4 The first step in getting fio to simulate a desired I/O workload, is writing a
5 job file describing that specific setup. A job file may contain any number of
6 threads and/or files -- the typical contents of the job file is a *global*
7 section defining shared parameters, and one or more job sections describing the
8 jobs involved. When run, fio parses this file and sets everything up as
9 described. If we break down a job from top to bottom, it contains the following
14 Defines the I/O pattern issued to the file(s). We may only be reading
15 sequentially from this file(s), or we may be writing randomly. Or even
16 mixing reads and writes, sequentially or randomly.
17 Should we be doing buffered I/O, or direct/raw I/O?
21 In how large chunks are we issuing I/O? This may be a single value,
22 or it may describe a range of block sizes.
26 How much data are we going to be reading/writing.
30 How do we issue I/O? We could be memory mapping the file, we could be
31 using regular read/write, we could be using splice, async I/O, or even
36 If the I/O engine is async, how large a queuing depth do we want to
42 How many files are we spreading the workload over.
44 `Threads, processes and job synchronization`_
46 How many threads or processes should we spread this workload over.
48 The above are the basic parameters defined for a workload, in addition there's a
49 multitude of parameters that modify other aspects of how this job behaves.
55 .. option:: --debug=type
57 Enable verbose tracing of various fio actions. May be ``all`` for all types
58 or individual types separated by a comma (e.g. ``--debug=file,mem`` will
59 enable file and memory debugging). Currently, additional logging is
63 Dump info related to processes.
65 Dump info related to file actions.
67 Dump info related to I/O queuing.
69 Dump info related to memory allocations.
71 Dump info related to blktrace setup.
73 Dump info related to I/O verification.
75 Enable all debug options.
77 Dump info related to random offset generation.
79 Dump info related to option matching and parsing.
81 Dump info related to disk utilization updates.
83 Dump info only related to job number x.
85 Dump info only related to mutex up/down ops.
87 Dump info related to profile extensions.
89 Dump info related to internal time keeping.
91 Dump info related to networking connections.
93 Dump info related to I/O rate switching.
95 Dump info related to log compress/decompress.
97 Show available debug options.
99 .. option:: --parse-only
101 Parse options only, don\'t start any I/O.
103 .. option:: --output=filename
105 Write output to file `filename`.
107 .. option:: --bandwidth-log
109 Generate aggregate bandwidth logs.
111 .. option:: --minimal
113 Print statistics in a terse, semicolon-delimited format.
115 .. option:: --append-terse
117 Print statistics in selected mode AND terse, semicolon-delimited format.
118 **deprecated**, use :option:`--output-format` instead to select multiple
121 .. option:: --output-format=type
123 Set the reporting format to `normal`, `terse`, `json`, or `json+`. Multiple
124 formats can be selected, separate by a comma. `terse` is a CSV based
125 format. `json+` is like `json`, except it adds a full dump of the latency
128 .. option:: --terse-version=type
130 Set terse version output format (default 3, or 2 or 4).
132 .. option:: --version
134 Print version info and exit.
140 .. option:: --cpuclock-test
142 Perform test and validation of internal CPU clock.
144 .. option:: --crctest=test
146 Test the speed of the builtin checksumming functions. If no argument is
147 given, all of them are tested. Or a comma separated list can be passed, in
148 which case the given ones are tested.
150 .. option:: --cmdhelp=command
152 Print help information for `command`. May be ``all`` for all commands.
154 .. option:: --enghelp=[ioengine[,command]]
156 List all commands defined by :option:`ioengine`, or print help for `command`
157 defined by :option:`ioengine`. If no :option:`ioengine` is given, list all
160 .. option:: --showcmd=jobfile
162 Turn a job file into command line options.
164 .. option:: --readonly
166 Turn on safety read-only checks, preventing writes. The ``--readonly``
167 option is an extra safety guard to prevent users from accidentally starting
168 a write workload when that is not desired. Fio will only write if
169 `rw=write/randwrite/rw/randrw` is given. This extra safety net can be used
170 as an extra precaution as ``--readonly`` will also enable a write check in
171 the I/O engine core to prevent writes due to unknown user space bug(s).
173 .. option:: --eta=when
175 When real-time ETA estimate should be printed. May be `always`, `never` or
178 .. option:: --eta-newline=time
180 Force a new line for every `time` period passed.
182 .. option:: --status-interval=time
184 Force full status dump every `time` period passed.
186 .. option:: --section=name
188 Only run specified section in job file. Multiple sections can be specified.
189 The ``--section`` option allows one to combine related jobs into one file.
190 E.g. one job file could define light, moderate, and heavy sections. Tell
191 fio to run only the "heavy" section by giving ``--section=heavy``
192 command line option. One can also specify the "write" operations in one
193 section and "verify" operation in another section. The ``--section`` option
194 only applies to job sections. The reserved *global* section is always
197 .. option:: --alloc-size=kb
199 Set the internal smalloc pool to this size in kb (def 1024). The
200 ``--alloc-size`` switch allows one to use a larger pool size for smalloc.
201 If running large jobs with randommap enabled, fio can run out of memory.
202 Smalloc is an internal allocator for shared structures from a fixed size
203 memory pool. The pool size defaults to 16M and can grow to 8 pools.
205 NOTE: While running :file:`.fio_smalloc.*` backing store files are visible
208 .. option:: --warnings-fatal
210 All fio parser warnings are fatal, causing fio to exit with an
213 .. option:: --max-jobs=nr
215 Maximum number of threads/processes to support.
217 .. option:: --server=args
219 Start a backend server, with `args` specifying what to listen to.
220 See `Client/Server`_ section.
222 .. option:: --daemonize=pidfile
224 Background a fio server, writing the pid to the given `pidfile` file.
226 .. option:: --client=hostname
228 Instead of running the jobs locally, send and run them on the given host or
229 set of hosts. See `Client/Server`_ section.
231 .. option:: --remote-config=file
233 Tell fio server to load this local file.
235 .. option:: --idle-prof=option
237 Report cpu idleness on a system or percpu basis
238 ``--idle-prof=system,percpu`` or
239 run unit work calibration only ``--idle-prof=calibrate``.
241 .. option:: --inflate-log=log
243 Inflate and output compressed log.
245 .. option:: --trigger-file=file
247 Execute trigger cmd when file exists.
249 .. option:: --trigger-timeout=t
251 Execute trigger at this time.
253 .. option:: --trigger=cmd
255 Set this command as local trigger.
257 .. option:: --trigger-remote=cmd
259 Set this command as remote trigger.
261 .. option:: --aux-path=path
263 Use this path for fio state generated files.
265 Any parameters following the options will be assumed to be job files, unless
266 they match a job file parameter. Multiple job files can be listed and each job
267 file will be regarded as a separate group. Fio will :option:`stonewall`
268 execution between each group.
274 As previously described, fio accepts one or more job files describing what it is
275 supposed to do. The job file format is the classic ini file, where the names
276 enclosed in [] brackets define the job name. You are free to use any ASCII name
277 you want, except *global* which has special meaning. Following the job name is
278 a sequence of zero or more parameters, one per line, that define the behavior of
279 the job. If the first character in a line is a ';' or a '#', the entire line is
280 discarded as a comment.
282 A *global* section sets defaults for the jobs described in that file. A job may
283 override a *global* section parameter, and a job file may even have several
284 *global* sections if so desired. A job is only affected by a *global* section
287 The :option:`--cmdhelp` option also lists all options. If used with an `option`
288 argument, :option:`--cmdhelp` will detail the given `option`.
290 See the `examples/` directory for inspiration on how to write job files. Note
291 the copyright and license requirements currently apply to `examples/` files.
293 So let's look at a really simple job file that defines two processes, each
294 randomly reading from a 128MiB file:
298 ; -- start job file --
309 As you can see, the job file sections themselves are empty as all the described
310 parameters are shared. As no :option:`filename` option is given, fio makes up a
311 `filename` for each of the jobs as it sees fit. On the command line, this job
312 would look as follows::
314 $ fio --name=global --rw=randread --size=128m --name=job1 --name=job2
317 Let's look at an example that has a number of processes writing randomly to
322 ; -- start job file --
333 Here we have no *global* section, as we only have one job defined anyway. We
334 want to use async I/O here, with a depth of 4 for each file. We also increased
335 the buffer size used to 32KiB and define numjobs to 4 to fork 4 identical
336 jobs. The result is 4 processes each randomly writing to their own 64MiB
337 file. Instead of using the above job file, you could have given the parameters
338 on the command line. For this case, you would specify::
340 $ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4
342 When fio is utilized as a basis of any reasonably large test suite, it might be
343 desirable to share a set of standardized settings across multiple job files.
344 Instead of copy/pasting such settings, any section may pull in an external
345 :file:`filename.fio` file with *include filename* directive, as in the following
348 ; -- start job file including.fio --
352 include glob-include.fio
359 include test-include.fio
360 ; -- end job file including.fio --
364 ; -- start job file glob-include.fio --
367 ; -- end job file glob-include.fio --
371 ; -- start job file test-include.fio --
374 ; -- end job file test-include.fio --
376 Settings pulled into a section apply to that section only (except *global*
377 section). Include directives may be nested in that any included file may contain
378 further include directive(s). Include files may not contain [] sections.
381 Environment variables
382 ~~~~~~~~~~~~~~~~~~~~~
384 Fio also supports environment variable expansion in job files. Any sub-string of
385 the form ``${VARNAME}`` as part of an option value (in other words, on the right
386 of the '='), will be expanded to the value of the environment variable called
387 `VARNAME`. If no such environment variable is defined, or `VARNAME` is the
388 empty string, the empty string will be substituted.
390 As an example, let's look at a sample fio invocation and job file::
392 $ SIZE=64m NUMJOBS=4 fio jobfile.fio
396 ; -- start job file --
403 This will expand to the following equivalent job file at runtime:
407 ; -- start job file --
414 Fio ships with a few example job files, you can also look there for inspiration.
419 Additionally, fio has a set of reserved keywords that will be replaced
420 internally with the appropriate value. Those keywords are:
424 The architecture page size of the running system.
428 Megabytes of total memory in the system.
432 Number of online available CPUs.
434 These can be used on the command line or in the job file, and will be
435 automatically substituted with the current system values when the job is
436 run. Simple math is also supported on these keywords, so you can perform actions
441 and get that properly expanded to 8 times the size of memory in the machine.
447 This section describes in details each parameter associated with a job. Some
448 parameters take an option of a given type, such as an integer or a
449 string. Anywhere a numeric value is required, an arithmetic expression may be
450 used, provided it is surrounded by parentheses. Supported operators are:
459 For time values in expressions, units are microseconds by default. This is
460 different than for time values not in expressions (not enclosed in
461 parentheses). The following types are used:
468 String. This is a sequence of alpha characters.
471 Integer with possible time suffix. In seconds unless otherwise
472 specified, use e.g. 10m for 10 minutes. Accepts s/m/h for seconds, minutes,
473 and hours, and accepts 'ms' (or 'msec') for milliseconds, and 'us' (or
474 'usec') for microseconds.
479 Integer. A whole number value, which may contain an integer prefix
480 and an integer suffix:
482 [*integer prefix*] **number** [*integer suffix*]
484 The optional *integer prefix* specifies the number's base. The default
485 is decimal. *0x* specifies hexadecimal.
487 The optional *integer suffix* specifies the number's units, and includes an
488 optional unit prefix and an optional unit. For quantities of data, the
489 default unit is bytes. For quantities of time, the default unit is seconds.
491 With :option:`kb_base` =1000, fio follows international standards for unit
492 prefixes. To specify power-of-10 decimal values defined in the
493 International System of Units (SI):
495 * *Ki* -- means kilo (K) or 1000
496 * *Mi* -- means mega (M) or 1000**2
497 * *Gi* -- means giga (G) or 1000**3
498 * *Ti* -- means tera (T) or 1000**4
499 * *Pi* -- means peta (P) or 1000**5
501 To specify power-of-2 binary values defined in IEC 80000-13:
503 * *k* -- means kibi (Ki) or 1024
504 * *M* -- means mebi (Mi) or 1024**2
505 * *G* -- means gibi (Gi) or 1024**3
506 * *T* -- means tebi (Ti) or 1024**4
507 * *P* -- means pebi (Pi) or 1024**5
509 With :option:`kb_base` =1024 (the default), the unit prefixes are opposite
510 from those specified in the SI and IEC 80000-13 standards to provide
511 compatibility with old scripts. For example, 4k means 4096.
513 For quantities of data, an optional unit of 'B' may be included
514 (e.g., 'kB' is the same as 'k').
516 The *integer suffix* is not case sensitive (e.g., m/mi mean mebi/mega,
517 not milli). 'b' and 'B' both mean byte, not bit.
519 Examples with :option:`kb_base` =1000:
521 * *4 KiB*: 4096, 4096b, 4096B, 4ki, 4kib, 4kiB, 4Ki, 4KiB
522 * *1 MiB*: 1048576, 1mi, 1024ki
523 * *1 MB*: 1000000, 1m, 1000k
524 * *1 TiB*: 1099511627776, 1ti, 1024gi, 1048576mi
525 * *1 TB*: 1000000000, 1t, 1000m, 1000000k
527 Examples with :option:`kb_base` =1024 (default):
529 * *4 KiB*: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB
530 * *1 MiB*: 1048576, 1m, 1024k
531 * *1 MB*: 1000000, 1mi, 1000ki
532 * *1 TiB*: 1099511627776, 1t, 1024g, 1048576m
533 * *1 TB*: 1000000000, 1ti, 1000mi, 1000000ki
535 To specify times (units are not case sensitive):
539 * *M* -- mean minutes
540 * *s* -- or sec means seconds (default)
541 * *ms* -- or *msec* means milliseconds
542 * *us* -- or *usec* means microseconds
544 If the option accepts an upper and lower range, use a colon ':' or
545 minus '-' to separate such values. See :ref:`irange <irange>`.
550 Boolean. Usually parsed as an integer, however only defined for
551 true and false (1 and 0).
