
2519 Use DAOS container's object class by default.
2524 .. option:: iodepth=int
2526 Number of I/O units to keep in flight against the file. Note that
2527 increasing *iodepth* beyond 1 will not affect synchronous ioengines (except
2528 for small degrees when :option:`verify_async` is in use). Even async
2529 engines may impose OS restrictions causing the desired depth not to be
2530 achieved. This may happen on Linux when using libaio and not setting
2531 :option:`direct`\=1, since buffered I/O is not async on that OS. Keep an
2532 eye on the I/O depth distribution in the fio output to verify that the
2533 achieved depth is as expected. Default: 1.
2535 .. option:: iodepth_batch_submit=int, iodepth_batch=int
2537 This defines how many pieces of I/O to submit at once. It defaults to 1
2538 which means that we submit each I/O as soon as it is available, but can be
2539 raised to submit bigger batches of I/O at the time. If it is set to 0 the
2540 :option:`iodepth` value will be used.
2542 .. option:: iodepth_batch_complete_min=int, iodepth_batch_complete=int
2544 This defines how many pieces of I/O to retrieve at once. It defaults to 1
2545 which means that we'll ask for a minimum of 1 I/O in the retrieval process
2546 from the kernel. The I/O retrieval will go on until we hit the limit set by
2547 :option:`iodepth_low`. If this variable is set to 0, then fio will always
2548 check for completed events before queuing more I/O. This helps reduce I/O
2549 latency, at the cost of more retrieval system calls.
2551 .. option:: iodepth_batch_complete_max=int
2553 This defines maximum pieces of I/O to retrieve at once. This variable should
2554 be used along with :option:`iodepth_batch_complete_min`\=int variable,
2555 specifying the range of min and max amount of I/O which should be
2556 retrieved. By default it is equal to the :option:`iodepth_batch_complete_min`
2561 iodepth_batch_complete_min=1
2562 iodepth_batch_complete_max=<iodepth>
2564 which means that we will retrieve at least 1 I/O and up to the whole
2565 submitted queue depth. If none of I/O has been completed yet, we will wait.
2569 iodepth_batch_complete_min=0
2570 iodepth_batch_complete_max=<iodepth>
2572 which means that we can retrieve up to the whole submitted queue depth, but
2573 if none of I/O has been completed yet, we will NOT wait and immediately exit
2574 the system call. In this example we simply do polling.
2576 .. option:: iodepth_low=int
2578 The low water mark indicating when to start filling the queue
2579 again. Defaults to the same as :option:`iodepth`, meaning that fio will
2580 attempt to keep the queue full at all times. If :option:`iodepth` is set to
2581 e.g. 16 and *iodepth_low* is set to 4, then after fio has filled the queue of
2582 16 requests, it will let the depth drain down to 4 before starting to fill
2585 .. option:: serialize_overlap=bool
2587 Serialize in-flight I/Os that might otherwise cause or suffer from data races.
2588 When two or more I/Os are submitted simultaneously, there is no guarantee that
2589 the I/Os will be processed or completed in the submitted order. Further, if
2590 two or more of those I/Os are writes, any overlapping region between them can
2591 become indeterminate/undefined on certain storage. These issues can cause
2592 verification to fail erratically when at least one of the racing I/Os is
2593 changing data and the overlapping region has a non-zero size. Setting
2594 ``serialize_overlap`` tells fio to avoid provoking this behavior by explicitly
2595 serializing in-flight I/Os that have a non-zero overlap. Note that setting
2596 this option can reduce both performance and the :option:`iodepth` achieved.
2598 This option only applies to I/Os issued for a single job except when it is
2599 enabled along with :option:`io_submit_mode`\=offload. In offload mode, fio
2600 will check for overlap among all I/Os submitted by offload jobs with :option:`serialize_overlap`
2605 .. option:: io_submit_mode=str
2607 This option controls how fio submits the I/O to the I/O engine. The default
2608 is `inline`, which means that the fio job threads submit and reap I/O
2609 directly. If set to `offload`, the job threads will offload I/O submission
2610 to a dedicated pool of I/O threads. This requires some coordination and thus
2611 has a bit of extra overhead, especially for lower queue depth I/O where it
2612 can increase latencies. The benefit is that fio can manage submission rates
2613 independently of the device completion rates. This avoids skewed latency
2614 reporting if I/O gets backed up on the device side (the coordinated omission
2615 problem). Note that this option cannot reliably be used with async IO
2622 .. option:: thinktime=time
2624 Stall the job for the specified period of time after an I/O has completed before issuing the
2625 next. May be used to simulate processing being done by an application.
2626 When the unit is omitted, the value is interpreted in microseconds. See
2627 :option:`thinktime_blocks` and :option:`thinktime_spin`.
2629 .. option:: thinktime_spin=time
2631 Only valid if :option:`thinktime` is set - pretend to spend CPU time doing
2632 something with the data received, before falling back to sleeping for the
2633 rest of the period specified by :option:`thinktime`. When the unit is
2634 omitted, the value is interpreted in microseconds.
2636 .. option:: thinktime_blocks=int
2638 Only valid if :option:`thinktime` is set - control how many blocks to issue,
2639 before waiting :option:`thinktime` usecs. If not set, defaults to 1 which will make
2640 fio wait :option:`thinktime` usecs after every block. This effectively makes any
2641 queue depth setting redundant, since no more than 1 I/O will be queued
2642 before we have to complete it and do our :option:`thinktime`. In other words, this
2643 setting effectively caps the queue depth if the latter is larger.
2645 .. option:: thinktime_blocks_type=str
2647 Only valid if :option:`thinktime` is set - control how :option:`thinktime_blocks`
2648 triggers. The default is `complete`, which triggers thinktime when fio completes
2649 :option:`thinktime_blocks` blocks. If this is set to `issue`, then the trigger happens
2652 .. option:: rate=int[,int][,int]
2654 Cap the bandwidth used by this job. The number is in bytes/sec, the normal
2655 suffix rules apply. Comma-separated values may be specified for reads,
2656 writes, and trims as described in :option:`blocksize`.
2658 For example, using `rate=1m,500k` would limit reads to 1MiB/sec and writes to
2659 500KiB/sec. Capping only reads or writes can be done with `rate=,500k` or
2660 `rate=500k,` where the former will only limit writes (to 500KiB/sec) and the
2661 latter will only limit reads.
2663 .. option:: rate_min=int[,int][,int]
2665 Tell fio to do whatever it can to maintain at least this bandwidth. Failing
2666 to meet this requirement will cause the job to exit. Comma-separated values
2667 may be specified for reads, writes, and trims as described in
2668 :option:`blocksize`.
2670 .. option:: rate_iops=int[,int][,int]
2672 Cap the bandwidth to this number of IOPS. Basically the same as
2673 :option:`rate`, just specified independently of bandwidth. If the job is
2674 given a block size range instead of a fixed value, the smallest block size
2675 is used as the metric. Comma-separated values may be specified for reads,
2676 writes, and trims as described in :option:`blocksize`.
2678 .. option:: rate_iops_min=int[,int][,int]
2680 If fio doesn't meet this rate of I/O, it will cause the job to exit.
2681 Comma-separated values may be specified for reads, writes, and trims as
2682 described in :option:`blocksize`.
2684 .. option:: rate_process=str
2686 This option controls how fio manages rated I/O submissions. The default is
2687 `linear`, which submits I/O in a linear fashion with fixed delays between
2688 I/Os that gets adjusted based on I/O completion rates. If this is set to
2689 `poisson`, fio will submit I/O based on a more real world random request
2690 flow, known as the Poisson process
2691 (https://en.wikipedia.org/wiki/Poisson_point_process). The lambda will be
2692 10^6 / IOPS for the given workload.
2694 .. option:: rate_ignore_thinktime=bool
2696 By default, fio will attempt to catch up to the specified rate setting,
2697 if any kind of thinktime setting was used. If this option is set, then
2698 fio will ignore the thinktime and continue doing IO at the specified
2699 rate, instead of entering a catch-up mode after thinktime is done.
2705 .. option:: latency_target=time
2707 If set, fio will attempt to find the max performance point that the given
2708 workload will run at while maintaining a latency below this target. When
2709 the unit is omitted, the value is interpreted in microseconds. See
2710 :option:`latency_window` and :option:`latency_percentile`.
2712 .. option:: latency_window=time
2714 Used with :option:`latency_target` to specify the sample window that the job
2715 is run at varying queue depths to test the performance. When the unit is
2716 omitted, the value is interpreted in microseconds.
2718 .. option:: latency_percentile=float
2720 The percentage of I/Os that must fall within the criteria specified by
2721 :option:`latency_target` and :option:`latency_window`. If not set, this
2722 defaults to 100.0, meaning that all I/Os must be equal or below to the value
2723 set by :option:`latency_target`.
2725 .. option:: latency_run=bool
2727 Used with :option:`latency_target`. If false (default), fio will find
2728 the highest queue depth that meets :option:`latency_target` and exit. If
2729 true, fio will continue running and try to meet :option:`latency_target`
2730 by adjusting queue depth.