556 Integer range with suffix. Allows value range to be given, such as
557 1024-4096. A colon may also be used as the separator, e.g. 1k:4k. If the
558 option allows two sets of ranges, they can be specified with a ',' or '/'
559 delimiter: 1k-4k/8k-32k. Also see :ref:`int <int>`.
562 A list of floating point numbers, separated by a ':' character.
568 .. option:: kb_base=int
570 Select the interpretation of unit prefixes in input parameters.
573 Inputs comply with IEC 80000-13 and the International
574 System of Units (SI). Use:
576 - power-of-2 values with IEC prefixes (e.g., KiB)
577 - power-of-10 values with SI prefixes (e.g., kB)
580 Compatibility mode (default). To avoid breaking old scripts:
582 - power-of-2 values with SI prefixes
583 - power-of-10 values with IEC prefixes
585 See :option:`bs` for more details on input parameters.
587 Outputs always use correct prefixes. Most outputs include both
590 bw=2383.3kB/s (2327.4KiB/s)
592 If only one value is reported, then kb_base selects the one to use:
594 **1000** -- SI prefixes
596 **1024** -- IEC prefixes
598 .. option:: unit_base=int
600 Base unit for reporting. Allowed values are:
603 Use auto-detection (default).
610 With the above in mind, here follows the complete list of fio job parameters.
618 ASCII name of the job. This may be used to override the name printed by fio
619 for this job. Otherwise the job name is used. On the command line this
620 parameter has the special purpose of also signaling the start of a new job.
622 .. option:: description=str
624 Text description of the job. Doesn't do anything except dump this text
625 description when this job is run. It's not parsed.
627 .. option:: loops=int
629 Run the specified number of iterations of this job. Used to repeat the same
630 workload a given number of times. Defaults to 1.
632 .. option:: numjobs=int
634 Create the specified number of clones of this job. May be used to setup a
635 larger number of threads/processes doing the same thing. Each thread is
636 reported separately; to see statistics for all clones as a whole, use
637 :option:`group_reporting` in conjunction with :option:`new_group`.
638 See :option:`--max-jobs`.
641 Time related parameters
642 ~~~~~~~~~~~~~~~~~~~~~~~
644 .. option:: runtime=time
646 Tell fio to terminate processing after the specified period of time. It
647 can be quite hard to determine for how long a specified job will run, so
648 this parameter is handy to cap the total runtime to a given time. When
649 the unit is omitted, the value is given in seconds.
651 .. option:: time_based
653 If set, fio will run for the duration of the :option:`runtime` specified
654 even if the file(s) are completely read or written. It will simply loop over
655 the same workload as many times as the :option:`runtime` allows.
657 .. option:: startdelay=irange(time)
659 Delay start of job for the specified number of seconds. Supports all time
660 suffixes to allow specification of hours, minutes, seconds and milliseconds
661 -- seconds are the default if a unit is omitted. Can be given as a range
662 which causes each thread to choose randomly out of the range.
664 .. option:: ramp_time=time
666 If set, fio will run the specified workload for this amount of time before
667 logging any performance numbers. Useful for letting performance settle
668 before logging results, thus minimizing the runtime required for stable
669 results. Note that the ``ramp_time`` is considered lead in time for a job,
670 thus it will increase the total runtime if a special timeout or
671 :option:`runtime` is specified. When the unit is omitted, the value is
674 .. option:: clocksource=str
676 Use the given clocksource as the base of timing. The supported options are:
679 :manpage:`gettimeofday(2)`
682 :manpage:`clock_gettime(2)`
685 Internal CPU clock source
687 cpu is the preferred clocksource if it is reliable, as it is very fast (and
688 fio is heavy on time calls). Fio will automatically use this clocksource if
689 it's supported and considered reliable on the system it is running on,
690 unless another clocksource is specifically set. For x86/x86-64 CPUs, this
691 means supporting TSC Invariant.
693 .. option:: gtod_reduce=bool
695 Enable all of the :manpage:`gettimeofday(2)` reducing options
696 (:option:`disable_clat`, :option:`disable_slat`, :option:`disable_bw_measurement`) plus
697 reduce precision of the timeout somewhat to really shrink the
698 :manpage:`gettimeofday(2)` call count. With this option enabled, we only do
699 about 0.4% of the :manpage:`gettimeofday(2)` calls we would have done if all
700 time keeping was enabled.
702 .. option:: gtod_cpu=int
704 Sometimes it's cheaper to dedicate a single thread of execution to just
705 getting the current time. Fio (and databases, for instance) are very
706 intensive on :manpage:`gettimeofday(2)` calls. With this option, you can set
707 one CPU aside for doing nothing but logging current time to a shared memory
708 location. Then the other threads/processes that run I/O workloads need only
709 copy that segment, instead of entering the kernel with a
710 :manpage:`gettimeofday(2)` call. The CPU set aside for doing these time
711 calls will be excluded from other uses. Fio will manually clear it from the
712 CPU mask of other jobs.
718 .. option:: directory=str
720 Prefix filenames with this directory. Used to place files in a different
721 location than :file:`./`. You can specify a number of directories by
722 separating the names with a ':' character. These directories will be
723 assigned equally distributed to job clones creates with :option:`numjobs` as
724 long as they are using generated filenames. If specific `filename(s)` are
725 set fio will use the first listed directory, and thereby matching the
726 `filename` semantic which generates a file each clone if not specified, but
727 let all clones use the same if set.
729 See the :option:`filename` option for escaping certain characters.
731 .. option:: filename=str
733 Fio normally makes up a `filename` based on the job name, thread number, and
734 file number. If you want to share files between threads in a job or several
735 jobs, specify a `filename` for each of them to override the default. If the
736 ioengine is file based, you can specify a number of files by separating the
737 names with a ':' colon. So if you wanted a job to open :file:`/dev/sda` and
738 :file:`/dev/sdb` as the two working files, you would use
739 ``filename=/dev/sda:/dev/sdb``.
740 On Windows, disk devices are accessed as :file:`\\\\.\\PhysicalDrive0` for
741 the first device, :file:`\\\\.\\PhysicalDrive1` for the second etc.
742 Note: Windows and FreeBSD prevent write access to areas
743 of the disk containing in-use data (e.g. filesystems). If the wanted
744 `filename` does need to include a colon, then escape that with a ``\``
745 character. For instance, if the `filename` is :file:`/dev/dsk/foo@3,0:c`,
746 then you would use ``filename="/dev/dsk/foo@3,0\:c"``. The
747 :file:`-` is a reserved name, meaning stdin or stdout. Which of the two
748 depends on the read/write direction set.
750 .. option:: filename_format=str
752 If sharing multiple files between jobs, it is usually necessary to have fio
753 generate the exact names that you want. By default, fio will name a file
754 based on the default file format specification of
755 :file:`jobname.jobnumber.filenumber`. With this option, that can be
756 customized. Fio will recognize and replace the following keywords in this
760 The name of the worker thread or process.
762 The incremental number of the worker thread or process.
764 The incremental number of the file for that worker thread or
767 To have dependent jobs share a set of files, this option can be set to have
768 fio generate filenames that are shared between the two. For instance, if
769 :file:`testfiles.$filenum` is specified, file number 4 for any job will be
770 named :file:`testfiles.4`. The default of :file:`$jobname.$jobnum.$filenum`
771 will be used if no other format specifier is given.
773 .. option:: unique_filename=bool
775 To avoid collisions between networked clients, fio defaults to prefixing any
776 generated filenames (with a directory specified) with the source of the
777 client connecting. To disable this behavior, set this option to 0.
779 .. option:: opendir=str
781 Recursively open any files below directory `str`.
783 .. option:: lockfile=str
785 Fio defaults to not locking any files before it does I/O to them. If a file
786 or file descriptor is shared, fio can serialize I/O to that file to make the
787 end result consistent. This is usual for emulating real workloads that share
788 files. The lock modes are:
791 No locking. The default.
793 Only one thread or process may do I/O at a time, excluding all
796 Read-write locking on the file. Many readers may
797 access the file at the same time, but writes get exclusive access.
799 .. option:: nrfiles=int
801 Number of files to use for this job. Defaults to 1.
803 .. option:: openfiles=int
805 Number of files to keep open at the same time. Defaults to the same as
806 :option:`nrfiles`, can be set smaller to limit the number simultaneous
809 .. option:: file_service_type=str
811 Defines how fio decides which file from a job to service next. The following
815 Choose a file at random.
818 Round robin over opened files. This is the default.
821 Finish one file before moving on to the next. Multiple files can
822 still be open depending on 'openfiles'.
825 Use a *Zipf* distribution to decide what file to access.
828 Use a *Pareto* distribution to decide what file to access.
831 Use a *Gaussian* (normal) distribution to decide what file to
834 For *random*, *roundrobin*, and *sequential*, a postfix can be appended to
835 tell fio how many I/Os to issue before switching to a new file. For example,
836 specifying ``file_service_type=random:8`` would cause fio to issue
837 8 I/Os before selecting a new file at random. For the non-uniform
838 distributions, a floating point postfix can be given to influence how the
839 distribution is skewed. See :option:`random_distribution` for a description
840 of how that would work.
842 .. option:: ioscheduler=str
844 Attempt to switch the device hosting the file to the specified I/O scheduler
847 .. option:: create_serialize=bool
849 If true, serialize the file creation for the jobs. This may be handy to
850 avoid interleaving of data files, which may greatly depend on the filesystem
851 used and even the number of processors in the system.
853 .. option:: create_fsync=bool
855 fsync the data file after creation. This is the default.
857 .. option:: create_on_open=bool
859 Don't pre-setup the files for I/O, just create open() when it's time to do
862 .. option:: create_only=bool
864 If true, fio will only run the setup phase of the job. If files need to be
865 laid out or updated on disk, only that will be done. The actual job contents
868 .. option:: allow_file_create=bool
870 If true, fio is permitted to create files as part of its workload. This is
871 the default behavior. If this option is false, then fio will error out if
872 the files it needs to use don't already exist. Default: true.
874 .. option:: allow_mounted_write=bool
876 If this isn't set, fio will abort jobs that are destructive (e.g. that write)
877 to what appears to be a mounted device or partition. This should help catch
878 creating inadvertently destructive tests, not realizing that the test will
879 destroy data on the mounted file system. Default: false.
881 .. option:: pre_read=bool
883 If this is given, files will be pre-read into memory before starting the
884 given I/O operation. This will also clear the :option:`invalidate` flag,
885 since it is pointless to pre-read and then drop the cache. This will only
886 work for I/O engines that are seek-able, since they allow you to read the
887 same data multiple times. Thus it will not work on e.g. network or splice I/O.
889 .. option:: unlink=bool
891 Unlink the job files when done. Not the default, as repeated runs of that
892 job would then waste time recreating the file set again and again.
894 .. option:: unlink_each_loop=bool
896 Unlink job files after each iteration or loop.
898 .. option:: zonesize=int
900 Divide a file into zones of the specified size. See :option:`zoneskip`.
902 .. option:: zonerange=int
904 Give size of an I/O zone. See :option:`zoneskip`.
906 .. option:: zoneskip=int
908 Skip the specified number of bytes when :option:`zonesize` data has been
909 read. The two zone options can be used to only do I/O on zones of a file.
915 .. option:: direct=bool
917 If value is true, use non-buffered I/O. This is usually O_DIRECT. Note that
918 ZFS on Solaris doesn't support direct I/O. On Windows the synchronous
919 ioengines don't support direct I/O. Default: false.
921 .. option:: atomic=bool
923 If value is true, attempt to use atomic direct I/O. Atomic writes are
924 guaranteed to be stable once acknowledged by the operating system. Only
925 Linux supports O_ATOMIC right now.
927 .. option:: buffered=bool
929 If value is true, use buffered I/O. This is the opposite of the
930 :option:`direct` option. Defaults to true.
932 .. option:: readwrite=str, rw=str
934 Type of I/O pattern. Accepted values are:
941 Sequential trims (Linux block devices only).
947 Random trims (Linux block devices only).
949 Sequential mixed reads and writes.
951 Random mixed reads and writes.
953 Sequential trim+write sequences. Blocks will be trimmed first,
954 then the same blocks will be written to.
956 Fio defaults to read if the option is not specified. For the mixed I/O
957 types, the default is to split them 50/50. For certain types of I/O the
958 result may still be skewed a bit, since the speed may be different. It is
959 possible to specify a number of I/O's to do before getting a new offset,
960 this is done by appending a ``:<nr>`` to the end of the string given. For a
961 random read, it would look like ``rw=randread:8`` for passing in an offset
962 modifier with a value of 8. If the suffix is used with a sequential I/O
963 pattern, then the value specified will be added to the generated offset for
964 each I/O. For instance, using ``rw=write:4k`` will skip 4k for every
965 write. It turns sequential I/O into sequential I/O with holes. See the
966 :option:`rw_sequencer` option.
968 .. option:: rw_sequencer=str
970 If an offset modifier is given by appending a number to the ``rw=<str>``
971 line, then this option controls how that number modifies the I/O offset
972 being generated. Accepted values are:
975 Generate sequential offset.
977 Generate the same offset.
979 ``sequential`` is only useful for random I/O, where fio would normally
980 generate a new random offset for every I/O. If you append e.g. 8 to randread,
981 you would get a new random offset for every 8 I/O's. The result would be a
982 seek for only every 8 I/O's, instead of for every I/O. Use ``rw=randread:8``
983 to specify that. As sequential I/O is already sequential, setting
984 ``sequential`` for that would not result in any differences. ``identical``
985 behaves in a similar fashion, except it sends the same offset 8 number of
986 times before generating a new offset.
988 .. option:: unified_rw_reporting=bool
990 Fio normally reports statistics on a per data direction basis, meaning that
991 reads, writes, and trims are accounted and reported separately. If this
992 option is set fio sums the results and report them as "mixed" instead.
994 .. option:: randrepeat=bool
996 Seed the random number generator used for random I/O patterns in a
997 predictable way so the pattern is repeatable across runs. Default: true.
999 .. option:: allrandrepeat=bool
1001 Seed all random number generators in a predictable way so results are
1002 repeatable across runs. Default: false.