2732 .. option:: max_latency=time[,time][,time]
2734 If set, fio will exit the job with an ETIMEDOUT error if it exceeds this
2735 maximum latency. When the unit is omitted, the value is interpreted in
2736 microseconds. Comma-separated values may be specified for reads, writes,
2737 and trims as described in :option:`blocksize`.
2739 .. option:: rate_cycle=int
2741 Average bandwidth for :option:`rate` and :option:`rate_min` over this number
2742 of milliseconds. Defaults to 1000.
2748 .. option:: write_iolog=str
2750 Write the issued I/O patterns to the specified file. See
2751 :option:`read_iolog`. Specify a separate file for each job, otherwise the
2752 iologs will be interspersed and the file may be corrupt.
2754 .. option:: read_iolog=str
2756 Open an iolog with the specified filename and replay the I/O patterns it
2757 contains. This can be used to store a workload and replay it sometime
2758 later. The iolog given may also be a blktrace binary file, which allows fio
2759 to replay a workload captured by :command:`blktrace`. See
2760 :manpage:`blktrace(8)` for how to capture such logging data. For blktrace
2761 replay, the file needs to be turned into a blkparse binary data file first
2762 (``blkparse <device> -o /dev/null -d file_for_fio.bin``).
2763 You can specify a number of files by separating the names with a ':'
2764 character. See the :option:`filename` option for information on how to
2765 escape ':' characters within the file names. These files will
2766 be sequentially assigned to job clones created by :option:`numjobs`.
2767 '-' is a reserved name, meaning read from stdin, notably if
2768 :option:`filename` is set to '-' which means stdin as well, then
2769 this flag can't be set to '-'.
2771 .. option:: read_iolog_chunked=bool
2773 Determines how iolog is read. If false(default) entire :option:`read_iolog`
2774 will be read at once. If selected true, input from iolog will be read
2775 gradually. Useful when iolog is very large, or it is generated.
2777 .. option:: merge_blktrace_file=str
2779 When specified, rather than replaying the logs passed to :option:`read_iolog`,
2780 the logs go through a merge phase which aggregates them into a single
2781 blktrace. The resulting file is then passed on as the :option:`read_iolog`
2782 parameter. The intention here is to make the order of events consistent.
2783 This limits the influence of the scheduler compared to replaying multiple
2784 blktraces via concurrent jobs.
2786 .. option:: merge_blktrace_scalars=float_list
2788 This is a percentage based option that is index paired with the list of
2789 files passed to :option:`read_iolog`. When merging is performed, scale
2790 the time of each event by the corresponding amount. For example,
2791 ``--merge_blktrace_scalars="50:100"`` runs the first trace in halftime
2792 and the second trace in realtime. This knob is separately tunable from
2793 :option:`replay_time_scale` which scales the trace during runtime and
2794 does not change the output of the merge unlike this option.
2796 .. option:: merge_blktrace_iters=float_list
2798 This is a whole number option that is index paired with the list of files
2799 passed to :option:`read_iolog`. When merging is performed, run each trace
2800 for the specified number of iterations. For example,
2801 ``--merge_blktrace_iters="2:1"`` runs the first trace for two iterations
2802 and the second trace for one iteration.
2804 .. option:: replay_no_stall=bool
2806 When replaying I/O with :option:`read_iolog` the default behavior is to
2807 attempt to respect the timestamps within the log and replay them with the
2808 appropriate delay between IOPS. By setting this variable fio will not
2809 respect the timestamps and attempt to replay them as fast as possible while
2810 still respecting ordering. The result is the same I/O pattern to a given
2811 device, but different timings.
2813 .. option:: replay_time_scale=int
2815 When replaying I/O with :option:`read_iolog`, fio will honor the
2816 original timing in the trace. With this option, it's possible to scale
2817 the time. It's a percentage option, if set to 50 it means run at 50%
2818 the original IO rate in the trace. If set to 200, run at twice the
2819 original IO rate. Defaults to 100.
2821 .. option:: replay_redirect=str
2823 While replaying I/O patterns using :option:`read_iolog` the default behavior
2824 is to replay the IOPS onto the major/minor device that each IOP was recorded
2825 from. This is sometimes undesirable because on a different machine those
2826 major/minor numbers can map to a different device. Changing hardware on the
2827 same system can also result in a different major/minor mapping.
2828 ``replay_redirect`` causes all I/Os to be replayed onto the single specified
2829 device regardless of the device it was recorded
2830 from. i.e. :option:`replay_redirect`\= :file:`/dev/sdc` would cause all I/O
2831 in the blktrace or iolog to be replayed onto :file:`/dev/sdc`. This means
2832 multiple devices will be replayed onto a single device, if the trace
2833 contains multiple devices. If you want multiple devices to be replayed
2834 concurrently to multiple redirected devices you must blkparse your trace
2835 into separate traces and replay them with independent fio invocations.
2836 Unfortunately this also breaks the strict time ordering between multiple
2839 .. option:: replay_align=int
2841 Force alignment of the byte offsets in a trace to this value. The value
2842 must be a power of 2.
2844 .. option:: replay_scale=int
2846 Scale byte offsets down by this factor when replaying traces. Should most
2847 likely use :option:`replay_align` as well.
2849 .. option:: replay_skip=str
2851 Sometimes it's useful to skip certain IO types in a replay trace.
2852 This could be, for instance, eliminating the writes in the trace.
2853 Or not replaying the trims/discards, if you are redirecting to
2854 a device that doesn't support them. This option takes a comma
2855 separated list of read, write, trim, sync.
2858 Threads, processes and job synchronization
2859 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2863 Fio defaults to creating jobs by using fork, however if this option is
2864 given, fio will create jobs by using POSIX Threads' function
2865 :manpage:`pthread_create(3)` to create threads instead.
2867 .. option:: wait_for=str
2869 If set, the current job won't be started until all workers of the specified
2870 waitee job are done.
2872 ``wait_for`` operates on the job name basis, so there are a few
2873 limitations. First, the waitee must be defined prior to the waiter job
2874 (meaning no forward references). Second, if a job is being referenced as a
2875 waitee, it must have a unique name (no duplicate waitees).
2877 .. option:: nice=int
2879 Run the job with the given nice value. See man :manpage:`nice(2)`.
2881 On Windows, values less than -15 set the process class to "High"; -1 through
2882 -15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle"
2885 .. option:: prio=int
2887 Set the I/O priority value of this job. Linux limits us to a positive value
2888 between 0 and 7, with 0 being the highest. See man
2889 :manpage:`ionice(1)`. Refer to an appropriate manpage for other operating
2890 systems since meaning of priority may differ. For per-command priority
2891 setting, see I/O engine specific `cmdprio_percentage` and `hipri_percentage`
2894 .. option:: prioclass=int
2896 Set the I/O priority class. See man :manpage:`ionice(1)`. For per-command
2897 priority setting, see I/O engine specific `cmdprio_percentage` and
2898 `hipri_percentage` options.
2900 .. option:: cpus_allowed=str
2902 Controls the same options as :option:`cpumask`, but accepts a textual
2903 specification of the permitted CPUs instead and CPUs are indexed from 0. So
2904 to use CPUs 0 and 5 you would specify ``cpus_allowed=0,5``. This option also
2905 allows a range of CPUs to be specified -- say you wanted a binding to CPUs
2906 0, 5, and 8 to 15, you would set ``cpus_allowed=0,5,8-15``.
2908 On Windows, when ``cpus_allowed`` is unset only CPUs from fio's current
2909 processor group will be used and affinity settings are inherited from the
2910 system. An fio build configured to target Windows 7 makes options that set
2911 CPUs processor group aware and values will set both the processor group
2912 and a CPU from within that group. For example, on a system where processor
2913 group 0 has 40 CPUs and processor group 1 has 32 CPUs, ``cpus_allowed``
2914 values between 0 and 39 will bind CPUs from processor group 0 and
2915 ``cpus_allowed`` values between 40 and 71 will bind CPUs from processor
2916 group 1. When using ``cpus_allowed_policy=shared`` all CPUs specified by a
2917 single ``cpus_allowed`` option must be from the same processor group. For
2918 Windows fio builds not built for Windows 7, CPUs will only be selected from
2919 (and be relative to) whatever processor group fio happens to be running in
2920 and CPUs from other processor groups cannot be used.
2922 .. option:: cpus_allowed_policy=str
2924 Set the policy of how fio distributes the CPUs specified by
2925 :option:`cpus_allowed` or :option:`cpumask`. Two policies are supported:
2928 All jobs will share the CPU set specified.
2930 Each job will get a unique CPU from the CPU set.
2932 **shared** is the default behavior, if the option isn't specified. If
2933 **split** is specified, then fio will assign one cpu per job. If not
2934 enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs
2937 .. option:: cpumask=int
2939 Set the CPU affinity of this job. The parameter given is a bit mask of
2940 allowed CPUs the job may run on. So if you want the allowed CPUs to be 1
2941 and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man
2942 :manpage:`sched_setaffinity(2)`. This may not work on all supported
2943 operating systems or kernel versions. This option doesn't work well for a
2944 higher CPU count than what you can store in an integer mask, so it can only
2945 control cpus 1-32. For boxes with larger CPU counts, use
2946 :option:`cpus_allowed`.