1004 .. option:: randseed=int
1006 Seed the random number generators based on this seed value, to be able to
1007 control what sequence of output is being generated. If not set, the random
1008 sequence depends on the :option:`randrepeat` setting.
1010 .. option:: fallocate=str
1012 Whether pre-allocation is performed when laying down files.
1013 Accepted values are:
1016 Do not pre-allocate space.
1019 Pre-allocate via :manpage:`posix_fallocate(3)`.
1022 Pre-allocate via :manpage:`fallocate(2)` with
1023 FALLOC_FL_KEEP_SIZE set.
1026 Backward-compatible alias for **none**.
1029 Backward-compatible alias for **posix**.
1031 May not be available on all supported platforms. **keep** is only available
1032 on Linux. If using ZFS on Solaris this must be set to **none** because ZFS
1033 doesn't support it. Default: **posix**.
1035 .. option:: fadvise_hint=str
1037 Use :manpage:`posix_fadvise(2)` to advise the kernel on what I/O patterns
1038 are likely to be issued. Accepted values are:
1041 Backwards-compatible hint for "no hint".
1044 Backwards compatible hint for "advise with fio workload type". This
1045 uses **FADV_RANDOM** for a random workload, and **FADV_SEQUENTIAL**
1046 for a sequential workload.
1049 Advise using **FADV_SEQUENTIAL**.
1052 Advise using **FADV_RANDOM**.
1054 .. option:: fadvise_stream=int
1056 Use :manpage:`posix_fadvise(2)` to advise the kernel what stream ID the
1057 writes issued belong to. Only supported on Linux. Note, this option may
1058 change going forward.
1060 .. option:: offset=int
1062 Start I/O at the given offset in the file. The data before the given offset
1063 will not be touched. This effectively caps the file size at `real_size -
1066 .. option:: offset_increment=int
1068 If this is provided, then the real offset becomes `offset + offset_increment
1069 * thread_number`, where the thread number is a counter that starts at 0 and
1070 is incremented for each sub-job (i.e. when :option:`numjobs` option is
1071 specified). This option is useful if there are several jobs which are
1072 intended to operate on a file in parallel disjoint segments, with even
1073 spacing between the starting points.
1075 .. option:: number_ios=int
1077 Fio will normally perform I/Os until it has exhausted the size of the region
1078 set by :option:`size`, or if it exhaust the allocated time (or hits an error
1079 condition). With this setting, the range/size can be set independently of
1080 the number of I/Os to perform. When fio reaches this number, it will exit
1081 normally and report status. Note that this does not extend the amount of I/O
1082 that will be done, it will only stop fio if this condition is met before
1083 other end-of-job criteria.
1085 .. option:: fsync=int
1087 If writing to a file, issue a sync of the dirty data for every number of
1088 blocks given. For example, if you give 32 as a parameter, fio will sync the
1089 file for every 32 writes issued. If fio is using non-buffered I/O, we may
1090 not sync the file. The exception is the sg I/O engine, which synchronizes
1091 the disk cache anyway.
1093 .. option:: fdatasync=int
1095 Like :option:`fsync` but uses :manpage:`fdatasync(2)` to only sync data and
1096 not metadata blocks. In FreeBSD and Windows there is no
1097 :manpage:`fdatasync(2)`, this falls back to using :manpage:`fsync(2)`.
1099 .. option:: write_barrier=int
1101 Make every `N-th` write a barrier write.
1103 .. option:: sync_file_range=str:val
1105 Use :manpage:`sync_file_range(2)` for every `val` number of write
1106 operations. Fio will track range of writes that have happened since the last
1107 :manpage:`sync_file_range(2)` call. `str` can currently be one or more of:
1110 SYNC_FILE_RANGE_WAIT_BEFORE
1112 SYNC_FILE_RANGE_WRITE
1114 SYNC_FILE_RANGE_WAIT_AFTER
1116 So if you do ``sync_file_range=wait_before,write:8``, fio would use
1117 ``SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE`` for every 8
1118 writes. Also see the :manpage:`sync_file_range(2)` man page. This option is
1121 .. option:: overwrite=bool
1123 If true, writes to a file will always overwrite existing data. If the file
1124 doesn't already exist, it will be created before the write phase begins. If
1125 the file exists and is large enough for the specified write phase, nothing
1128 .. option:: end_fsync=bool
1130 If true, fsync file contents when a write stage has completed.
1132 .. option:: fsync_on_close=bool
1134 If true, fio will :manpage:`fsync(2)` a dirty file on close. This differs
1135 from end_fsync in that it will happen on every file close, not just at the
1138 .. option:: rwmixread=int
1140 Percentage of a mixed workload that should be reads. Default: 50.
1142 .. option:: rwmixwrite=int
1144 Percentage of a mixed workload that should be writes. If both
1145 :option:`rwmixread` and :option:`rwmixwrite` is given and the values do not
1146 add up to 100%, the latter of the two will be used to override the
1147 first. This may interfere with a given rate setting, if fio is asked to
1148 limit reads or writes to a certain rate. If that is the case, then the
1149 distribution may be skewed. Default: 50.
1151 .. option:: random_distribution=str:float[,str:float][,str:float]
1153 By default, fio will use a completely uniform random distribution when asked
1154 to perform random I/O. Sometimes it is useful to skew the distribution in
1155 specific ways, ensuring that some parts of the data is more hot than others.
1156 fio includes the following distribution models:
1159 Uniform random distribution
1168 Normal (Gaussian) distribution
1171 Zoned random distribution
1173 When using a **zipf** or **pareto** distribution, an input value is also
1174 needed to define the access pattern. For **zipf**, this is the `zipf
1175 theta`. For **pareto**, it's the `Pareto power`. Fio includes a test
1176 program, :command:`genzipf`, that can be used visualize what the given input
1177 values will yield in terms of hit rates. If you wanted to use **zipf** with
1178 a `theta` of 1.2, you would use ``random_distribution=zipf:1.2`` as the
1179 option. If a non-uniform model is used, fio will disable use of the random
1180 map. For the **gauss** distribution, a normal deviation is supplied as a
1181 value between 0 and 100.
1183 For a **zoned** distribution, fio supports specifying percentages of I/O
1184 access that should fall within what range of the file or device. For
1185 example, given a criteria of:
1187 * 60% of accesses should be to the first 10%
1188 * 30% of accesses should be to the next 20%
1189 * 8% of accesses should be to to the next 30%
1190 * 2% of accesses should be to the next 40%
1192 we can define that through zoning of the random accesses. For the above
1193 example, the user would do::
1195 random_distribution=zoned:60/10:30/20:8/30:2/40
1197 similarly to how :option:`bssplit` works for setting ranges and percentages
1198 of block sizes. Like :option:`bssplit`, it's possible to specify separate
1199 zones for reads, writes, and trims. If just one set is given, it'll apply to
1202 .. option:: percentage_random=int[,int][,int]
1204 For a random workload, set how big a percentage should be random. This
1205 defaults to 100%, in which case the workload is fully random. It can be set
1206 from anywhere from 0 to 100. Setting it to 0 would make the workload fully
1207 sequential. Any setting in between will result in a random mix of sequential
1208 and random I/O, at the given percentages. Comma-separated values may be
1209 specified for reads, writes, and trims as described in :option:`blocksize`.
1211 .. option:: norandommap
1213 Normally fio will cover every block of the file when doing random I/O. If
1214 this option is given, fio will just get a new random offset without looking
1215 at past I/O history. This means that some blocks may not be read or written,
1216 and that some blocks may be read/written more than once. If this option is
1217 used with :option:`verify` and multiple blocksizes (via :option:`bsrange`),
1218 only intact blocks are verified, i.e., partially-overwritten blocks are
1221 .. option:: softrandommap=bool
1223 See :option:`norandommap`. If fio runs with the random block map enabled and
1224 it fails to allocate the map, if this option is set it will continue without
1225 a random block map. As coverage will not be as complete as with random maps,
1226 this option is disabled by default.
1228 .. option:: random_generator=str
1230 Fio supports the following engines for generating
1231 I/O offsets for random I/O:
1234 Strong 2^88 cycle random number generator
1236 Linear feedback shift register generator
1238 Strong 64-bit 2^258 cycle random number generator
1240 **tausworthe** is a strong random number generator, but it requires tracking
1241 on the side if we want to ensure that blocks are only read or written
1242 once. **LFSR** guarantees that we never generate the same offset twice, and
1243 it's also less computationally expensive. It's not a true random generator,
1244 however, though for I/O purposes it's typically good enough. **LFSR** only
1245 works with single block sizes, not with workloads that use multiple block
1246 sizes. If used with such a workload, fio may read or write some blocks
1247 multiple times. The default value is **tausworthe**, unless the required
1248 space exceeds 2^32 blocks. If it does, then **tausworthe64** is
1249 selected automatically.
1255 .. option:: blocksize=int[,int][,int], bs=int[,int][,int]
1257 The block size in bytes used for I/O units. Default: 4096. A single value
1258 applies to reads, writes, and trims. Comma-separated values may be
1259 specified for reads, writes, and trims. A value not terminated in a comma
1260 applies to subsequent types.
1265 means 256k for reads, writes and trims.
1268 means 8k for reads, 32k for writes and trims.
1271 means 8k for reads, 32k for writes, and default for trims.
1274 means default for reads, 8k for writes and trims.
1277 means default for reads, 8k for writes, and default for writes.
1279 .. option:: blocksize_range=irange[,irange][,irange], bsrange=irange[,irange][,irange]
1281 A range of block sizes in bytes for I/O units. The issued I/O unit will
1282 always be a multiple of the minimum size, unless
1283 :option:`blocksize_unaligned` is set.
1285 Comma-separated ranges may be specified for reads, writes, and trims as
1286 described in :option:`blocksize`.
1288 Example: ``bsrange=1k-4k,2k-8k``.
1290 .. option:: bssplit=str[,str][,str]
1292 Sometimes you want even finer grained control of the block sizes issued, not
1293 just an even split between them. This option allows you to weight various
1294 block sizes, so that you are able to define a specific amount of block sizes
1295 issued. The format for this option is::
1297 bssplit=blocksize/percentage:blocksize/percentage
1299 for as many block sizes as needed. So if you want to define a workload that
1300 has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would write::
1302 bssplit=4k/10:64k/50:32k/40
1304 Ordering does not matter. If the percentage is left blank, fio will fill in
1305 the remaining values evenly. So a bssplit option like this one::
1307 bssplit=4k/50:1k/:32k/
1309 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always add up
1310 to 100, if bssplit is given a range that adds up to more, it will error out.
1312 Comma-separated values may be specified for reads, writes, and trims as
1313 described in :option:`blocksize`.
1315 If you want a workload that has 50% 2k reads and 50% 4k reads, while having
1316 90% 4k writes and 10% 8k writes, you would specify::
1318 bssplit=2k/50:4k/50,4k/90,8k/10
1320 .. option:: blocksize_unaligned, bs_unaligned
1322 If set, fio will issue I/O units with any size within
1323 :option:`blocksize_range`, not just multiples of the minimum size. This
1324 typically won't work with direct I/O, as that normally requires sector
1327 .. option:: bs_is_seq_rand
1329 If this option is set, fio will use the normal read,write blocksize settings
1330 as sequential,random blocksize settings instead. Any random read or write
1331 will use the WRITE blocksize settings, and any sequential read or write will
1332 use the READ blocksize settings.
1334 .. option:: blockalign=int[,int][,int], ba=int[,int][,int]
1336 Boundary to which fio will align random I/O units. Default:
1337 :option:`blocksize`. Minimum alignment is typically 512b for using direct
1338 I/O, though it usually depends on the hardware block size. This option is
1339 mutually exclusive with using a random map for files, so it will turn off
1340 that option. Comma-separated values may be specified for reads, writes, and
1341 trims as described in :option:`blocksize`.
1347 .. option:: zero_buffers
1349 Initialize buffers with all zeros. Default: fill buffers with random data.
1351 .. option:: refill_buffers
1353 If this option is given, fio will refill the I/O buffers on every
1354 submit. The default is to only fill it at init time and reuse that
1355 data. Only makes sense if zero_buffers isn't specified, naturally. If data
1356 verification is enabled, `refill_buffers` is also automatically enabled.
1358 .. option:: scramble_buffers=bool
1360 If :option:`refill_buffers` is too costly and the target is using data
1361 deduplication, then setting this option will slightly modify the I/O buffer
1362 contents to defeat normal de-dupe attempts. This is not enough to defeat
1363 more clever block compression attempts, but it will stop naive dedupe of
1364 blocks. Default: true.
1366 .. option:: buffer_compress_percentage=int
1368 If this is set, then fio will attempt to provide I/O buffer content (on
1369 WRITEs) that compress to the specified level. Fio does this by providing a
1370 mix of random data and a fixed pattern. The fixed pattern is either zeroes,
1371 or the pattern specified by :option:`buffer_pattern`. If the pattern option
1372 is used, it might skew the compression ratio slightly. Note that this is per
1373 block size unit, for file/disk wide compression level that matches this
1374 setting, you'll also want to set :option:`refill_buffers`.
1376 .. option:: buffer_compress_chunk=int
1378 See :option:`buffer_compress_percentage`. This setting allows fio to manage
1379 how big the ranges of random data and zeroed data is. Without this set, fio
1380 will provide :option:`buffer_compress_percentage` of blocksize random data,
1381 followed by the remaining zeroed. With this set to some chunk size smaller
1382 than the block size, fio can alternate random and zeroed data throughout the
1385 .. option:: buffer_pattern=str
1387 If set, fio will fill the I/O buffers with this pattern. If not set, the
1388 contents of I/O buffers is defined by the other options related to buffer
1389 contents. The setting can be any pattern of bytes, and can be prefixed with
1390 0x for hex values. It may also be a string, where the string must then be
1391 wrapped with ``""``, e.g.::
1393 buffer_pattern="abcd"
1401 buffer_pattern=0xdeadface
1403 Also you can combine everything together in any order::
1405 buffer_pattern=0xdeadface"abcd"-12
1407 .. option:: dedupe_percentage=int
1409 If set, fio will generate this percentage of identical buffers when
1410 writing. These buffers will be naturally dedupable. The contents of the
1411 buffers depend on what other buffer compression settings have been set. It's
1412 possible to have the individual buffers either fully compressible, or not at
1413 all. This option only controls the distribution of unique buffers.
1415 .. option:: invalidate=bool
1417 Invalidate the buffer/page cache parts for this file prior to starting
1418 I/O. Defaults to true.
1420 .. option:: sync=bool
1422 Use synchronous I/O for buffered writes. For the majority of I/O engines,
1423 this means using O_SYNC. Default: false.
1425 .. option:: iomem=str, mem=str
1427 Fio can use various types of memory as the I/O unit buffer. The allowed
1431 Use memory from :manpage:`malloc(3)` as the buffers. Default memory
1435 Use shared memory as the buffers. Allocated through
1436 :manpage:`shmget(2)`.
1439 Same as shm, but use huge pages as backing.
1442 Use mmap to allocate buffers. May either be anonymous memory, or can
1443 be file backed if a filename is given after the option. The format
1444 is `mem=mmap:/path/to/file`.
1447 Use a memory mapped huge file as the buffer backing. Append filename
1448 after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file`.
1451 Same as mmap, but use a MMAP_SHARED mapping.
1453 The area allocated is a function of the maximum allowed bs size for the job,
1454 multiplied by the I/O depth given. Note that for **shmhuge** and
1455 **mmaphuge** to work, the system must have free huge pages allocated. This
1456 can normally be checked and set by reading/writing
1457 :file:`/proc/sys/vm/nr_hugepages` on a Linux system. Fio assumes a huge page
1458 is 4MiB in size. So to calculate the number of huge pages you need for a
1459 given job file, add up the I/O depth of all jobs (normally one unless
1460 :option:`iodepth` is used) and multiply by the maximum bs set. Then divide
1461 that number by the huge page size. You can see the size of the huge pages in
1462 :file:`/proc/meminfo`. If no huge pages are allocated by having a non-zero
1463 number in `nr_hugepages`, using **mmaphuge** or **shmhuge** will fail. Also
1464 see :option:`hugepage-size`.
1466 **mmaphuge** also needs to have hugetlbfs mounted and the file location
1467 should point there. So if it's mounted in :file:`/huge`, you would use
1468 `mem=mmaphuge:/huge/somefile`.
1470 .. option:: iomem_align=int
1472 This indicates the memory alignment of the I/O memory buffers. Note that
1473 the given alignment is applied to the first I/O unit buffer, if using
1474 :option:`iodepth` the alignment of the following buffers are given by the
1475 :option:`bs` used. In other words, if using a :option:`bs` that is a
1476 multiple of the page sized in the system, all buffers will be aligned to
1477 this value. If using a :option:`bs` that is not page aligned, the alignment
1478 of subsequent I/O memory buffers is the sum of the :option:`iomem_align` and
1481 .. option:: hugepage-size=int
1483 Defines the size of a huge page. Must at least be equal to the system
1484 setting, see :file:`/proc/meminfo`. Defaults to 4MiB. Should probably
1485 always be a multiple of megabytes, so using ``hugepage-size=Xm`` is the
1486 preferred way to set this to avoid setting a non-pow-2 bad value.
1488 .. option:: lockmem=int
1490 Pin the specified amount of memory with :manpage:`mlock(2)`. Can be used to
1491 simulate a smaller amount of memory. The amount specified is per worker.
1497 .. option:: size=int
1499 The total size of file I/O for this job. Fio will run until this many bytes
1500 has been transferred, unless runtime is limited by other options (such as
1501 :option:`runtime`, for instance, or increased/decreased by
1502 :option:`io_size`). Unless specific :option:`nrfiles` and :option:`filesize`
1503 options are given, fio will divide this size between the available files
1504 specified by the job. If not set, fio will use the full size of the given
1505 files or devices. If the files do not exist, size must be given. It is also
1506 possible to give size as a percentage between 1 and 100. If ``size=20%`` is
1507 given, fio will use 20% of the full size of the given files or devices.
1509 .. option:: io_size=int, io_limit=int
1511 Normally fio operates within the region set by :option:`size`, which means
1512 that the :option:`size` option sets both the region and size of I/O to be
1513 performed. Sometimes that is not what you want. With this option, it is
1514 possible to define just the amount of I/O that fio should do. For instance,
1515 if :option:`size` is set to 20GiB and :option:`io_size` is set to 5GiB, fio
1516 will perform I/O within the first 20GiB but exit when 5GiB have been
1517 done. The opposite is also possible -- if :option:`size` is set to 20GiB,
1518 and :option:`io_size` is set to 40GiB, then fio will do 40GiB of I/O within
1519 the 0..20GiB region.
1521 .. option:: filesize=int
1523 Individual file sizes. May be a range, in which case fio will select sizes
1524 for files at random within the given range and limited to :option:`size` in
1525 total (if that is given). If not given, each created file is the same size.
1527 .. option:: file_append=bool
1529 Perform I/O after the end of the file. Normally fio will operate within the
1530 size of a file. If this option is set, then fio will append to the file
1531 instead. This has identical behavior to setting :option:`offset` to the size
1532 of a file. This option is ignored on non-regular files.
1534 .. option:: fill_device=bool, fill_fs=bool
1536 Sets size to something really large and waits for ENOSPC (no space left on
1537 device) as the terminating condition. Only makes sense with sequential
1538 write. For a read workload, the mount point will be filled first then I/O
1539 started on the result. This option doesn't make sense if operating on a raw
1540 device node, since the size of that is already known by the file system.
1541 Additionally, writing beyond end-of-device will not return ENOSPC there.
1547 .. option:: ioengine=str
1549 Defines how the job issues I/O to the file. The following types are defined:
1552 Basic :manpage:`read(2)` or :manpage:`write(2)`
1553 I/O. :manpage:`lseek(2)` is used to position the I/O location.
1556 Basic :manpage:`pread(2)` or :manpage:`pwrite(2)` I/O. Default on
1557 all supported operating systems except for Windows.
1560 Basic :manpage:`readv(2)` or :manpage:`writev(2)` I/O. Will emulate
1561 queuing by coalescing adjacent I/Os into a single submission.
1564 Basic :manpage:`preadv(2)` or :manpage:`pwritev(2)` I/O.
1567 Basic :manpage:`preadv2(2)` or :manpage:`pwritev2(2)` I/O.
1570 Linux native asynchronous I/O. Note that Linux may only support
1571 queued behaviour with non-buffered I/O (set ``direct=1`` or
1573 This engine defines engine specific options.
1576 POSIX asynchronous I/O using :manpage:`aio_read(3)` and
1577 :manpage:`aio_write(3)`.
1580 Solaris native asynchronous I/O.
1583 Windows native asynchronous I/O. Default on Windows.
1586 File is memory mapped with :manpage:`mmap(2)` and data copied
1587 to/from using :manpage:`memcpy(3)`.
1590 :manpage:`splice(2)` is used to transfer the data and
1591 :manpage:`vmsplice(2)` to transfer data from user space to the
1595 SCSI generic sg v3 I/O. May either be synchronous using the SG_IO
1596 ioctl, or if the target is an sg character device we use
1597 :manpage:`read(2)` and :manpage:`write(2)` for asynchronous
1598 I/O. Requires filename option to specify either block or character
1602 Doesn't transfer any data, just pretends to. This is mainly used to
1603 exercise fio itself and for debugging/testing purposes.
1606 Transfer over the network to given ``host:port``. Depending on the
1607 :option:`protocol` used, the :option:`hostname`, :option:`port`,
1608 :option:`listen` and :option:`filename` options are used to specify
1609 what sort of connection to make, while the :option:`protocol` option
1610 determines which protocol will be used. This engine defines engine
1614 Like **net**, but uses :manpage:`splice(2)` and
1615 :manpage:`vmsplice(2)` to map data and send/receive.
1616 This engine defines engine specific options.
1619 Doesn't transfer any data, but burns CPU cycles according to the
1620 :option:`cpuload` and :option:`cpuchunks` options. Setting
1621 :option:`cpuload` =85 will cause that job to do nothing but burn 85%
1622 of the CPU. In case of SMP machines, use :option:`numjobs`
1623 =<no_of_cpu> to get desired CPU usage, as the cpuload only loads a
1624 single CPU at the desired rate. A job never finishes unless there is
1625 at least one non-cpuio job.
1628 The GUASI I/O engine is the Generic Userspace Asyncronous Syscall
1629 Interface approach to async I/O. See
1631 http://www.xmailserver.org/guasi-lib.html
1633 for more info on GUASI.
1636 The RDMA I/O engine supports both RDMA memory semantics
1637 (RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the
1638 InfiniBand, RoCE and iWARP protocols.
1641 I/O engine that does regular fallocate to simulate data transfer as
1645 does fallocate(,mode = FALLOC_FL_KEEP_SIZE,).
1648 does fallocate(,mode = 0).
1651 does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE).
1654 I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate
1655 defragment activity in request to DDIR_WRITE event.
1658 I/O engine supporting direct access to Ceph Rados Block Devices
1659 (RBD) via librbd without the need to use the kernel rbd driver. This
1660 ioengine defines engine specific options.
1663 Using Glusterfs libgfapi sync interface to direct access to
1664 Glusterfs volumes without having to go through FUSE. This ioengine
1665 defines engine specific options.
1668 Using Glusterfs libgfapi async interface to direct access to
1669 Glusterfs volumes without having to go through FUSE. This ioengine
1670 defines engine specific options.
1673 Read and write through Hadoop (HDFS). The :file:`filename` option
1674 is used to specify host,port of the hdfs name-node to connect. This
1675 engine interprets offsets a little differently. In HDFS, files once
1676 created cannot be modified. So random writes are not possible. To
1677 imitate this, libhdfs engine expects bunch of small files to be
1678 created over HDFS, and engine will randomly pick a file out of those
1679 files based on the offset generated by fio backend. (see the example
1680 job file to create such files, use ``rw=write`` option). Please
1681 note, you might want to set necessary environment variables to work
1682 with hdfs/libhdfs properly. Each jobs uses it's own connection to
1686 Read, write and erase an MTD character device (e.g.,
1687 :file:`/dev/mtd0`). Discards are treated as erases. Depending on the
1688 underlying device type, the I/O may have to go in a certain pattern,
1689 e.g., on NAND, writing sequentially to erase blocks and discarding
1690 before overwriting. The writetrim mode works well for this
1694 Read and write using filesystem DAX to a file on a filesystem
1695 mounted with DAX on a persistent memory device through the NVML
1699 Read and write using device DAX to a persistent memory device (e.g.,
1700 /dev/dax0.0) through the NVML libpmem library.
1703 Prefix to specify loading an external I/O engine object file. Append
1704 the engine filename, e.g. ``ioengine=external:/tmp/foo.o`` to load
1705 ioengine :file:`foo.o` in :file:`/tmp`.
1708 I/O engine specific parameters
1709 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1711 In addition, there are some parameters which are only valid when a specific
1712 ioengine is in use. These are used identically to normal parameters, with the
1713 caveat that when used on the command line, they must come after the
1714 :option:`ioengine` that defines them is selected.
1716 .. option:: userspace_reap : [libaio]
1718 Normally, with the libaio engine in use, fio will use the
1719 :manpage:`io_getevents(2)` system call to reap newly returned events. With
1720 this flag turned on, the AIO ring will be read directly from user-space to
1721 reap events. The reaping mode is only enabled when polling for a minimum of
1722 0 events (e.g. when :option:`iodepth_batch_complete` `=0`).
1724 .. option:: hipri : [psyncv2]
1726 Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority
1729 .. option:: cpuload=int : [cpuio]
1731 Attempt to use the specified percentage of CPU cycles.
1733 .. option:: cpuchunks=int : [cpuio]
1735 Split the load into cycles of the given time. In microseconds.
1737 .. option:: exit_on_io_done=bool : [cpuio]
1739 Detect when I/O threads are done, then exit.
1741 .. option:: hostname=str : [netsplice] [net]
1743 The host name or IP address to use for TCP or UDP based I/O. If the job is
1744 a TCP listener or UDP reader, the host name is not used and must be omitted
1745 unless it is a valid UDP multicast address.
1747 .. option:: namenode=str : [libhdfs]
1749 The host name or IP address of a HDFS cluster namenode to contact.
1751 .. option:: port=int
1755 The TCP or UDP port to bind to or connect to. If this is used with
1756 :option:`numjobs` to spawn multiple instances of the same job type, then
1757 this will be the starting port number since fio will use a range of
1762 the listening port of the HFDS cluster namenode.
1764 .. option:: interface=str : [netsplice] [net]
1766 The IP address of the network interface used to send or receive UDP
1769 .. option:: ttl=int : [netsplice] [net]
1771 Time-to-live value for outgoing UDP multicast packets. Default: 1.
1773 .. option:: nodelay=bool : [netsplice] [net]
1775 Set TCP_NODELAY on TCP connections.
1777 .. option:: protocol=str : [netsplice] [net]
1779 .. option:: proto=str : [netsplice] [net]
1781 The network protocol to use. Accepted values are:
1784 Transmission control protocol.
1786 Transmission control protocol V6.
1788 User datagram protocol.
1790 User datagram protocol V6.
1794 When the protocol is TCP or UDP, the port must also be given, as well as the
1795 hostname if the job is a TCP listener or UDP reader. For unix sockets, the
1796 normal filename option should be used and the port is invalid.
1798 .. option:: listen : [net]
1800 For TCP network connections, tell fio to listen for incoming connections
1801 rather than initiating an outgoing connection. The :option:`hostname` must
1802 be omitted if this option is used.
1804 .. option:: pingpong : [net]
1806 Normally a network writer will just continue writing data, and a network
1807 reader will just consume packages. If ``pingpong=1`` is set, a writer will
1808 send its normal payload to the reader, then wait for the reader to send the
1809 same payload back. This allows fio to measure network latencies. The
1810 submission and completion latencies then measure local time spent sending or
1811 receiving, and the completion latency measures how long it took for the
1812 other end to receive and send back. For UDP multicast traffic
1813 ``pingpong=1`` should only be set for a single reader when multiple readers
1814 are listening to the same address.