2948 .. option:: numa_cpu_nodes=str
2950 Set this job running on specified NUMA nodes' CPUs. The arguments allow
2951 comma delimited list of cpu numbers, A-B ranges, or `all`. Note, to enable
2952 NUMA options support, fio must be built on a system with libnuma-dev(el)
2955 .. option:: numa_mem_policy=str
2957 Set this job's memory policy and corresponding NUMA nodes. Format of the
2962 ``mode`` is one of the following memory policies: ``default``, ``prefer``,
2963 ``bind``, ``interleave`` or ``local``. For ``default`` and ``local`` memory
2964 policies, no node needs to be specified. For ``prefer``, only one node is
2965 allowed. For ``bind`` and ``interleave`` the ``nodelist`` may be as
2966 follows: a comma delimited list of numbers, A-B ranges, or `all`.
2968 .. option:: cgroup=str
2970 Add job to this control group. If it doesn't exist, it will be created. The
2971 system must have a mounted cgroup blkio mount point for this to work. If
2972 your system doesn't have it mounted, you can do so with::
2974 # mount -t cgroup -o blkio none /cgroup
2976 .. option:: cgroup_weight=int
2978 Set the weight of the cgroup to this value. See the documentation that comes
2979 with the kernel, allowed values are in the range of 100..1000.
2981 .. option:: cgroup_nodelete=bool
2983 Normally fio will delete the cgroups it has created after the job
2984 completion. To override this behavior and to leave cgroups around after the
2985 job completion, set ``cgroup_nodelete=1``. This can be useful if one wants
2986 to inspect various cgroup files after job completion. Default: false.
2988 .. option:: flow_id=int
2990 The ID of the flow. If not specified, it defaults to being a global
2991 flow. See :option:`flow`.
2993 .. option:: flow=int
2995 Weight in token-based flow control. If this value is used, then there is a
2996 'flow counter' which is used to regulate the proportion of activity between
2997 two or more jobs. Fio attempts to keep this flow counter near zero. The
2998 ``flow`` parameter stands for how much should be added or subtracted to the
2999 flow counter on each iteration of the main I/O loop. That is, if one job has
3000 ``flow=8`` and another job has ``flow=-1``, then there will be a roughly 1:8
3001 ratio in how much one runs vs the other.
3003 .. option:: flow_sleep=int
3005 The period of time, in microseconds, to wait after the flow counter
3006 has exceeded its proportion before retrying operations.
3008 .. option:: stonewall, wait_for_previous
3010 Wait for preceding jobs in the job file to exit, before starting this
3011 one. Can be used to insert serialization points in the job file. A stone
3012 wall also implies starting a new reporting group, see
3013 :option:`group_reporting`.
3017 By default, fio will continue running all other jobs when one job finishes.
3018 Sometimes this is not the desired action. Setting ``exitall`` will instead
3019 make fio terminate all jobs in the same group, as soon as one job of that
3022 .. option:: exit_what
3024 By default, fio will continue running all other jobs when one job finishes.
3025 Sometimes this is not the desired action. Setting ``exit_all`` will
3026 instead make fio terminate all jobs in the same group. The option
3027 ``exit_what`` allows to control which jobs get terminated when ``exitall`` is
3028 enabled. The default is ``group`` and does not change the behaviour of
3029 ``exitall``. The setting ``all`` terminates all jobs. The setting ``stonewall``
3030 terminates all currently running jobs across all groups and continues execution
3031 with the next stonewalled group.
3033 .. option:: exec_prerun=str
3035 Before running this job, issue the command specified through
3036 :manpage:`system(3)`. Output is redirected in a file called
3037 :file:`jobname.prerun.txt`.
3039 .. option:: exec_postrun=str
3041 After the job completes, issue the command specified though
3042 :manpage:`system(3)`. Output is redirected in a file called
3043 :file:`jobname.postrun.txt`.
3047 Instead of running as the invoking user, set the user ID to this value
3048 before the thread/process does any work.
3052 Set group ID, see :option:`uid`.
3058 .. option:: verify_only
3060 Do not perform specified workload, only verify data still matches previous
3061 invocation of this workload. This option allows one to check data multiple
3062 times at a later date without overwriting it. This option makes sense only
3063 for workloads that write data, and does not support workloads with the
3064 :option:`time_based` option set.
3066 .. option:: do_verify=bool
3068 Run the verify phase after a write phase. Only valid if :option:`verify` is
3071 .. option:: verify=str
3073 If writing to a file, fio can verify the file contents after each iteration
3074 of the job. Each verification method also implies verification of special
3075 header, which is written to the beginning of each block. This header also
3076 includes meta information, like offset of the block, block number, timestamp
3077 when block was written, etc. :option:`verify` can be combined with
3078 :option:`verify_pattern` option. The allowed values are:
3081 Use an md5 sum of the data area and store it in the header of
3085 Use an experimental crc64 sum of the data area and store it in the
3086 header of each block.
3089 Use a crc32c sum of the data area and store it in the header of
3090 each block. This will automatically use hardware acceleration
3091 (e.g. SSE4.2 on an x86 or CRC crypto extensions on ARM64) but will
3092 fall back to software crc32c if none is found. Generally the
3093 fastest checksum fio supports when hardware accelerated.
3099 Use a crc32 sum of the data area and store it in the header of each
3103 Use a crc16 sum of the data area and store it in the header of each
3107 Use a crc7 sum of the data area and store it in the header of each
3111 Use xxhash as the checksum function. Generally the fastest software
3112 checksum that fio supports.
3115 Use sha512 as the checksum function.
3118 Use sha256 as the checksum function.
3121 Use optimized sha1 as the checksum function.
3124 Use optimized sha3-224 as the checksum function.
3127 Use optimized sha3-256 as the checksum function.
3130 Use optimized sha3-384 as the checksum function.
3133 Use optimized sha3-512 as the checksum function.
3136 This option is deprecated, since now meta information is included in
3137 generic verification header and meta verification happens by
3138 default. For detailed information see the description of the
3139 :option:`verify` setting. This option is kept because of
3140 compatibility's sake with old configurations. Do not use it.
3143 Verify a strict pattern. Normally fio includes a header with some
3144 basic information and checksumming, but if this option is set, only
3145 the specific pattern set with :option:`verify_pattern` is verified.
3148 Only pretend to verify. Useful for testing internals with
3149 :option:`ioengine`\=null, not for much else.
3151 This option can be used for repeated burn-in tests of a system to make sure
3152 that the written data is also correctly read back. If the data direction
3153 given is a read or random read, fio will assume that it should verify a
3154 previously written file. If the data direction includes any form of write,
3155 the verify will be of the newly written data.
3157 To avoid false verification errors, do not use the norandommap option when
3158 verifying data with async I/O engines and I/O depths > 1. Or use the
3159 norandommap and the lfsr random generator together to avoid writing to the
3160 same offset with muliple outstanding I/Os.
3162 .. option:: verify_offset=int
3164 Swap the verification header with data somewhere else in the block before
3165 writing. It is swapped back before verifying.
3167 .. option:: verify_interval=int
3169 Write the verification header at a finer granularity than the
3170 :option:`blocksize`. It will be written for chunks the size of
3171 ``verify_interval``. :option:`blocksize` should divide this evenly.
3173 .. option:: verify_pattern=str
3175 If set, fio will fill the I/O buffers with this pattern. Fio defaults to
3176 filling with totally random bytes, but sometimes it's interesting to fill
3177 with a known pattern for I/O verification purposes. Depending on the width
3178 of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time (it can
3179 be either a decimal or a hex number). The ``verify_pattern`` if larger than
3180 a 32-bit quantity has to be a hex number that starts with either "0x" or
3181 "0X". Use with :option:`verify`. Also, ``verify_pattern`` supports %o
3182 format, which means that for each block offset will be written and then
3183 verified back, e.g.::
3187 Or use combination of everything::
3189 verify_pattern=0xff%o"abcd"-12
3191 .. option:: verify_fatal=bool
3193 Normally fio will keep checking the entire contents before quitting on a
3194 block verification failure. If this option is set, fio will exit the job on
3195 the first observed failure. Default: false.
3197 .. option:: verify_dump=bool
3199 If set, dump the contents of both the original data block and the data block
3200 we read off disk to files. This allows later analysis to inspect just what
3201 kind of data corruption occurred. Off by default.
3203 .. option:: verify_async=int
3205 Fio will normally verify I/O inline from the submitting thread. This option
3206 takes an integer describing how many async offload threads to create for I/O
3207 verification instead, causing fio to offload the duty of verifying I/O
3208 contents to one or more separate threads. If using this offload option, even
3209 sync I/O engines can benefit from using an :option:`iodepth` setting higher
3210 than 1, as it allows them to have I/O in flight while verifies are running.
3211 Defaults to 0 async threads, i.e. verification is not asynchronous.
3213 .. option:: verify_async_cpus=str
3215 Tell fio to set the given CPU affinity on the async I/O verification
3216 threads. See :option:`cpus_allowed` for the format used.
3218 .. option:: verify_backlog=int
3220 Fio will normally verify the written contents of a job that utilizes verify
3221 once that job has completed. In other words, everything is written then
3222 everything is read back and verified. You may want to verify continually
3223 instead for a variety of reasons. Fio stores the meta data associated with
3224 an I/O block in memory, so for large verify workloads, quite a bit of memory
3225 would be used up holding this meta data. If this option is enabled, fio will
3226 write only N blocks before verifying these blocks.
3228 .. option:: verify_backlog_batch=int
3230 Control how many blocks fio will verify if :option:`verify_backlog` is
3231 set. If not set, will default to the value of :option:`verify_backlog`
3232 (meaning the entire queue is read back and verified). If
3233 ``verify_backlog_batch`` is less than :option:`verify_backlog` then not all
3234 blocks will be verified, if ``verify_backlog_batch`` is larger than
3235 :option:`verify_backlog`, some blocks will be verified more than once.
3237 .. option:: verify_state_save=bool
3239 When a job exits during the write phase of a verify workload, save its
3240 current state. This allows fio to replay up until that point, if the verify
3241 state is loaded for the verify read phase. The format of the filename is,
3244 <type>-<jobname>-<jobindex>-verify.state.
3246 <type> is "local" for a local run, "sock" for a client/server socket
3247 connection, and "ip" (192.168.0.1, for instance) for a networked
3248 client/server connection. Defaults to true.
3250 .. option:: verify_state_load=bool
3252 If a verify termination trigger was used, fio stores the current write state
3253 of each thread. This can be used at verification time so that fio knows how
3254 far it should verify. Without this information, fio will run a full
3255 verification pass, according to the settings in the job file used. Default
3258 .. option:: trim_percentage=int
3260 Number of verify blocks to discard/trim.
3262 .. option:: trim_verify_zero=bool
3264 Verify that trim/discarded blocks are returned as zeros.
3266 .. option:: trim_backlog=int
3268 Trim after this number of blocks are written.
3270 .. option:: trim_backlog_batch=int
3272 Trim this number of I/O blocks.
3274 .. option:: experimental_verify=bool
3276 Enable experimental verification.
3281 .. option:: steadystate=str:float, ss=str:float
3283 Define the criterion and limit for assessing steady state performance. The
3284 first parameter designates the criterion whereas the second parameter sets
3285 the threshold. When the criterion falls below the threshold for the
3286 specified duration, the job will stop. For example, `iops_slope:0.1%` will
3287 direct fio to terminate the job when the least squares regression slope
3288 falls below 0.1% of the mean IOPS. If :option:`group_reporting` is enabled
3289 this will apply to all jobs in the group. Below is the list of available
3290 steady state assessment criteria. All assessments are carried out using only
3291 data from the rolling collection window. Threshold limits can be expressed
3292 as a fixed value or as a percentage of the mean in the collection window.
3294 When using this feature, most jobs should include the :option:`time_based`
3295 and :option:`runtime` options or the :option:`loops` option so that fio does not
3296 stop running after it has covered the full size of the specified file(s) or device(s).
3299 Collect IOPS data. Stop the job if all individual IOPS measurements
3300 are within the specified limit of the mean IOPS (e.g., ``iops:2``
3301 means that all individual IOPS values must be within 2 of the mean,
3302 whereas ``iops:0.2%`` means that all individual IOPS values must be
3303 within 0.2% of the mean IOPS to terminate the job).
3306 Collect IOPS data and calculate the least squares regression
3307 slope. Stop the job if the slope falls below the specified limit.
3310 Collect bandwidth data. Stop the job if all individual bandwidth
3311 measurements are within the specified limit of the mean bandwidth.
3314 Collect bandwidth data and calculate the least squares regression
3315 slope. Stop the job if the slope falls below the specified limit.
3317 .. option:: steadystate_duration=time, ss_dur=time
3319 A rolling window of this duration will be used to judge whether steady state
3320 has been reached. Data will be collected once per second. The default is 0
3321 which disables steady state detection. When the unit is omitted, the
3322 value is interpreted in seconds.
3324 .. option:: steadystate_ramp_time=time, ss_ramp=time
3326 Allow the job to run for the specified duration before beginning data
3327 collection for checking the steady state job termination criterion. The
3328 default is 0. When the unit is omitted, the value is interpreted in seconds.
3331 Measurements and reporting
3332 ~~~~~~~~~~~~~~~~~~~~~~~~~~
3334 .. option:: per_job_logs=bool
3336 If set, this generates bw/clat/iops log with per file private filenames. If
3337 not set, jobs with identical names will share the log filename. Default:
3340 .. option:: group_reporting
3342 It may sometimes be interesting to display statistics for groups of jobs as
3343 a whole instead of for each individual job. This is especially true if
3344 :option:`numjobs` is used; looking at individual thread/process output
3345 quickly becomes unwieldy. To see the final report per-group instead of
3346 per-job, use :option:`group_reporting`. Jobs in a file will be part of the
3347 same reporting group, unless if separated by a :option:`stonewall`, or by
3348 using :option:`new_group`.
3350 .. option:: new_group
3352 Start a new reporting group. See: :option:`group_reporting`. If not given,
3353 all jobs in a file will be part of the same reporting group, unless
3354 separated by a :option:`stonewall`.
3356 .. option:: stats=bool
3358 By default, fio collects and shows final output results for all jobs
3359 that run. If this option is set to 0, then fio will ignore it in
3360 the final stat output.
3362 .. option:: write_bw_log=str
3364 If given, write a bandwidth log for this job. Can be used to store data of
3365 the bandwidth of the jobs in their lifetime.
3367 If no str argument is given, the default filename of
3368 :file:`jobname_type.x.log` is used. Even when the argument is given, fio
3369 will still append the type of log. So if one specifies::
3373 The actual log name will be :file:`foo_bw.x.log` where `x` is the index
3374 of the job (`1..N`, where `N` is the number of jobs). If
3375 :option:`per_job_logs` is false, then the filename will not include the
3378 The included :command:`fio_generate_plots` script uses :command:`gnuplot` to turn these
3379 text files into nice graphs. See `Log File Formats`_ for how data is
3380 structured within the file.
3382 .. option:: write_lat_log=str
3384 Same as :option:`write_bw_log`, except this option creates I/O
3385 submission (e.g., :file:`name_slat.x.log`), completion (e.g.,
3386 :file:`name_clat.x.log`), and total (e.g., :file:`name_lat.x.log`)
3387 latency files instead. See :option:`write_bw_log` for details about
3388 the filename format and `Log File Formats`_ for how data is structured
3391 .. option:: write_hist_log=str
3393 Same as :option:`write_bw_log` but writes an I/O completion latency
3394 histogram file (e.g., :file:`name_hist.x.log`) instead. Note that this
3395 file will be empty unless :option:`log_hist_msec` has also been set.
3396 See :option:`write_bw_log` for details about the filename format and
3397 `Log File Formats`_ for how data is structured within the file.
3399 .. option:: write_iops_log=str
3401 Same as :option:`write_bw_log`, but writes an IOPS file (e.g.
3402 :file:`name_iops.x.log`) instead. Because fio defaults to individual
3403 I/O logging, the value entry in the IOPS log will be 1 unless windowed
3404 logging (see :option:`log_avg_msec`) has been enabled. See
3405 :option:`write_bw_log` for details about the filename format and `Log
3406 File Formats`_ for how data is structured within the file.
3408 .. option:: log_avg_msec=int
3410 By default, fio will log an entry in the iops, latency, or bw log for every
3411 I/O that completes. When writing to the disk log, that can quickly grow to a
3412 very large size. Setting this option makes fio average the each log entry
3413 over the specified period of time, reducing the resolution of the log. See
3414 :option:`log_max_value` as well. Defaults to 0, logging all entries.
3415 Also see `Log File Formats`_.
3417 .. option:: log_hist_msec=int
3419 Same as :option:`log_avg_msec`, but logs entries for completion latency
3420 histograms. Computing latency percentiles from averages of intervals using
3421 :option:`log_avg_msec` is inaccurate. Setting this option makes fio log
3422 histogram entries over the specified period of time, reducing log sizes for
3423 high IOPS devices while retaining percentile accuracy. See
3424 :option:`log_hist_coarseness` and :option:`write_hist_log` as well.
3425 Defaults to 0, meaning histogram logging is disabled.
3427 .. option:: log_hist_coarseness=int
3429 Integer ranging from 0 to 6, defining the coarseness of the resolution of
3430 the histogram logs enabled with :option:`log_hist_msec`. For each increment
3431 in coarseness, fio outputs half as many bins. Defaults to 0, for which
3432 histogram logs contain 1216 latency bins. See :option:`write_hist_log`
3433 and `Log File Formats`_.
3435 .. option:: log_max_value=bool
3437 If :option:`log_avg_msec` is set, fio logs the average over that window. If
3438 you instead want to log the maximum value, set this option to 1. Defaults to
3439 0, meaning that averaged values are logged.
3441 .. option:: log_offset=bool
3443 If this is set, the iolog options will include the byte offset for the I/O
3444 entry as well as the other data values. Defaults to 0 meaning that
3445 offsets are not present in logs. Also see `Log File Formats`_.