1816 .. option:: window_size : [net]
1818 Set the desired socket buffer size for the connection.
1820 .. option:: mss : [net]
1822 Set the TCP maximum segment size (TCP_MAXSEG).
1824 .. option:: donorname=str : [e4defrag]
1826 File will be used as a block donor(swap extents between files).
1828 .. option:: inplace=int : [e4defrag]
1830 Configure donor file blocks allocation strategy:
1833 Default. Preallocate donor's file on init.
1835 Allocate space immediately inside defragment event, and free right
1838 .. option:: clustername=str : [rbd]
1840 Specifies the name of the Ceph cluster.
1842 .. option:: rbdname=str : [rbd]
1844 Specifies the name of the RBD.
1846 .. option:: pool=str : [rbd]
1848 Specifies the name of the Ceph pool containing RBD.
1850 .. option:: clientname=str : [rbd]
1852 Specifies the username (without the 'client.' prefix) used to access the
1853 Ceph cluster. If the *clustername* is specified, the *clientname* shall be
1854 the full *type.id* string. If no type. prefix is given, fio will add
1855 'client.' by default.
1857 .. option:: skip_bad=bool : [mtd]
1859 Skip operations against known bad blocks.
1861 .. option:: hdfsdirectory : [libhdfs]
1863 libhdfs will create chunk in this HDFS directory.
1865 .. option:: chunk_size : [libhdfs]
1867 the size of the chunk to use for each file.
1873 .. option:: iodepth=int
1875 Number of I/O units to keep in flight against the file. Note that
1876 increasing *iodepth* beyond 1 will not affect synchronous ioengines (except
1877 for small degrees when :option:`verify_async` is in use). Even async
1878 engines may impose OS restrictions causing the desired depth not to be
1879 achieved. This may happen on Linux when using libaio and not setting
1880 :option:`direct` =1, since buffered I/O is not async on that OS. Keep an
1881 eye on the I/O depth distribution in the fio output to verify that the
1882 achieved depth is as expected. Default: 1.
1884 .. option:: iodepth_batch_submit=int, iodepth_batch=int
1886 This defines how many pieces of I/O to submit at once. It defaults to 1
1887 which means that we submit each I/O as soon as it is available, but can be
1888 raised to submit bigger batches of I/O at the time. If it is set to 0 the
1889 :option:`iodepth` value will be used.
1891 .. option:: iodepth_batch_complete_min=int, iodepth_batch_complete=int
1893 This defines how many pieces of I/O to retrieve at once. It defaults to 1
1894 which means that we'll ask for a minimum of 1 I/O in the retrieval process
1895 from the kernel. The I/O retrieval will go on until we hit the limit set by
1896 :option:`iodepth_low`. If this variable is set to 0, then fio will always
1897 check for completed events before queuing more I/O. This helps reduce I/O
1898 latency, at the cost of more retrieval system calls.
1900 .. option:: iodepth_batch_complete_max=int
1902 This defines maximum pieces of I/O to retrieve at once. This variable should
1903 be used along with :option:`iodepth_batch_complete_min` =int variable,
1904 specifying the range of min and max amount of I/O which should be
1905 retrieved. By default it is equal to :option:`iodepth_batch_complete_min`
1910 iodepth_batch_complete_min=1
1911 iodepth_batch_complete_max=<iodepth>
1913 which means that we will retrieve at least 1 I/O and up to the whole
1914 submitted queue depth. If none of I/O has been completed yet, we will wait.
1918 iodepth_batch_complete_min=0
1919 iodepth_batch_complete_max=<iodepth>
1921 which means that we can retrieve up to the whole submitted queue depth, but
1922 if none of I/O has been completed yet, we will NOT wait and immediately exit
1923 the system call. In this example we simply do polling.
1925 .. option:: iodepth_low=int
1927 The low water mark indicating when to start filling the queue
1928 again. Defaults to the same as :option:`iodepth`, meaning that fio will
1929 attempt to keep the queue full at all times. If :option:`iodepth` is set to
1930 e.g. 16 and *iodepth_low* is set to 4, then after fio has filled the queue of
1931 16 requests, it will let the depth drain down to 4 before starting to fill
1934 .. option:: io_submit_mode=str
1936 This option controls how fio submits the I/O to the I/O engine. The default
1937 is `inline`, which means that the fio job threads submit and reap I/O
1938 directly. If set to `offload`, the job threads will offload I/O submission
1939 to a dedicated pool of I/O threads. This requires some coordination and thus
1940 has a bit of extra overhead, especially for lower queue depth I/O where it
1941 can increase latencies. The benefit is that fio can manage submission rates
1942 independently of the device completion rates. This avoids skewed latency
1943 reporting if I/O gets back up on the device side (the coordinated omission
1950 .. option:: thinktime=time
1952 Stall the job for the specified period of time after an I/O has completed before issuing the
1953 next. May be used to simulate processing being done by an application.
1954 When the unit is omitted, the value is given in microseconds. See
1955 :option:`thinktime_blocks` and :option:`thinktime_spin`.
1957 .. option:: thinktime_spin=time
1959 Only valid if :option:`thinktime` is set - pretend to spend CPU time doing
1960 something with the data received, before falling back to sleeping for the
1961 rest of the period specified by :option:`thinktime`. When the unit is
1962 omitted, the value is given in microseconds.
1964 .. option:: thinktime_blocks=int
1966 Only valid if :option:`thinktime` is set - control how many blocks to issue,
1967 before waiting `thinktime` usecs. If not set, defaults to 1 which will make
1968 fio wait `thinktime` usecs after every block. This effectively makes any
1969 queue depth setting redundant, since no more than 1 I/O will be queued
1970 before we have to complete it and do our thinktime. In other words, this
1971 setting effectively caps the queue depth if the latter is larger.
1973 .. option:: rate=int[,int][,int]
1975 Cap the bandwidth used by this job. The number is in bytes/sec, the normal
1976 suffix rules apply. Comma-separated values may be specified for reads,
1977 writes, and trims as described in :option:`blocksize`.
1979 .. option:: rate_min=int[,int][,int]
1981 Tell fio to do whatever it can to maintain at least this bandwidth. Failing
1982 to meet this requirement will cause the job to exit. Comma-separated values
1983 may be specified for reads, writes, and trims as described in
1984 :option:`blocksize`.
1986 .. option:: rate_iops=int[,int][,int]
1988 Cap the bandwidth to this number of IOPS. Basically the same as
1989 :option:`rate`, just specified independently of bandwidth. If the job is
1990 given a block size range instead of a fixed value, the smallest block size
1991 is used as the metric. Comma-separated values may be specified for reads,
1992 writes, and trims as described in :option:`blocksize`.
1994 .. option:: rate_iops_min=int[,int][,int]
1996 If fio doesn't meet this rate of I/O, it will cause the job to exit.
1997 Comma-separated values may be specified for reads, writes, and trims as
1998 described in :option:`blocksize`.
2000 .. option:: rate_process=str
2002 This option controls how fio manages rated I/O submissions. The default is
2003 `linear`, which submits I/O in a linear fashion with fixed delays between
2004 I/Os that gets adjusted based on I/O completion rates. If this is set to
2005 `poisson`, fio will submit I/O based on a more real world random request
2006 flow, known as the Poisson process
2007 (https://en.wikipedia.org/wiki/Poisson_point_process). The lambda will be
2008 10^6 / IOPS for the given workload.
2014 .. option:: latency_target=time
2016 If set, fio will attempt to find the max performance point that the given
2017 workload will run at while maintaining a latency below this target. When
2018 the unit is omitted, the value is given in microseconds. See
2019 :option:`latency_window` and :option:`latency_percentile`.
2021 .. option:: latency_window=time
2023 Used with :option:`latency_target` to specify the sample window that the job
2024 is run at varying queue depths to test the performance. When the unit is
2025 omitted, the value is given in microseconds.
2027 .. option:: latency_percentile=float
2029 The percentage of I/Os that must fall within the criteria specified by
2030 :option:`latency_target` and :option:`latency_window`. If not set, this
2031 defaults to 100.0, meaning that all I/Os must be equal or below to the value
2032 set by :option:`latency_target`.
2034 .. option:: max_latency=time
2036 If set, fio will exit the job with an ETIMEDOUT error if it exceeds this
2037 maximum latency. When the unit is omitted, the value is given in
2040 .. option:: rate_cycle=int
2042 Average bandwidth for :option:`rate` and :option:`rate_min` over this number
2049 .. option:: write_iolog=str
2051 Write the issued I/O patterns to the specified file. See
2052 :option:`read_iolog`. Specify a separate file for each job, otherwise the
2053 iologs will be interspersed and the file may be corrupt.
2055 .. option:: read_iolog=str
2057 Open an iolog with the specified file name and replay the I/O patterns it
2058 contains. This can be used to store a workload and replay it sometime
2059 later. The iolog given may also be a blktrace binary file, which allows fio
2060 to replay a workload captured by :command:`blktrace`. See
2061 :manpage:`blktrace(8)` for how to capture such logging data. For blktrace
2062 replay, the file needs to be turned into a blkparse binary data file first
2063 (``blkparse <device> -o /dev/null -d file_for_fio.bin``).
2065 .. option:: replay_no_stall=int
2067 When replaying I/O with :option:`read_iolog` the default behavior is to
2068 attempt to respect the time stamps within the log and replay them with the
2069 appropriate delay between IOPS. By setting this variable fio will not
2070 respect the timestamps and attempt to replay them as fast as possible while
2071 still respecting ordering. The result is the same I/O pattern to a given
2072 device, but different timings.
2074 .. option:: replay_redirect=str
2076 While replaying I/O patterns using :option:`read_iolog` the default behavior
2077 is to replay the IOPS onto the major/minor device that each IOP was recorded
2078 from. This is sometimes undesirable because on a different machine those
2079 major/minor numbers can map to a different device. Changing hardware on the
2080 same system can also result in a different major/minor mapping.
2081 ``replay_redirect`` causes all IOPS to be replayed onto the single specified
2082 device regardless of the device it was recorded
2083 from. i.e. :option:`replay_redirect` = :file:`/dev/sdc` would cause all I/O
2084 in the blktrace or iolog to be replayed onto :file:`/dev/sdc`. This means
2085 multiple devices will be replayed onto a single device, if the trace
2086 contains multiple devices. If you want multiple devices to be replayed
2087 concurrently to multiple redirected devices you must blkparse your trace
2088 into separate traces and replay them with independent fio invocations.
2089 Unfortunately this also breaks the strict time ordering between multiple
2092 .. option:: replay_align=int
2094 Force alignment of I/O offsets and lengths in a trace to this power of 2
2097 .. option:: replay_scale=int
2099 Scale sector offsets down by this factor when replaying traces.
2102 Threads, processes and job synchronization
2103 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2107 Fio defaults to forking jobs, however if this option is given, fio will use
2108 :manpage:`pthread_create(3)` to create threads instead.
2110 .. option:: wait_for=str
2112 Specifies the name of the already defined job to wait for. Single waitee
2113 name only may be specified. If set, the job won't be started until all
2114 workers of the waitee job are done.
2116 ``wait_for`` operates on the job name basis, so there are a few
2117 limitations. First, the waitee must be defined prior to the waiter job
2118 (meaning no forward references). Second, if a job is being referenced as a
2119 waitee, it must have a unique name (no duplicate waitees).
2121 .. option:: nice=int
2123 Run the job with the given nice value. See man :manpage:`nice(2)`.
2125 On Windows, values less than -15 set the process class to "High"; -1 through
2126 -15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle"
2129 .. option:: prio=int
2131 Set the I/O priority value of this job. Linux limits us to a positive value
2132 between 0 and 7, with 0 being the highest. See man
2133 :manpage:`ionice(1)`. Refer to an appropriate manpage for other operating
2134 systems since meaning of priority may differ.
2136 .. option:: prioclass=int
2138 Set the I/O priority class. See man :manpage:`ionice(1)`.
2140 .. option:: cpumask=int
2142 Set the CPU affinity of this job. The parameter given is a bitmask of
2143 allowed CPU's the job may run on. So if you want the allowed CPUs to be 1
2144 and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man
2145 :manpage:`sched_setaffinity(2)`. This may not work on all supported
2146 operating systems or kernel versions. This option doesn't work well for a
2147 higher CPU count than what you can store in an integer mask, so it can only
2148 control cpus 1-32. For boxes with larger CPU counts, use
2149 :option:`cpus_allowed`.
2151 .. option:: cpus_allowed=str
2153 Controls the same options as :option:`cpumask`, but it allows a text setting
2154 of the permitted CPUs instead. So to use CPUs 1 and 5, you would specify
2155 ``cpus_allowed=1,5``. This options also allows a range of CPUs. Say you
2156 wanted a binding to CPUs 1, 5, and 8-15, you would set
2157 ``cpus_allowed=1,5,8-15``.
2159 .. option:: cpus_allowed_policy=str
2161 Set the policy of how fio distributes the CPUs specified by
2162 :option:`cpus_allowed` or cpumask. Two policies are supported:
2165 All jobs will share the CPU set specified.
2167 Each job will get a unique CPU from the CPU set.
2169 **shared** is the default behaviour, if the option isn't specified. If
2170 **split** is specified, then fio will will assign one cpu per job. If not
2171 enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs
2174 .. option:: numa_cpu_nodes=str
2176 Set this job running on specified NUMA nodes' CPUs. The arguments allow
2177 comma delimited list of cpu numbers, A-B ranges, or `all`. Note, to enable
2178 numa options support, fio must be built on a system with libnuma-dev(el)
2181 .. option:: numa_mem_policy=str
2183 Set this job's memory policy and corresponding NUMA nodes. Format of the
2188 ``mode`` is one of the following memory policy: ``default``, ``prefer``,
2189 ``bind``, ``interleave``, ``local`` For ``default`` and ``local`` memory
2190 policy, no node is needed to be specified. For ``prefer``, only one node is
2191 allowed. For ``bind`` and ``interleave``, it allow comma delimited list of
2192 numbers, A-B ranges, or `all`.