3447 .. option:: log_compression=int
3449 If this is set, fio will compress the I/O logs as it goes, to keep the
3450 memory footprint lower. When a log reaches the specified size, that chunk is
3451 removed and compressed in the background. Given that I/O logs are fairly
3452 highly compressible, this yields a nice memory savings for longer runs. The
3453 downside is that the compression will consume some background CPU cycles, so
3454 it may impact the run. This, however, is also true if the logging ends up
3455 consuming most of the system memory. So pick your poison. The I/O logs are
3456 saved normally at the end of a run, by decompressing the chunks and storing
3457 them in the specified log file. This feature depends on the availability of
3460 .. option:: log_compression_cpus=str
3462 Define the set of CPUs that are allowed to handle online log compression for
3463 the I/O jobs. This can provide better isolation between performance
3464 sensitive jobs, and background compression work. See
3465 :option:`cpus_allowed` for the format used.
3467 .. option:: log_store_compressed=bool
3469 If set, fio will store the log files in a compressed format. They can be
3470 decompressed with fio, using the :option:`--inflate-log` command line
3471 parameter. The files will be stored with a :file:`.fz` suffix.
3473 .. option:: log_unix_epoch=bool
3475 If set, fio will log Unix timestamps to the log files produced by enabling
3476 write_type_log for each log type, instead of the default zero-based
3479 .. option:: block_error_percentiles=bool
3481 If set, record errors in trim block-sized units from writes and trims and
3482 output a histogram of how many trims it took to get to errors, and what kind
3483 of error was encountered.
3485 .. option:: bwavgtime=int
3487 Average the calculated bandwidth over the given time. Value is specified in
3488 milliseconds. If the job also does bandwidth logging through
3489 :option:`write_bw_log`, then the minimum of this option and
3490 :option:`log_avg_msec` will be used. Default: 500ms.
3492 .. option:: iopsavgtime=int
3494 Average the calculated IOPS over the given time. Value is specified in
3495 milliseconds. If the job also does IOPS logging through
3496 :option:`write_iops_log`, then the minimum of this option and
3497 :option:`log_avg_msec` will be used. Default: 500ms.
3499 .. option:: disk_util=bool
3501 Generate disk utilization statistics, if the platform supports it.
3504 .. option:: disable_lat=bool
3506 Disable measurements of total latency numbers. Useful only for cutting back
3507 the number of calls to :manpage:`gettimeofday(2)`, as that does impact
3508 performance at really high IOPS rates. Note that to really get rid of a
3509 large amount of these calls, this option must be used with
3510 :option:`disable_slat` and :option:`disable_bw_measurement` as well.
3512 .. option:: disable_clat=bool
3514 Disable measurements of completion latency numbers. See
3515 :option:`disable_lat`.
3517 .. option:: disable_slat=bool
3519 Disable measurements of submission latency numbers. See
3520 :option:`disable_lat`.
3522 .. option:: disable_bw_measurement=bool, disable_bw=bool
3524 Disable measurements of throughput/bandwidth numbers. See
3525 :option:`disable_lat`.
3527 .. option:: slat_percentiles=bool
3529 Report submission latency percentiles. Submission latency is not recorded
3530 for synchronous ioengines.
3532 .. option:: clat_percentiles=bool
3534 Report completion latency percentiles.
3536 .. option:: lat_percentiles=bool
3538 Report total latency percentiles. Total latency is the sum of submission
3539 latency and completion latency.
3541 .. option:: percentile_list=float_list
3543 Overwrite the default list of percentiles for latencies and the block error
3544 histogram. Each number is a floating point number in the range (0,100], and
3545 the maximum length of the list is 20. Use ``:`` to separate the numbers. For
3546 example, ``--percentile_list=99.5:99.9`` will cause fio to report the
3547 latency durations below which 99.5% and 99.9% of the observed latencies fell,
3550 .. option:: significant_figures=int
3552 If using :option:`--output-format` of `normal`, set the significant
3553 figures to this value. Higher values will yield more precise IOPS and
3554 throughput units, while lower values will round. Requires a minimum
3555 value of 1 and a maximum value of 10. Defaults to 4.
3561 .. option:: exitall_on_error
3563 When one job finishes in error, terminate the rest. The default is to wait
3564 for each job to finish.
3566 .. option:: continue_on_error=str
3568 Normally fio will exit the job on the first observed failure. If this option
3569 is set, fio will continue the job when there is a 'non-fatal error' (EIO or
3570 EILSEQ) until the runtime is exceeded or the I/O size specified is
3571 completed. If this option is used, there are two more stats that are
3572 appended, the total error count and the first error. The error field given
3573 in the stats is the first error that was hit during the run.
3575 The allowed values are:
3578 Exit on any I/O or verify errors.
3581 Continue on read errors, exit on all others.
3584 Continue on write errors, exit on all others.
3587 Continue on any I/O error, exit on all others.
3590 Continue on verify errors, exit on all others.
3593 Continue on all errors.
3596 Backward-compatible alias for 'none'.
3599 Backward-compatible alias for 'all'.
3601 .. option:: ignore_error=str
3603 Sometimes you want to ignore some errors during test in that case you can
3604 specify error list for each error type, instead of only being able to
3605 ignore the default 'non-fatal error' using :option:`continue_on_error`.
3606 ``ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST`` errors for
3607 given error type is separated with ':'. Error may be symbol ('ENOSPC',
3608 'ENOMEM') or integer. Example::
3610 ignore_error=EAGAIN,ENOSPC:122
3612 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from
3613 WRITE. This option works by overriding :option:`continue_on_error` with
3614 the list of errors for each error type if any.
3616 .. option:: error_dump=bool
3618 If set dump every error even if it is non fatal, true by default. If
3619 disabled only fatal error will be dumped.
3621 Running predefined workloads
3622 ----------------------------
3624 Fio includes predefined profiles that mimic the I/O workloads generated by
3627 .. option:: profile=str
3629 The predefined workload to run. Current profiles are:
3632 Threaded I/O bench (tiotest/tiobench) like workload.
3635 Aerospike Certification Tool (ACT) like workload.
3637 To view a profile's additional options use :option:`--cmdhelp` after specifying
3638 the profile. For example::
3640 $ fio --profile=act --cmdhelp
3645 .. option:: device-names=str
3650 .. option:: load=int
3653 ACT load multiplier. Default: 1.
3655 .. option:: test-duration=time
3658 How long the entire test takes to run. When the unit is omitted, the value
3659 is given in seconds. Default: 24h.
3661 .. option:: threads-per-queue=int
3664 Number of read I/O threads per device. Default: 8.
3666 .. option:: read-req-num-512-blocks=int
3669 Number of 512B blocks to read at the time. Default: 3.
3671 .. option:: large-block-op-kbytes=int
3674 Size of large block ops in KiB (writes). Default: 131072.
3679 Set to run ACT prep phase.
3681 Tiobench profile options
3682 ~~~~~~~~~~~~~~~~~~~~~~~~
3684 .. option:: size=str
3689 .. option:: block=int
3692 Block size in bytes. Default: 4096.
3694 .. option:: numruns=int
3704 .. option:: threads=int
3709 Interpreting the output
3710 -----------------------
3713 Example output was based on the following:
3714 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --time_based \
3715 --rate=1256k --bs=14K --name=quick --runtime=1s --name=mixed \
3716 --runtime=2m --rw=rw
3718 Fio spits out a lot of output. While running, fio will display the status of the
3719 jobs created. An example of that would be::
3721 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]
3723 The characters inside the first set of square brackets denote the current status of
3724 each thread. The first character is the first job defined in the job file, and so
3725 forth. The possible values (in typical life cycle order) are:
3727 +------+-----+-----------------------------------------------------------+
3729 +======+=====+===========================================================+
3730 | P | | Thread setup, but not started. |
3731 +------+-----+-----------------------------------------------------------+
3732 | C | | Thread created. |
3733 +------+-----+-----------------------------------------------------------+
3734 | I | | Thread initialized, waiting or generating necessary data. |
3735 +------+-----+-----------------------------------------------------------+
3736 | | p | Thread running pre-reading file(s). |
3737 +------+-----+-----------------------------------------------------------+
3738 | | / | Thread is in ramp period. |
3739 +------+-----+-----------------------------------------------------------+
3740 | | R | Running, doing sequential reads. |
3741 +------+-----+-----------------------------------------------------------+
3742 | | r | Running, doing random reads. |
3743 +------+-----+-----------------------------------------------------------+
3744 | | W | Running, doing sequential writes. |
3745 +------+-----+-----------------------------------------------------------+
3746 | | w | Running, doing random writes. |
3747 +------+-----+-----------------------------------------------------------+
3748 | | M | Running, doing mixed sequential reads/writes. |
3749 +------+-----+-----------------------------------------------------------+
3750 | | m | Running, doing mixed random reads/writes. |
3751 +------+-----+-----------------------------------------------------------+
3752 | | D | Running, doing sequential trims. |
3753 +------+-----+-----------------------------------------------------------+
3754 | | d | Running, doing random trims. |
3755 +------+-----+-----------------------------------------------------------+
3756 | | F | Running, currently waiting for :manpage:`fsync(2)`. |
3757 +------+-----+-----------------------------------------------------------+
3758 | | V | Running, doing verification of written data. |
3759 +------+-----+-----------------------------------------------------------+
3760 | f | | Thread finishing. |
3761 +------+-----+-----------------------------------------------------------+
3762 | E | | Thread exited, not reaped by main thread yet. |
3763 +------+-----+-----------------------------------------------------------+
3764 | _ | | Thread reaped. |
3765 +------+-----+-----------------------------------------------------------+
3766 | X | | Thread reaped, exited with an error. |
3767 +------+-----+-----------------------------------------------------------+
3768 | K | | Thread reaped, exited due to signal. |
3769 +------+-----+-----------------------------------------------------------+
3772 Example output was based on the following:
3773 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --runtime=58m \
3774 --time_based --rate=2512k --bs=256K --numjobs=10 \
3775 --name=readers --rw=read --name=writers --rw=write
3777 Fio will condense the thread string as not to take up more space on the command
3778 line than needed. For instance, if you have 10 readers and 10 writers running,
3779 the output would look like this::
3781 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]
3783 Note that the status string is displayed in order, so it's possible to tell which of
3784 the jobs are currently doing what. In the example above this means that jobs 1--10
3785 are readers and 11--20 are writers.
3787 The other values are fairly self explanatory -- number of threads currently
3788 running and doing I/O, the number of currently open files (f=), the estimated
3789 completion percentage, the rate of I/O since last check (read speed listed first,
3790 then write speed and optionally trim speed) in terms of bandwidth and IOPS,
3791 and time to completion for the current running group. It's impossible to estimate
3792 runtime of the following groups (if any).
3795 Example output was based on the following:
3796 TZ=UTC fio --iodepth=16 --ioengine=posixaio --filename=/tmp/fiofile \
3797 --direct=1 --size=100M --time_based --runtime=50s --rate_iops=89 \
3798 --bs=7K --name=Client1 --rw=write
3800 When fio is done (or interrupted by :kbd:`Ctrl-C`), it will show the data for
3801 each thread, group of threads, and disks in that order. For each overall thread (or
3802 group) the output looks like::
3804 Client1: (groupid=0, jobs=1): err= 0: pid=16109: Sat Jun 24 12:07:54 2017
3805 write: IOPS=88, BW=623KiB/s (638kB/s)(30.4MiB/50032msec)
3806 slat (nsec): min=500, max=145500, avg=8318.00, stdev=4781.50
3807 clat (usec): min=170, max=78367, avg=4019.02, stdev=8293.31
3808 lat (usec): min=174, max=78375, avg=4027.34, stdev=8291.79
3809 clat percentiles (usec):
3810 | 1.00th=[ 302], 5.00th=[ 326], 10.00th=[ 343], 20.00th=[ 363],
3811 | 30.00th=[ 392], 40.00th=[ 404], 50.00th=[ 416], 60.00th=[ 445],
3812 | 70.00th=[ 816], 80.00th=[ 6718], 90.00th=[12911], 95.00th=[21627],
3813 | 99.00th=[43779], 99.50th=[51643], 99.90th=[68682], 99.95th=[72877],
3815 bw ( KiB/s): min= 532, max= 686, per=0.10%, avg=622.87, stdev=24.82, samples= 100
3816 iops : min= 76, max= 98, avg=88.98, stdev= 3.54, samples= 100
3817 lat (usec) : 250=0.04%, 500=64.11%, 750=4.81%, 1000=2.79%
3818 lat (msec) : 2=4.16%, 4=1.84%, 10=4.90%, 20=11.33%, 50=5.37%
3819 lat (msec) : 100=0.65%
3820 cpu : usr=0.27%, sys=0.18%, ctx=12072, majf=0, minf=21
3821 IO depths : 1=85.0%, 2=13.1%, 4=1.8%, 8=0.1%, 16=0.0%, 32=0.0%, >=64=0.0%
3822 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
3823 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
3824 issued rwt: total=0,4450,0, short=0,0,0, dropped=0,0,0
3825 latency : target=0, window=0, percentile=100.00%, depth=8
3827 The job name (or first job's name when using :option:`group_reporting`) is printed,
3828 along with the group id, count of jobs being aggregated, last error id seen (which
3829 is 0 when there are no errors), pid/tid of that thread and the time the job/group
3830 completed. Below are the I/O statistics for each data direction performed (showing
3831 writes in the example above). In the order listed, they denote:
3834 The string before the colon shows the I/O direction the statistics
3835 are for. **IOPS** is the average I/Os performed per second. **BW**
3836 is the average bandwidth rate shown as: value in power of 2 format
3837 (value in power of 10 format). The last two values show: (**total
3838 I/O performed** in power of 2 format / **runtime** of that thread).
3841 Submission latency (**min** being the minimum, **max** being the
3842 maximum, **avg** being the average, **stdev** being the standard
3843 deviation). This is the time it took to submit the I/O. For
3844 sync I/O this row is not displayed as the slat is really the
3845 completion latency (since queue/complete is one operation there).
3846 This value can be in nanoseconds, microseconds or milliseconds ---
3847 fio will choose the most appropriate base and print that (in the
3848 example above nanoseconds was the best scale). Note: in :option:`--minimal` mode
3849 latencies are always expressed in microseconds.
3852 Completion latency. Same names as slat, this denotes the time from
3853 submission to completion of the I/O pieces. For sync I/O, clat will
3854 usually be equal (or very close) to 0, as the time from submit to
3855 complete is basically just CPU time (I/O has already been done, see slat
3859 Total latency. Same names as slat and clat, this denotes the time from
3860 when fio created the I/O unit to completion of the I/O operation.
3863 Bandwidth statistics based on samples. Same names as the xlat stats,
3864 but also includes the number of samples taken (**samples**) and an
3865 approximate percentage of total aggregate bandwidth this thread
3866 received in its group (**per**). This last value is only really
3867 useful if the threads in this group are on the same disk, since they
3868 are then competing for disk access.
3871 IOPS statistics based on samples. Same names as bw.
3873 **lat (nsec/usec/msec)**
3874 The distribution of I/O completion latencies. This is the time from when
3875 I/O leaves fio and when it gets completed. Unlike the separate
3876 read/write/trim sections above, the data here and in the remaining
3877 sections apply to all I/Os for the reporting group. 250=0.04% means that
3878 0.04% of the I/Os completed in under 250us. 500=64.11% means that 64.11%
3879 of the I/Os required 250 to 499us for completion.
3882 CPU usage. User and system time, along with the number of context
3883 switches this thread went through, usage of system and user time, and
3884 finally the number of major and minor page faults. The CPU utilization
3885 numbers are averages for the jobs in that reporting group, while the
3886 context and fault counters are summed.
3889 The distribution of I/O depths over the job lifetime. The numbers are
3890 divided into powers of 2 and each entry covers depths from that value
3891 up to those that are lower than the next entry -- e.g., 16= covers
3892 depths from 16 to 31. Note that the range covered by a depth
3893 distribution entry can be different to the range covered by the
3894 equivalent submit/complete distribution entry.
3897 How many pieces of I/O were submitting in a single submit call. Each
3898 entry denotes that amount and below, until the previous entry -- e.g.,
3899 16=100% means that we submitted anywhere between 9 to 16 I/Os per submit
3900 call. Note that the range covered by a submit distribution entry can
3901 be different to the range covered by the equivalent depth distribution
3905 Like the above submit number, but for completions instead.
3908 The number of read/write/trim requests issued, and how many of them were
3912 These values are for :option:`latency_target` and related options. When
3913 these options are engaged, this section describes the I/O depth required
3914 to meet the specified latency target.
3917 Example output was based on the following:
3918 TZ=UTC fio --ioengine=null --iodepth=2 --size=100M --numjobs=2 \
3919 --rate_process=poisson --io_limit=32M --name=read --bs=128k \
3920 --rate=11M --name=write --rw=write --bs=2k --rate=700k
3922 After each client has been listed, the group statistics are printed. They
3923 will look like this::
3925 Run status group 0 (all jobs):
3926 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
3927 WRITE: bw=1231KiB/s (1261kB/s), 616KiB/s-621KiB/s (630kB/s-636kB/s), io=64.0MiB (67.1MB), run=52747-53223msec
3929 For each data direction it prints:
3932 Aggregate bandwidth of threads in this group followed by the
3933 minimum and maximum bandwidth of all the threads in this group.
3934 Values outside of brackets are power-of-2 format and those
3935 within are the equivalent value in a power-of-10 format.
3937 Aggregate I/O performed of all threads in this group. The
3938 format is the same as bw.
3940 The smallest and longest runtimes of the threads in this group.
3942 And finally, the disk statistics are printed. This is Linux specific. They will look like this::
3944 Disk stats (read/write):
3945 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
3947 Each value is printed for both reads and writes, with reads first. The
3951 Number of I/Os performed by all groups.
3953 Number of merges performed by the I/O scheduler.
3955 Number of ticks we kept the disk busy.
3957 Total time spent in the disk queue.
3959 The disk utilization. A value of 100% means we kept the disk
3960 busy constantly, 50% would be a disk idling half of the time.
3962 It is also possible to get fio to dump the current output while it is running,
3963 without terminating the job. To do that, send fio the **USR1** signal. You can
3964 also get regularly timed dumps by using the :option:`--status-interval`
3965 parameter, or by creating a file in :file:`/tmp` named
3966 :file:`fio-dump-status`. If fio sees this file, it will unlink it and dump the
3967 current output status.