2194 .. option:: cgroup=str
2196 Add job to this control group. If it doesn't exist, it will be created. The
2197 system must have a mounted cgroup blkio mount point for this to work. If
2198 your system doesn't have it mounted, you can do so with::
2200 # mount -t cgroup -o blkio none /cgroup
2202 .. option:: cgroup_weight=int
2204 Set the weight of the cgroup to this value. See the documentation that comes
2205 with the kernel, allowed values are in the range of 100..1000.
2207 .. option:: cgroup_nodelete=bool
2209 Normally fio will delete the cgroups it has created after the job
2210 completion. To override this behavior and to leave cgroups around after the
2211 job completion, set ``cgroup_nodelete=1``. This can be useful if one wants
2212 to inspect various cgroup files after job completion. Default: false.
2214 .. option:: flow_id=int
2216 The ID of the flow. If not specified, it defaults to being a global
2217 flow. See :option:`flow`.
2219 .. option:: flow=int
2221 Weight in token-based flow control. If this value is used, then there is a
2222 'flow counter' which is used to regulate the proportion of activity between
2223 two or more jobs. Fio attempts to keep this flow counter near zero. The
2224 ``flow`` parameter stands for how much should be added or subtracted to the
2225 flow counter on each iteration of the main I/O loop. That is, if one job has
2226 ``flow=8`` and another job has ``flow=-1``, then there will be a roughly 1:8
2227 ratio in how much one runs vs the other.
2229 .. option:: flow_watermark=int
2231 The maximum value that the absolute value of the flow counter is allowed to
2232 reach before the job must wait for a lower value of the counter.
2234 .. option:: flow_sleep=int
2236 The period of time, in microseconds, to wait after the flow watermark has
2237 been exceeded before retrying operations.
2239 .. option:: stonewall, wait_for_previous
2241 Wait for preceding jobs in the job file to exit, before starting this
2242 one. Can be used to insert serialization points in the job file. A stone
2243 wall also implies starting a new reporting group, see
2244 :option:`group_reporting`.
2248 When one job finishes, terminate the rest. The default is to wait for each
2249 job to finish, sometimes that is not the desired action.
2251 .. option:: exec_prerun=str
2253 Before running this job, issue the command specified through
2254 :manpage:`system(3)`. Output is redirected in a file called
2255 :file:`jobname.prerun.txt`.
2257 .. option:: exec_postrun=str
2259 After the job completes, issue the command specified though
2260 :manpage:`system(3)`. Output is redirected in a file called
2261 :file:`jobname.postrun.txt`.
2265 Instead of running as the invoking user, set the user ID to this value
2266 before the thread/process does any work.
2270 Set group ID, see :option:`uid`.
2276 .. option:: verify_only
2278 Do not perform specified workload, only verify data still matches previous
2279 invocation of this workload. This option allows one to check data multiple
2280 times at a later date without overwriting it. This option makes sense only
2281 for workloads that write data, and does not support workloads with the
2282 :option:`time_based` option set.
2284 .. option:: do_verify=bool
2286 Run the verify phase after a write phase. Only valid if :option:`verify` is
2289 .. option:: verify=str
2291 If writing to a file, fio can verify the file contents after each iteration
2292 of the job. Each verification method also implies verification of special
2293 header, which is written to the beginning of each block. This header also
2294 includes meta information, like offset of the block, block number, timestamp
2295 when block was written, etc. :option:`verify` can be combined with
2296 :option:`verify_pattern` option. The allowed values are:
2299 Use an md5 sum of the data area and store it in the header of
2303 Use an experimental crc64 sum of the data area and store it in the
2304 header of each block.
2307 Use a crc32c sum of the data area and store it in the header of each
2311 Use hardware assisted crc32c calculation provided on SSE4.2 enabled
2312 processors. Falls back to regular software crc32c, if not supported
2316 Use a crc32 sum of the data area and store it in the header of each
2320 Use a crc16 sum of the data area and store it in the header of each
2324 Use a crc7 sum of the data area and store it in the header of each
2328 Use xxhash as the checksum function. Generally the fastest software
2329 checksum that fio supports.
2332 Use sha512 as the checksum function.
2335 Use sha256 as the checksum function.
2338 Use optimized sha1 as the checksum function.
2341 This option is deprecated, since now meta information is included in
2342 generic verification header and meta verification happens by
2343 default. For detailed information see the description of the
2344 :option:`verify` setting. This option is kept because of
2345 compatibility's sake with old configurations. Do not use it.
2348 Verify a strict pattern. Normally fio includes a header with some
2349 basic information and checksumming, but if this option is set, only
2350 the specific pattern set with :option:`verify_pattern` is verified.
2353 Only pretend to verify. Useful for testing internals with
2354 :option:`ioengine` `=null`, not for much else.
2356 This option can be used for repeated burn-in tests of a system to make sure
2357 that the written data is also correctly read back. If the data direction
2358 given is a read or random read, fio will assume that it should verify a
2359 previously written file. If the data direction includes any form of write,
2360 the verify will be of the newly written data.
2362 .. option:: verifysort=bool
2364 If true, fio will sort written verify blocks when it deems it faster to read
2365 them back in a sorted manner. This is often the case when overwriting an
2366 existing file, since the blocks are already laid out in the file system. You
2367 can ignore this option unless doing huge amounts of really fast I/O where
2368 the red-black tree sorting CPU time becomes significant. Default: true.
2370 .. option:: verifysort_nr=int
2372 Pre-load and sort verify blocks for a read workload.
2374 .. option:: verify_offset=int
2376 Swap the verification header with data somewhere else in the block before
2377 writing. It is swapped back before verifying.
2379 .. option:: verify_interval=int
2381 Write the verification header at a finer granularity than the
2382 :option:`blocksize`. It will be written for chunks the size of
2383 ``verify_interval``. :option:`blocksize` should divide this evenly.
2385 .. option:: verify_pattern=str
2387 If set, fio will fill the I/O buffers with this pattern. Fio defaults to
2388 filling with totally random bytes, but sometimes it's interesting to fill
2389 with a known pattern for I/O verification purposes. Depending on the width
2390 of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time(it can
2391 be either a decimal or a hex number). The ``verify_pattern`` if larger than
2392 a 32-bit quantity has to be a hex number that starts with either "0x" or
2393 "0X". Use with :option:`verify`. Also, ``verify_pattern`` supports %o
2394 format, which means that for each block offset will be written and then
2395 verified back, e.g.::
2399 Or use combination of everything::
2401 verify_pattern=0xff%o"abcd"-12
2403 .. option:: verify_fatal=bool
2405 Normally fio will keep checking the entire contents before quitting on a
2406 block verification failure. If this option is set, fio will exit the job on
2407 the first observed failure. Default: false.
2409 .. option:: verify_dump=bool
2411 If set, dump the contents of both the original data block and the data block
2412 we read off disk to files. This allows later analysis to inspect just what
2413 kind of data corruption occurred. Off by default.
2415 .. option:: verify_async=int
2417 Fio will normally verify I/O inline from the submitting thread. This option
2418 takes an integer describing how many async offload threads to create for I/O
2419 verification instead, causing fio to offload the duty of verifying I/O
2420 contents to one or more separate threads. If using this offload option, even
2421 sync I/O engines can benefit from using an :option:`iodepth` setting higher
2422 than 1, as it allows them to have I/O in flight while verifies are running.
2424 .. option:: verify_async_cpus=str
2426 Tell fio to set the given CPU affinity on the async I/O verification
2427 threads. See :option:`cpus_allowed` for the format used.
2429 .. option:: verify_backlog=int
2431 Fio will normally verify the written contents of a job that utilizes verify
2432 once that job has completed. In other words, everything is written then
2433 everything is read back and verified. You may want to verify continually
2434 instead for a variety of reasons. Fio stores the meta data associated with
2435 an I/O block in memory, so for large verify workloads, quite a bit of memory
2436 would be used up holding this meta data. If this option is enabled, fio will
2437 write only N blocks before verifying these blocks.
2439 .. option:: verify_backlog_batch=int
2441 Control how many blocks fio will verify if :option:`verify_backlog` is
2442 set. If not set, will default to the value of :option:`verify_backlog`
2443 (meaning the entire queue is read back and verified). If
2444 ``verify_backlog_batch`` is less than :option:`verify_backlog` then not all
2445 blocks will be verified, if ``verify_backlog_batch`` is larger than
2446 :option:`verify_backlog`, some blocks will be verified more than once.
2448 .. option:: verify_state_save=bool
2450 When a job exits during the write phase of a verify workload, save its
2451 current state. This allows fio to replay up until that point, if the verify
2452 state is loaded for the verify read phase. The format of the filename is,
2455 <type>-<jobname>-<jobindex>-verify.state.
2457 <type> is "local" for a local run, "sock" for a client/server socket
2458 connection, and "ip" (192.168.0.1, for instance) for a networked
2459 client/server connection.
2461 .. option:: verify_state_load=bool
2463 If a verify termination trigger was used, fio stores the current write state
2464 of each thread. This can be used at verification time so that fio knows how
2465 far it should verify. Without this information, fio will run a full
2466 verification pass, according to the settings in the job file used.
2468 .. option:: trim_percentage=int
2470 Number of verify blocks to discard/trim.
2472 .. option:: trim_verify_zero=bool
2474 Verify that trim/discarded blocks are returned as zeroes.
2476 .. option:: trim_backlog=int
2478 Verify that trim/discarded blocks are returned as zeroes.
2480 .. option:: trim_backlog_batch=int
2482 Trim this number of I/O blocks.
2484 .. option:: experimental_verify=bool
2486 Enable experimental verification.
2492 .. option:: steadystate=str:float, ss=str:float
2494 Define the criterion and limit for assessing steady state performance. The
2495 first parameter designates the criterion whereas the second parameter sets
2496 the threshold. When the criterion falls below the threshold for the
2497 specified duration, the job will stop. For example, `iops_slope:0.1%` will
2498 direct fio to terminate the job when the least squares regression slope
2499 falls below 0.1% of the mean IOPS. If :option:`group_reporting` is enabled
2500 this will apply to all jobs in the group. Below is the list of available
2501 steady state assessment criteria. All assessments are carried out using only
2502 data from the rolling collection window. Threshold limits can be expressed
2503 as a fixed value or as a percentage of the mean in the collection window.
2506 Collect IOPS data. Stop the job if all individual IOPS measurements
2507 are within the specified limit of the mean IOPS (e.g., ``iops:2``
2508 means that all individual IOPS values must be within 2 of the mean,
2509 whereas ``iops:0.2%`` means that all individual IOPS values must be
2510 within 0.2% of the mean IOPS to terminate the job).
2513 Collect IOPS data and calculate the least squares regression
2514 slope. Stop the job if the slope falls below the specified limit.
2517 Collect bandwidth data. Stop the job if all individual bandwidth
2518 measurements are within the specified limit of the mean bandwidth.
2521 Collect bandwidth data and calculate the least squares regression
2522 slope. Stop the job if the slope falls below the specified limit.
2524 .. option:: steadystate_duration=time, ss_dur=time
2526 A rolling window of this duration will be used to judge whether steady state
2527 has been reached. Data will be collected once per second. The default is 0
2528 which disables steady state detection. When the unit is omitted, the
2529 value is given in seconds.
2531 .. option:: steadystate_ramp_time=time, ss_ramp=time
2533 Allow the job to run for the specified duration before beginning data
2534 collection for checking the steady state job termination criterion. The
2535 default is 0. When the unit is omitted, the value is given in seconds.
2538 Measurements and reporting
2539 ~~~~~~~~~~~~~~~~~~~~~~~~~~
2541 .. option:: per_job_logs=bool
2543 If set, this generates bw/clat/iops log with per file private filenames. If
2544 not set, jobs with identical names will share the log filename. Default:
2547 .. option:: group_reporting
2549 It may sometimes be interesting to display statistics for groups of jobs as
2550 a whole instead of for each individual job. This is especially true if
2551 :option:`numjobs` is used; looking at individual thread/process output
2552 quickly becomes unwieldy. To see the final report per-group instead of
2553 per-job, use :option:`group_reporting`. Jobs in a file will be part of the
2554 same reporting group, unless if separated by a :option:`stonewall`, or by
2555 using :option:`new_group`.
2557 .. option:: new_group
2559 Start a new reporting group. See: :option:`group_reporting`. If not given,
2560 all jobs in a file will be part of the same reporting group, unless
2561 separated by a :option:`stonewall`.
2563 .. option:: write_bw_log=str
2565 If given, write a bandwidth log for this job. Can be used to store data of
2566 the bandwidth of the jobs in their lifetime. The included
2567 :command:`fio_generate_plots` script uses :command:`gnuplot` to turn these
2568 text files into nice graphs. See :option:`write_lat_log` for behaviour of
2569 given filename. For this option, the postfix is :file:`_bw.x.log`, where `x`
2570 is the index of the job (`1..N`, where `N` is the number of jobs). If
2571 :option:`per_job_logs` is false, then the filename will not include the job
2572 index. See `Log File Formats`_.
2574 .. option:: write_lat_log=str
2576 Same as :option:`write_bw_log`, except that this option stores I/O
2577 submission, completion, and total latencies instead. If no filename is given
2578 with this option, the default filename of :file:`jobname_type.log` is
2579 used. Even if the filename is given, fio will still append the type of
2580 log. So if one specifies::
2584 The actual log names will be :file:`foo_slat.x.log`, :file:`foo_clat.x.log`,
2585 and :file:`foo_lat.x.log`, where `x` is the index of the job (1..N, where N
2586 is the number of jobs). This helps :command:`fio_generate_plot` find the
2587 logs automatically. If :option:`per_job_logs` is false, then the filename
2588 will not include the job index. See `Log File Formats`_.