3973 For scripted usage where you typically want to generate tables or graphs of the
3974 results, fio can output the results in a semicolon separated format. The format
3975 is one long line of values, such as::
3977 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%
3978 A description of this job goes here.
3980 The job description (if provided) follows on a second line for terse v2.
3981 It appears on the same line for other terse versions.
3983 To enable terse output, use the :option:`--minimal` or
3984 :option:`--output-format`\=terse command line options. The
3985 first value is the version of the terse output format. If the output has to be
3986 changed for some reason, this number will be incremented by 1 to signify that
3989 Split up, the format is as follows (comments in brackets denote when a
3990 field was introduced or whether it's specific to some terse version):
3994 terse version, fio version [v3], jobname, groupid, error
3998 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3999 Submission latency: min, max, mean, stdev (usec)
4000 Completion latency: min, max, mean, stdev (usec)
4001 Completion latency percentiles: 20 fields (see below)
4002 Total latency: min, max, mean, stdev (usec)
4003 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
4004 IOPS [v5]: min, max, mean, stdev, number of samples
4010 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
4011 Submission latency: min, max, mean, stdev (usec)
4012 Completion latency: min, max, mean, stdev (usec)
4013 Completion latency percentiles: 20 fields (see below)
4014 Total latency: min, max, mean, stdev (usec)
4015 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
4016 IOPS [v5]: min, max, mean, stdev, number of samples
4018 TRIM status [all but version 3]:
4020 Fields are similar to READ/WRITE status.
4024 user, system, context switches, major faults, minor faults
4028 <=1, 2, 4, 8, 16, 32, >=64
4030 I/O latencies microseconds::
4032 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
4034 I/O latencies milliseconds::
4036 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
4038 Disk utilization [v3]::
4040 disk name, read ios, write ios, read merges, write merges, read ticks, write ticks,
4041 time spent in queue, disk utilization percentage
4043 Additional Info (dependent on continue_on_error, default off)::
4045 total # errors, first error code
4047 Additional Info (dependent on description being set)::
4051 Completion latency percentiles can be a grouping of up to 20 sets, so for the
4052 terse output fio writes all of them. Each field will look like this::
4056 which is the Xth percentile, and the `usec` latency associated with it.
4058 For `Disk utilization`, all disks used by fio are shown. So for each disk there
4059 will be a disk utilization section.
4061 Below is a single line containing short names for each of the fields in the
4062 minimal output v3, separated by semicolons::
4064 terse_version_3;fio_version;jobname;groupid;error;read_kb;read_bandwidth_kb;read_iops;read_runtime_ms;read_slat_min_us;read_slat_max_us;read_slat_mean_us;read_slat_dev_us;read_clat_min_us;read_clat_max_us;read_clat_mean_us;read_clat_dev_us;read_clat_pct01;read_clat_pct02;read_clat_pct03;read_clat_pct04;read_clat_pct05;read_clat_pct06;read_clat_pct07;read_clat_pct08;read_clat_pct09;read_clat_pct10;read_clat_pct11;read_clat_pct12;read_clat_pct13;read_clat_pct14;read_clat_pct15;read_clat_pct16;read_clat_pct17;read_clat_pct18;read_clat_pct19;read_clat_pct20;read_tlat_min_us;read_lat_max_us;read_lat_mean_us;read_lat_dev_us;read_bw_min_kb;read_bw_max_kb;read_bw_agg_pct;read_bw_mean_kb;read_bw_dev_kb;write_kb;write_bandwidth_kb;write_iops;write_runtime_ms;write_slat_min_us;write_slat_max_us;write_slat_mean_us;write_slat_dev_us;write_clat_min_us;write_clat_max_us;write_clat_mean_us;write_clat_dev_us;write_clat_pct01;write_clat_pct02;write_clat_pct03;write_clat_pct04;write_clat_pct05;write_clat_pct06;write_clat_pct07;write_clat_pct08;write_clat_pct09;write_clat_pct10;write_clat_pct11;write_clat_pct12;write_clat_pct13;write_clat_pct14;write_clat_pct15;write_clat_pct16;write_clat_pct17;write_clat_pct18;write_clat_pct19;write_clat_pct20;write_tlat_min_us;write_lat_max_us;write_lat_mean_us;write_lat_dev_us;write_bw_min_kb;write_bw_max_kb;write_bw_agg_pct;write_bw_mean_kb;write_bw_dev_kb;cpu_user;cpu_sys;cpu_csw;cpu_mjf;cpu_minf;iodepth_1;iodepth_2;iodepth_4;iodepth_8;iodepth_16;iodepth_32;iodepth_64;lat_2us;lat_4us;lat_10us;lat_20us;lat_50us;lat_100us;lat_250us;lat_500us;lat_750us;lat_1000us;lat_2ms;lat_4ms;lat_10ms;lat_20ms;lat_50ms;lat_100ms;lat_250ms;lat_500ms;lat_750ms;lat_1000ms;lat_2000ms;lat_over_2000ms;disk_name;disk_read_iops;disk_write_iops;disk_read_merges;disk_write_merges;disk_read_ticks;write_ticks;disk_queue_time;disk_util
4066 In client/server mode terse output differs from what appears when jobs are run
4067 locally. Disk utilization data is omitted from the standard terse output and
4068 for v3 and later appears on its own separate line at the end of each terse
4075 The `json` output format is intended to be both human readable and convenient
4076 for automated parsing. For the most part its sections mirror those of the
4077 `normal` output. The `runtime` value is reported in msec and the `bw` value is
4078 reported in 1024 bytes per second units.
4084 The `json+` output format is identical to the `json` output format except that it
4085 adds a full dump of the completion latency bins. Each `bins` object contains a
4086 set of (key, value) pairs where keys are latency durations and values count how
4087 many I/Os had completion latencies of the corresponding duration. For example,
4090 "bins" : { "87552" : 1, "89600" : 1, "94720" : 1, "96768" : 1, "97792" : 1, "99840" : 1, "100864" : 2, "103936" : 6, "104960" : 534, "105984" : 5995, "107008" : 7529, ... }
4092 This data indicates that one I/O required 87,552ns to complete, two I/Os required
4093 100,864ns to complete, and 7529 I/Os required 107,008ns to complete.
4095 Also included with fio is a Python script `fio_jsonplus_clat2csv` that takes
4096 json+ output and generates CSV-formatted latency data suitable for plotting.
4098 The latency durations actually represent the midpoints of latency intervals.
4099 For details refer to :file:`stat.h`.
4105 There are two trace file format that you can encounter. The older (v1) format is
4106 unsupported since version 1.20-rc3 (March 2008). It will still be described
4107 below in case that you get an old trace and want to understand it.
4109 In any case the trace is a simple text file with a single action per line.
4112 Trace file format v1
4113 ~~~~~~~~~~~~~~~~~~~~
4115 Each line represents a single I/O action in the following format::
4119 where `rw=0/1` for read/write, and the `offset` and `length` entries being in bytes.
4121 This format is not supported in fio versions >= 1.20-rc3.
4124 Trace file format v2
4125 ~~~~~~~~~~~~~~~~~~~~
4127 The second version of the trace file format was added in fio version 1.17. It
4128 allows to access more then one file per trace and has a bigger set of possible
4131 The first line of the trace file has to be::
4135 Following this can be lines in two different formats, which are described below.
4137 The file management format::
4141 The `filename` is given as an absolute path. The `action` can be one of these:
4144 Add the given `filename` to the trace.
4146 Open the file with the given `filename`. The `filename` has to have
4147 been added with the **add** action before.
4149 Close the file with the given `filename`. The file has to have been
4153 The file I/O action format::
4155 filename action offset length
4157 The `filename` is given as an absolute path, and has to have been added and
4158 opened before it can be used with this format. The `offset` and `length` are
4159 given in bytes. The `action` can be one of these:
4162 Wait for `offset` microseconds. Everything below 100 is discarded.
4163 The time is relative to the previous `wait` statement.
4165 Read `length` bytes beginning from `offset`.
4167 Write `length` bytes beginning from `offset`.
4169 :manpage:`fsync(2)` the file.
4171 :manpage:`fdatasync(2)` the file.
4173 Trim the given file from the given `offset` for `length` bytes.
4176 I/O Replay - Merging Traces
4177 ---------------------------
4179 Colocation is a common practice used to get the most out of a machine.
4180 Knowing which workloads play nicely with each other and which ones don't is
4181 a much harder task. While fio can replay workloads concurrently via multiple
4182 jobs, it leaves some variability up to the scheduler making results harder to
4183 reproduce. Merging is a way to make the order of events consistent.
4185 Merging is integrated into I/O replay and done when a
4186 :option:`merge_blktrace_file` is specified. The list of files passed to
4187 :option:`read_iolog` go through the merge process and output a single file
4188 stored to the specified file. The output file is passed on as if it were the
4189 only file passed to :option:`read_iolog`. An example would look like::
4191 $ fio --read_iolog="<file1>:<file2>" --merge_blktrace_file="<output_file>"
4193 Creating only the merged file can be done by passing the command line argument
4194 :option:`--merge-blktrace-only`.
4196 Scaling traces can be done to see the relative impact of any particular trace
4197 being slowed down or sped up. :option:`merge_blktrace_scalars` takes in a colon
4198 separated list of percentage scalars. It is index paired with the files passed
4199 to :option:`read_iolog`.