2590 .. option:: write_hist_log=str
2592 Same as :option:`write_lat_log`, but writes I/O completion latency
2593 histograms. If no filename is given with this option, the default filename
2594 of :file:`jobname_clat_hist.x.log` is used, where `x` is the index of the
2595 job (1..N, where `N` is the number of jobs). Even if the filename is given,
2596 fio will still append the type of log. If :option:`per_job_logs` is false,
2597 then the filename will not include the job index. See `Log File Formats`_.
2599 .. option:: write_iops_log=str
2601 Same as :option:`write_bw_log`, but writes IOPS. If no filename is given
2602 with this option, the default filename of :file:`jobname_type.x.log` is
2603 used,where `x` is the index of the job (1..N, where `N` is the number of
2604 jobs). Even if the filename is given, fio will still append the type of
2605 log. If :option:`per_job_logs` is false, then the filename will not include
2606 the job index. See `Log File Formats`_.
2608 .. option:: log_avg_msec=int
2610 By default, fio will log an entry in the iops, latency, or bw log for every
2611 I/O that completes. When writing to the disk log, that can quickly grow to a
2612 very large size. Setting this option makes fio average the each log entry
2613 over the specified period of time, reducing the resolution of the log. See
2614 :option:`log_max_value` as well. Defaults to 0, logging all entries.
2616 .. option:: log_hist_msec=int
2618 Same as :option:`log_avg_msec`, but logs entries for completion latency
2619 histograms. Computing latency percentiles from averages of intervals using
2620 :option:`log_avg_msec` is inaccurate. Setting this option makes fio log
2621 histogram entries over the specified period of time, reducing log sizes for
2622 high IOPS devices while retaining percentile accuracy. See
2623 :option:`log_hist_coarseness` as well. Defaults to 0, meaning histogram
2624 logging is disabled.
2626 .. option:: log_hist_coarseness=int
2628 Integer ranging from 0 to 6, defining the coarseness of the resolution of
2629 the histogram logs enabled with :option:`log_hist_msec`. For each increment
2630 in coarseness, fio outputs half as many bins. Defaults to 0, for which
2631 histogram logs contain 1216 latency bins. See `Log File Formats`_.
2633 .. option:: log_max_value=bool
2635 If :option:`log_avg_msec` is set, fio logs the average over that window. If
2636 you instead want to log the maximum value, set this option to 1. Defaults to
2637 0, meaning that averaged values are logged.
2639 .. option:: log_offset=int
2641 If this is set, the iolog options will include the byte offset for the I/O
2642 entry as well as the other data values.
2644 .. option:: log_compression=int
2646 If this is set, fio will compress the I/O logs as it goes, to keep the
2647 memory footprint lower. When a log reaches the specified size, that chunk is
2648 removed and compressed in the background. Given that I/O logs are fairly
2649 highly compressible, this yields a nice memory savings for longer runs. The
2650 downside is that the compression will consume some background CPU cycles, so
2651 it may impact the run. This, however, is also true if the logging ends up
2652 consuming most of the system memory. So pick your poison. The I/O logs are
2653 saved normally at the end of a run, by decompressing the chunks and storing
2654 them in the specified log file. This feature depends on the availability of
2657 .. option:: log_compression_cpus=str
2659 Define the set of CPUs that are allowed to handle online log compression for
2660 the I/O jobs. This can provide better isolation between performance
2661 sensitive jobs, and background compression work.
2663 .. option:: log_store_compressed=bool
2665 If set, fio will store the log files in a compressed format. They can be
2666 decompressed with fio, using the :option:`--inflate-log` command line
2667 parameter. The files will be stored with a :file:`.fz` suffix.
2669 .. option:: log_unix_epoch=bool
2671 If set, fio will log Unix timestamps to the log files produced by enabling
2672 write_type_log for each log type, instead of the default zero-based
2675 .. option:: block_error_percentiles=bool
2677 If set, record errors in trim block-sized units from writes and trims and
2678 output a histogram of how many trims it took to get to errors, and what kind
2679 of error was encountered.
2681 .. option:: bwavgtime=int
2683 Average the calculated bandwidth over the given time. Value is specified in
2684 milliseconds. If the job also does bandwidth logging through
2685 :option:`write_bw_log`, then the minimum of this option and
2686 :option:`log_avg_msec` will be used. Default: 500ms.
2688 .. option:: iopsavgtime=int
2690 Average the calculated IOPS over the given time. Value is specified in
2691 milliseconds. If the job also does IOPS logging through
2692 :option:`write_iops_log`, then the minimum of this option and
2693 :option:`log_avg_msec` will be used. Default: 500ms.
2695 .. option:: disk_util=bool
2697 Generate disk utilization statistics, if the platform supports it.
2700 .. option:: disable_lat=bool
2702 Disable measurements of total latency numbers. Useful only for cutting back
2703 the number of calls to :manpage:`gettimeofday(2)`, as that does impact
2704 performance at really high IOPS rates. Note that to really get rid of a
2705 large amount of these calls, this option must be used with
2706 :option:`disable_slat` and :option:`disable_bw_measurement` as well.
2708 .. option:: disable_clat=bool
2710 Disable measurements of completion latency numbers. See
2711 :option:`disable_lat`.
2713 .. option:: disable_slat=bool
2715 Disable measurements of submission latency numbers. See
2716 :option:`disable_slat`.
2718 .. option:: disable_bw_measurement=bool, disable_bw=bool
2720 Disable measurements of throughput/bandwidth numbers. See
2721 :option:`disable_lat`.
2723 .. option:: clat_percentiles=bool
2725 Enable the reporting of percentiles of completion latencies.
2727 .. option:: percentile_list=float_list
2729 Overwrite the default list of percentiles for completion latencies and the
2730 block error histogram. Each number is a floating number in the range
2731 (0,100], and the maximum length of the list is 20. Use ``:`` to separate the
2732 numbers, and list the numbers in ascending order. For example,
2733 ``--percentile_list=99.5:99.9`` will cause fio to report the values of
2734 completion latency below which 99.5% and 99.9% of the observed latencies
2741 .. option:: exitall_on_error
2743 When one job finishes in error, terminate the rest. The default is to wait
2744 for each job to finish.
2746 .. option:: continue_on_error=str
2748 Normally fio will exit the job on the first observed failure. If this option
2749 is set, fio will continue the job when there is a 'non-fatal error' (EIO or
2750 EILSEQ) until the runtime is exceeded or the I/O size specified is
2751 completed. If this option is used, there are two more stats that are
2752 appended, the total error count and the first error. The error field given
2753 in the stats is the first error that was hit during the run.
2755 The allowed values are:
2758 Exit on any I/O or verify errors.
2761 Continue on read errors, exit on all others.
2764 Continue on write errors, exit on all others.
2767 Continue on any I/O error, exit on all others.
2770 Continue on verify errors, exit on all others.
2773 Continue on all errors.
2776 Backward-compatible alias for 'none'.
2779 Backward-compatible alias for 'all'.
2781 .. option:: ignore_error=str
2783 Sometimes you want to ignore some errors during test in that case you can
2784 specify error list for each error type.
2785 ``ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST`` errors for
2786 given error type is separated with ':'. Error may be symbol ('ENOSPC',
2787 'ENOMEM') or integer. Example::
2789 ignore_error=EAGAIN,ENOSPC:122
2791 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from
2794 .. option:: error_dump=bool
2796 If set dump every error even if it is non fatal, true by default. If
2797 disabled only fatal error will be dumped.
2799 Running predefined workloads
2800 ----------------------------
2802 Fio includes predefined profiles that mimic the I/O workloads generated by
2805 .. option:: profile=str
2807 The predefined workload to run. Current profiles are:
2810 Threaded I/O bench (tiotest/tiobench) like workload.
2813 Aerospike Certification Tool (ACT) like workload.
2815 To view a profile's additional options use :option:`--cmdhelp` after specifying
2816 the profile. For example::
2818 $ fio --profile=act --cmdhelp
2823 .. option:: device-names=str
2828 .. option:: load=int
2831 ACT load multiplier. Default: 1.
2833 .. option:: test-duration=time
2836 How long the entire test takes to run. Default: 24h.
2838 .. option:: threads-per-queue=int
2841 Number of read IO threads per device. Default: 8.
2843 .. option:: read-req-num-512-blocks=int
2846 Number of 512B blocks to read at the time. Default: 3.
2848 .. option:: large-block-op-kbytes=int
2851 Size of large block ops in KiB (writes). Default: 131072.
2856 Set to run ACT prep phase.
2858 Tiobench profile options
2859 ~~~~~~~~~~~~~~~~~~~~~~~~
2861 .. option:: size=str
2866 .. option:: block=int
2869 Block size in bytes. Default: 4096.
2871 .. option:: numruns=int
2881 .. option:: threads=int
2886 Interpreting the output
2887 -----------------------
2889 Fio spits out a lot of output. While running, fio will display the status of the
2890 jobs created. An example of that would be::
2892 Jobs: 1: [_r] [24.8% done] [r=20992KiB/s,w=24064KiB/s,t=0KiB/s] [r=82,w=94,t=0 iops] [eta 00h:01m:31s]
2894 The characters inside the square brackets denote the current status of each
2895 thread. The possible values (in typical life cycle order) are:
2897 +------+-----+-----------------------------------------------------------+
2899 +======+=====+===========================================================+
2900 | P | | Thread setup, but not started. |
2901 +------+-----+-----------------------------------------------------------+
2902 | C | | Thread created. |
2903 +------+-----+-----------------------------------------------------------+
2904 | I | | Thread initialized, waiting or generating necessary data. |
2905 +------+-----+-----------------------------------------------------------+
2906 | | p | Thread running pre-reading file(s). |
2907 +------+-----+-----------------------------------------------------------+
2908 | | R | Running, doing sequential reads. |
2909 +------+-----+-----------------------------------------------------------+
2910 | | r | Running, doing random reads. |
2911 +------+-----+-----------------------------------------------------------+
2912 | | W | Running, doing sequential writes. |
2913 +------+-----+-----------------------------------------------------------+
2914 | | w | Running, doing random writes. |
2915 +------+-----+-----------------------------------------------------------+
2916 | | M | Running, doing mixed sequential reads/writes. |
2917 +------+-----+-----------------------------------------------------------+
2918 | | m | Running, doing mixed random reads/writes. |
2919 +------+-----+-----------------------------------------------------------+
2920 | | F | Running, currently waiting for :manpage:`fsync(2)` |
2921 +------+-----+-----------------------------------------------------------+
2922 | | V | Running, doing verification of written data. |
2923 +------+-----+-----------------------------------------------------------+
2924 | E | | Thread exited, not reaped by main thread yet. |
2925 +------+-----+-----------------------------------------------------------+
2926 | _ | | Thread reaped, or |
2927 +------+-----+-----------------------------------------------------------+
2928 | X | | Thread reaped, exited with an error. |
2929 +------+-----+-----------------------------------------------------------+
2930 | K | | Thread reaped, exited due to signal. |
2931 +------+-----+-----------------------------------------------------------+
2933 Fio will condense the thread string as not to take up more space on the command
2934 line as is needed. For instance, if you have 10 readers and 10 writers running,
2935 the output would look like this::
2937 Jobs: 20 (f=20): [R(10),W(10)] [4.0% done] [r=20992KiB/s,w=24064KiB/s,t=0KiB/s] [r=82,w=94,t=0 iops] [eta 57m:36s]
2939 Fio will still maintain the ordering, though. So the above means that jobs 1..10
2940 are readers, and 11..20 are writers.
2942 The other values are fairly self explanatory -- number of threads currently
2943 running and doing I/O, rate of I/O since last check (read speed listed first,
2944 then write speed), and the estimated completion percentage and time for the
2945 running group. It's impossible to estimate runtime of the following groups (if
2946 any). Note that the string is displayed in order, so it's possible to tell which
2947 of the jobs are currently doing what. The first character is the first job
2948 defined in the job file, and so forth.
2950 When fio is done (or interrupted by :kbd:`ctrl-c`), it will show the data for
2951 each thread, group of threads, and disks in that order. For each data direction,
2952 the output looks like::
2954 Client1 (g=0): err= 0:
2955 write: io= 32MiB, bw= 666KiB/s, iops=89 , runt= 50320msec
2956 slat (msec): min= 0, max= 136, avg= 0.03, stdev= 1.92
2957 clat (msec): min= 0, max= 631, avg=48.50, stdev=86.82
2958 bw (KiB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, stdev=681.68
2959 cpu : usr=1.49%, sys=0.25%, ctx=7969, majf=0, minf=17
2960 IO depths : 1=0.1%, 2=0.3%, 4=0.5%, 8=99.0%, 16=0.0%, 32=0.0%, >32=0.0%
2961 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
2962 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
2963 issued r/w: total=0/32768, short=0/0
2964 lat (msec): 2=1.6%, 4=0.0%, 10=3.2%, 20=12.8%, 50=38.4%, 100=24.8%,
2965 lat (msec): 250=15.2%, 500=0.0%, 750=0.0%, 1000=0.0%, >=2048=0.0%
2967 The client number is printed, along with the group id and error of that
2968 thread. Below is the I/O statistics, here for writes. In the order listed, they
2972 Number of megabytes I/O performed.
2975 Average bandwidth rate.
2978 Average I/Os performed per second.
2981 The runtime of that thread.
2984 Submission latency (avg being the average, stdev being the standard
2985 deviation). This is the time it took to submit the I/O. For sync I/O,
2986 the slat is really the completion latency, since queue/complete is one
2987 operation there. This value can be in milliseconds or microseconds, fio
2988 will choose the most appropriate base and print that. In the example
2989 above, milliseconds is the best scale. Note: in :option:`--minimal` mode
2990 latencies are always expressed in microseconds.
2993 Completion latency. Same names as slat, this denotes the time from
2994 submission to completion of the I/O pieces. For sync I/O, clat will
2995 usually be equal (or very close) to 0, as the time from submit to
2996 complete is basically just CPU time (I/O has already been done, see slat
3000 Bandwidth. Same names as the xlat stats, but also includes an
3001 approximate percentage of total aggregate bandwidth this thread received
3002 in this group. This last value is only really useful if the threads in
3003 this group are on the same disk, since they are then competing for disk
3007 CPU usage. User and system time, along with the number of context
3008 switches this thread went through, usage of system and user time, and
3009 finally the number of major and minor page faults. The CPU utilization
3010 numbers are averages for the jobs in that reporting group, while the
3011 context and fault counters are summed.
3014 The distribution of I/O depths over the job life time. The numbers are
3015 divided into powers of 2, so for example the 16= entries includes depths
3016 up to that value but higher than the previous entry. In other words, it
3017 covers the range from 16 to 31.
3020 How many pieces of I/O were submitting in a single submit call. Each
3021 entry denotes that amount and below, until the previous entry -- e.g.,
3022 8=100% mean that we submitted anywhere in between 5-8 I/Os per submit
3026 Like the above submit number, but for completions instead.
3029 The number of read/write requests issued, and how many of them were
3033 The distribution of I/O completion latencies. This is the time from when
3034 I/O leaves fio and when it gets completed. The numbers follow the same
3035 pattern as the I/O depths, meaning that 2=1.6% means that 1.6% of the
3036 I/O completed within 2 msecs, 20=12.8% means that 12.8% of the I/O took
3037 more than 10 msecs, but less than (or equal to) 20 msecs.
3039 After each client has been listed, the group statistics are printed. They
3040 will look like this::
3042 Run status group 0 (all jobs):
3043 READ: io=64MB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
3044 WRITE: io=64MB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
3046 For each data direction, it prints:
3049 Number of megabytes I/O performed.
3051 Aggregate bandwidth of threads in this group.
3053 The minimum average bandwidth a thread saw.
3055 The maximum average bandwidth a thread saw.
3057 The smallest runtime of the threads in that group.
3059 The longest runtime of the threads in that group.
3061 And finally, the disk statistics are printed. They will look like this::
3063 Disk stats (read/write):
3064 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
3066 Each value is printed for both reads and writes, with reads first. The
3070 Number of I/Os performed by all groups.
3072 Number of merges I/O the I/O scheduler.
3074 Number of ticks we kept the disk busy.
3076 Total time spent in the disk queue.
3078 The disk utilization. A value of 100% means we kept the disk
3079 busy constantly, 50% would be a disk idling half of the time.
3081 It is also possible to get fio to dump the current output while it is running,
3082 without terminating the job. To do that, send fio the **USR1** signal. You can
3083 also get regularly timed dumps by using the :option:`--status-interval`
3084 parameter, or by creating a file in :file:`/tmp` named
3085 :file:`fio-dump-status`. If fio sees this file, it will unlink it and dump the
3086 current output status.
3092 For scripted usage where you typically want to generate tables or graphs of the
3093 results, fio can output the results in a semicolon separated format. The format
3094 is one long line of values, such as::
3096 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%
3097 A description of this job goes here.
3099 The job description (if provided) follows on a second line.
3101 To enable terse output, use the :option:`--minimal` command line option. The
3102 first value is the version of the terse output format. If the output has to be
3103 changed for some reason, this number will be incremented by 1 to signify that
3106 Split up, the format is as follows:
3110 terse version, fio version, jobname, groupid, error
3114 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3115 Submission latency: min, max, mean, stdev (usec)
3116 Completion latency: min, max, mean, stdev (usec)
3117 Completion latency percentiles: 20 fields (see below)
3118 Total latency: min, max, mean, stdev (usec)
3119 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev
3125 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3126 Submission latency: min, max, mean, stdev (usec)
3127 Completion latency: min, max, mean, stdev(usec)
3128 Completion latency percentiles: 20 fields (see below)
3129 Total latency: min, max, mean, stdev (usec)
3130 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev
3134 user, system, context switches, major faults, minor faults
3138 <=1, 2, 4, 8, 16, 32, >=64
3140 I/O latencies microseconds::
3142 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
3144 I/O latencies milliseconds::
3146 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
3150 Disk name, Read ios, write ios,
3151 Read merges, write merges,
3152 Read ticks, write ticks,
3153 Time spent in queue, disk utilization percentage
3155 Additional Info (dependent on continue_on_error, default off)::
3157 total # errors, first error code
3159 Additional Info (dependent on description being set)::
3163 Completion latency percentiles can be a grouping of up to 20 sets, so for the
3164 terse output fio writes all of them. Each field will look like this::
3168 which is the Xth percentile, and the `usec` latency associated with it.
3170 For disk utilization, all disks used by fio are shown. So for each disk there
3171 will be a disk utilization section.
3177 There are two trace file format that you can encounter. The older (v1) format is
3178 unsupported since version 1.20-rc3 (March 2008). It will still be described
3179 below in case that you get an old trace and want to understand it.
3181 In any case the trace is a simple text file with a single action per line.
3184 Trace file format v1
3185 ~~~~~~~~~~~~~~~~~~~~
3187 Each line represents a single I/O action in the following format::
3191 where `rw=0/1` for read/write, and the offset and length entries being in bytes.
3193 This format is not supported in fio versions => 1.20-rc3.
3196 Trace file format v2
3197 ~~~~~~~~~~~~~~~~~~~~
3199 The second version of the trace file format was added in fio version 1.17. It
3200 allows to access more then one file per trace and has a bigger set of possible
3203 The first line of the trace file has to be::
3207 Following this can be lines in two different formats, which are described below.
3209 The file management format::
3213 The filename is given as an absolute path. The action can be one of these:
3216 Add the given filename to the trace.
3218 Open the file with the given filename. The filename has to have
3219 been added with the **add** action before.
3221 Close the file with the given filename. The file has to have been
3225 The file I/O action format::
3227 filename action offset length
3229 The `filename` is given as an absolute path, and has to have been added and
3230 opened before it can be used with this format. The `offset` and `length` are
3231 given in bytes. The `action` can be one of these:
3234 Wait for `offset` microseconds. Everything below 100 is discarded.
3235 The time is relative to the previous `wait` statement.
3237 Read `length` bytes beginning from `offset`.
3239 Write `length` bytes beginning from `offset`.
3241 :manpage:`fsync(2)` the file.
3243 :manpage:`fdatasync(2)` the file.
3245 Trim the given file from the given `offset` for `length` bytes.
3247 CPU idleness profiling
3248 ----------------------
3250 In some cases, we want to understand CPU overhead in a test. For example, we
3251 test patches for the specific goodness of whether they reduce CPU usage.
3252 Fio implements a balloon approach to create a thread per CPU that runs at idle
3253 priority, meaning that it only runs when nobody else needs the cpu.
3254 By measuring the amount of work completed by the thread, idleness of each CPU
3255 can be derived accordingly.
3257 An unit work is defined as touching a full page of unsigned characters. Mean and
3258 standard deviation of time to complete an unit work is reported in "unit work"
3259 section. Options can be chosen to report detailed percpu idleness or overall
3260 system idleness by aggregating percpu stats.
3263 Verification and triggers
3264 -------------------------
3266 Fio is usually run in one of two ways, when data verification is done. The first
3267 is a normal write job of some sort with verify enabled. When the write phase has
3268 completed, fio switches to reads and verifies everything it wrote. The second
3269 model is running just the write phase, and then later on running the same job
3270 (but with reads instead of writes) to repeat the same I/O patterns and verify
3271 the contents. Both of these methods depend on the write phase being completed,
3272 as fio otherwise has no idea how much data was written.
3274 With verification triggers, fio supports dumping the current write state to
3275 local files. Then a subsequent read verify workload can load this state and know
3276 exactly where to stop. This is useful for testing cases where power is cut to a
3277 server in a managed fashion, for instance.
3279 A verification trigger consists of two things:
3281 1) Storing the write state of each job.
3282 2) Executing a trigger command.
3284 The write state is relatively small, on the order of hundreds of bytes to single
3285 kilobytes. It contains information on the number of completions done, the last X
3288 A trigger is invoked either through creation ('touch') of a specified file in
3289 the system, or through a timeout setting. If fio is run with
3290 :option:`--trigger-file` = :file:`/tmp/trigger-file`, then it will continually
3291 check for the existence of :file:`/tmp/trigger-file`. When it sees this file, it
3292 will fire off the trigger (thus saving state, and executing the trigger
3295 For client/server runs, there's both a local and remote trigger. If fio is
3296 running as a server backend, it will send the job states back to the client for
3297 safe storage, then execute the remote trigger, if specified. If a local trigger
3298 is specified, the server will still send back the write state, but the client
3299 will then execute the trigger.
3301 Verification trigger example
3302 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3304 Lets say we want to run a powercut test on the remote machine 'server'. Our
3305 write workload is in :file:`write-test.fio`. We want to cut power to 'server' at
3306 some point during the run, and we'll run this test from the safety or our local
3307 machine, 'localbox'. On the server, we'll start the fio backend normally::
3309 server# fio --server
3311 and on the client, we'll fire off the workload::
3313 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger-remote="bash -c \"echo b > /proc/sysrq-triger\""
3315 We set :file:`/tmp/my-trigger` as the trigger file, and we tell fio to execute::
3317 echo b > /proc/sysrq-trigger
3319 on the server once it has received the trigger and sent us the write state. This
3320 will work, but it's not **really** cutting power to the server, it's merely
3321 abruptly rebooting it. If we have a remote way of cutting power to the server
3322 through IPMI or similar, we could do that through a local trigger command
3323 instead. Lets assume we have a script that does IPMI reboot of a given hostname,
3324 ipmi-reboot. On localbox, we could then have run fio with a local trigger
3327 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger="ipmi-reboot server"
3329 For this case, fio would wait for the server to send us the write state, then
3330 execute ``ipmi-reboot server`` when that happened.
3332 Loading verify state
3333 ~~~~~~~~~~~~~~~~~~~~
3335 To load store write state, read verification job file must contain the
3336 :option:`verify_state_load` option. If that is set, fio will load the previously
3337 stored state. For a local fio run this is done by loading the files directly,
3338 and on a client/server run, the server backend will ask the client to send the
3339 files over and load them from there.
3345 Fio supports a variety of log file formats, for logging latencies, bandwidth,
3346 and IOPS. The logs share a common format, which looks like this:
3348 *time* (`msec`), *value*, *data direction*, *offset*
3350 Time for the log entry is always in milliseconds. The *value* logged depends
3351 on the type of log, it will be one of the following:
3354 Value is latency in usecs
3360 *Data direction* is one of the following:
3369 The *offset* is the offset, in bytes, from the start of the file, for that
3370 particular I/O. The logging of the offset can be toggled with
3371 :option:`log_offset`.
3373 If windowed logging is enabled through :option:`log_avg_msec` then fio doesn't
3374 log individual I/Os. Instead of logs the average values over the specified period
3375 of time. Since 'data direction' and 'offset' are per-I/O values, they aren't
3376 applicable if windowed logging is enabled. If windowed logging is enabled and
3377 :option:`log_max_value` is set, then fio logs maximum values in that window
3378 instead of averages.
3384 Normally fio is invoked as a stand-alone application on the machine where the
3385 I/O workload should be generated. However, the frontend and backend of fio can
3386 be run separately. Ie the fio server can generate an I/O workload on the "Device
3387 Under Test" while being controlled from another machine.
3389 Start the server on the machine which has access to the storage DUT::
3393 where args defines what fio listens to. The arguments are of the form
3394 ``type,hostname`` or ``IP,port``. *type* is either ``ip`` (or ip4) for TCP/IP
3395 v4, ``ip6`` for TCP/IP v6, or ``sock`` for a local unix domain socket.
3396 *hostname* is either a hostname or IP address, and *port* is the port to listen
3397 to (only valid for TCP/IP, not a local socket). Some examples:
3401 Start a fio server, listening on all interfaces on the default port (8765).
3403 2) ``fio --server=ip:hostname,4444``
3405 Start a fio server, listening on IP belonging to hostname and on port 4444.
3407 3) ``fio --server=ip6:::1,4444``
3409 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
3411 4) ``fio --server=,4444``
3413 Start a fio server, listening on all interfaces on port 4444.
3415 5) ``fio --server=1.2.3.4``
3417 Start a fio server, listening on IP 1.2.3.4 on the default port.
3419 6) ``fio --server=sock:/tmp/fio.sock``
3421 Start a fio server, listening on the local socket /tmp/fio.sock.
3423 Once a server is running, a "client" can connect to the fio server with::
3425 fio <local-args> --client=<server> <remote-args> <job file(s)>
3427 where `local-args` are arguments for the client where it is running, `server`
3428 is the connect string, and `remote-args` and `job file(s)` are sent to the
3429 server. The `server` string follows the same format as it does on the server
3430 side, to allow IP/hostname/socket and port strings.
3432 Fio can connect to multiple servers this way::
3434 fio --client=<server1> <job file(s)> --client=<server2> <job file(s)>
3436 If the job file is located on the fio server, then you can tell the server to
3437 load a local file as well. This is done by using :option:`--remote-config` ::
3439 fio --client=server --remote-config /path/to/file.fio
3441 Then fio will open this local (to the server) job file instead of being passed
3442 one from the client.
3444 If you have many servers (example: 100 VMs/containers), you can input a pathname
3445 of a file containing host IPs/names as the parameter value for the
3446 :option:`--client` option. For example, here is an example :file:`host.list`
3447 file containing 2 hostnames::
3449 host1.your.dns.domain
3450 host2.your.dns.domain
3452 The fio command would then be::
3454 fio --client=host.list <job file(s)>
3456 In this mode, you cannot input server-specific parameters or job files -- all
3457 servers receive the same job file.
3459 In order to let ``fio --client`` runs use a shared filesystem from multiple
3460 hosts, ``fio --client`` now prepends the IP address of the server to the
3461 filename. For example, if fio is using directory :file:`/mnt/nfs/fio` and is
3462 writing filename :file:`fileio.tmp`, with a :option:`--client` `hostfile`
3463 containing two hostnames ``h1`` and ``h2`` with IP addresses 192.168.10.120 and
3464 192.168.10.121, then fio will create two files::
3466 /mnt/nfs/fio/192.168.10.120.fileio.tmp
3467 /mnt/nfs/fio/192.168.10.121.fileio.tmp