4201 With scaling, it may be desirable to match the running time of all traces.
4202 This can be done with :option:`merge_blktrace_iters`. It is index paired with
4203 :option:`read_iolog` just like :option:`merge_blktrace_scalars`.
4205 In an example, given two traces, A and B, each 60s long. If we want to see
4206 the impact of trace A issuing IOs twice as fast and repeat trace A over the
4207 runtime of trace B, the following can be done::
4209 $ fio --read_iolog="<trace_a>:"<trace_b>" --merge_blktrace_file"<output_file>" --merge_blktrace_scalars="50:100" --merge_blktrace_iters="2:1"
4211 This runs trace A at 2x the speed twice for approximately the same runtime as
4212 a single run of trace B.
4215 CPU idleness profiling
4216 ----------------------
4218 In some cases, we want to understand CPU overhead in a test. For example, we
4219 test patches for the specific goodness of whether they reduce CPU usage.
4220 Fio implements a balloon approach to create a thread per CPU that runs at idle
4221 priority, meaning that it only runs when nobody else needs the cpu.
4222 By measuring the amount of work completed by the thread, idleness of each CPU
4223 can be derived accordingly.
4225 An unit work is defined as touching a full page of unsigned characters. Mean and
4226 standard deviation of time to complete an unit work is reported in "unit work"
4227 section. Options can be chosen to report detailed percpu idleness or overall
4228 system idleness by aggregating percpu stats.
4231 Verification and triggers
4232 -------------------------
4234 Fio is usually run in one of two ways, when data verification is done. The first
4235 is a normal write job of some sort with verify enabled. When the write phase has
4236 completed, fio switches to reads and verifies everything it wrote. The second
4237 model is running just the write phase, and then later on running the same job
4238 (but with reads instead of writes) to repeat the same I/O patterns and verify
4239 the contents. Both of these methods depend on the write phase being completed,
4240 as fio otherwise has no idea how much data was written.
4242 With verification triggers, fio supports dumping the current write state to
4243 local files. Then a subsequent read verify workload can load this state and know
4244 exactly where to stop. This is useful for testing cases where power is cut to a
4245 server in a managed fashion, for instance.
4247 A verification trigger consists of two things:
4249 1) Storing the write state of each job.
4250 2) Executing a trigger command.
4252 The write state is relatively small, on the order of hundreds of bytes to single
4253 kilobytes. It contains information on the number of completions done, the last X
4256 A trigger is invoked either through creation ('touch') of a specified file in
4257 the system, or through a timeout setting. If fio is run with
4258 :option:`--trigger-file`\= :file:`/tmp/trigger-file`, then it will continually
4259 check for the existence of :file:`/tmp/trigger-file`. When it sees this file, it
4260 will fire off the trigger (thus saving state, and executing the trigger
4263 For client/server runs, there's both a local and remote trigger. If fio is
4264 running as a server backend, it will send the job states back to the client for
4265 safe storage, then execute the remote trigger, if specified. If a local trigger
4266 is specified, the server will still send back the write state, but the client
4267 will then execute the trigger.
4269 Verification trigger example
4270 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
4272 Let's say we want to run a powercut test on the remote Linux machine 'server'.
4273 Our write workload is in :file:`write-test.fio`. We want to cut power to 'server' at
4274 some point during the run, and we'll run this test from the safety or our local
4275 machine, 'localbox'. On the server, we'll start the fio backend normally::
4277 server# fio --server
4279 and on the client, we'll fire off the workload::
4281 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger-remote="bash -c \"echo b > /proc/sysrq-triger\""
4283 We set :file:`/tmp/my-trigger` as the trigger file, and we tell fio to execute::
4285 echo b > /proc/sysrq-trigger
4287 on the server once it has received the trigger and sent us the write state. This
4288 will work, but it's not **really** cutting power to the server, it's merely
4289 abruptly rebooting it. If we have a remote way of cutting power to the server
4290 through IPMI or similar, we could do that through a local trigger command
4291 instead. Let's assume we have a script that does IPMI reboot of a given hostname,
4292 ipmi-reboot. On localbox, we could then have run fio with a local trigger
4295 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger="ipmi-reboot server"
4297 For this case, fio would wait for the server to send us the write state, then
4298 execute ``ipmi-reboot server`` when that happened.
4300 Loading verify state
4301 ~~~~~~~~~~~~~~~~~~~~
4303 To load stored write state, a read verification job file must contain the
4304 :option:`verify_state_load` option. If that is set, fio will load the previously
4305 stored state. For a local fio run this is done by loading the files directly,
4306 and on a client/server run, the server backend will ask the client to send the
4307 files over and load them from there.
4313 Fio supports a variety of log file formats, for logging latencies, bandwidth,
4314 and IOPS. The logs share a common format, which looks like this:
4316 *time* (`msec`), *value*, *data direction*, *block size* (`bytes`),
4317 *offset* (`bytes`), *command priority*
4319 *Time* for the log entry is always in milliseconds. The *value* logged depends
4320 on the type of log, it will be one of the following:
4323 Value is latency in nsecs
4329 *Data direction* is one of the following:
4338 The entry's *block size* is always in bytes. The *offset* is the position in bytes
4339 from the start of the file for that particular I/O. The logging of the offset can be
4340 toggled with :option:`log_offset`.
4342 *Command priority* is 0 for normal priority and 1 for high priority. This is controlled
4343 by the ioengine specific :option:`cmdprio_percentage`.
4345 Fio defaults to logging every individual I/O but when windowed logging is set
4346 through :option:`log_avg_msec`, either the average (by default) or the maximum
4347 (:option:`log_max_value` is set) *value* seen over the specified period of time
4348 is recorded. Each *data direction* seen within the window period will aggregate
4349 its values in a separate row. Further, when using windowed logging the *block
4350 size* and *offset* entries will always contain 0.
4356 Normally fio is invoked as a stand-alone application on the machine where the
4357 I/O workload should be generated. However, the backend and frontend of fio can
4358 be run separately i.e., the fio server can generate an I/O workload on the "Device
4359 Under Test" while being controlled by a client on another machine.
4361 Start the server on the machine which has access to the storage DUT::
4365 where `args` defines what fio listens to. The arguments are of the form
4366 ``type,hostname`` or ``IP,port``. *type* is either ``ip`` (or ip4) for TCP/IP
4367 v4, ``ip6`` for TCP/IP v6, or ``sock`` for a local unix domain socket.
4368 *hostname* is either a hostname or IP address, and *port* is the port to listen
4369 to (only valid for TCP/IP, not a local socket). Some examples:
4373 Start a fio server, listening on all interfaces on the default port (8765).
4375 2) ``fio --server=ip:hostname,4444``
4377 Start a fio server, listening on IP belonging to hostname and on port 4444.
4379 3) ``fio --server=ip6:::1,4444``
4381 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
4383 4) ``fio --server=,4444``
4385 Start a fio server, listening on all interfaces on port 4444.
4387 5) ``fio --server=1.2.3.4``
4389 Start a fio server, listening on IP 1.2.3.4 on the default port.
4391 6) ``fio --server=sock:/tmp/fio.sock``
4393 Start a fio server, listening on the local socket :file:`/tmp/fio.sock`.
4395 Once a server is running, a "client" can connect to the fio server with::
4397 fio <local-args> --client=<server> <remote-args> <job file(s)>
4399 where `local-args` are arguments for the client where it is running, `server`
4400 is the connect string, and `remote-args` and `job file(s)` are sent to the
4401 server. The `server` string follows the same format as it does on the server
4402 side, to allow IP/hostname/socket and port strings.
4404 Fio can connect to multiple servers this way::
4406 fio --client=<server1> <job file(s)> --client=<server2> <job file(s)>
4408 If the job file is located on the fio server, then you can tell the server to
4409 load a local file as well. This is done by using :option:`--remote-config` ::
4411 fio --client=server --remote-config /path/to/file.fio
4413 Then fio will open this local (to the server) job file instead of being passed
4414 one from the client.
4416 If you have many servers (example: 100 VMs/containers), you can input a pathname
4417 of a file containing host IPs/names as the parameter value for the
4418 :option:`--client` option. For example, here is an example :file:`host.list`
4419 file containing 2 hostnames::
4421 host1.your.dns.domain
4422 host2.your.dns.domain
4424 The fio command would then be::
4426 fio --client=host.list <job file(s)>
4428 In this mode, you cannot input server-specific parameters or job files -- all
4429 servers receive the same job file.
4431 In order to let ``fio --client`` runs use a shared filesystem from multiple
4432 hosts, ``fio --client`` now prepends the IP address of the server to the
4433 filename. For example, if fio is using the directory :file:`/mnt/nfs/fio` and is
4434 writing filename :file:`fileio.tmp`, with a :option:`--client` `hostfile`
4435 containing two hostnames ``h1`` and ``h2`` with IP addresses 192.168.10.120 and
4436 192.168.10.121, then fio will create two files::
4438 /mnt/nfs/fio/192.168.10.120.fileio.tmp
4439 /mnt/nfs/fio/192.168.10.121.fileio.tmp
4441 Terse output in client/server mode will differ slightly from what is produced
4442 when fio is run in stand-alone mode. See the terse output section for details.