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>`.
546 If the lower value specified happens to be larger than the upper value,
547 two values are swapped.
552 Boolean. Usually parsed as an integer, however only defined for
553 true and false (1 and 0).
558 Integer range with suffix. Allows value range to be given, such as
559 1024-4096. A colon may also be used as the separator, e.g. 1k:4k. If the
560 option allows two sets of ranges, they can be specified with a ',' or '/'
561 delimiter: 1k-4k/8k-32k. Also see :ref:`int <int>`.
564 A list of floating point numbers, separated by a ':' character.
570 .. option:: kb_base=int
572 Select the interpretation of unit prefixes in input parameters.
575 Inputs comply with IEC 80000-13 and the International
576 System of Units (SI). Use:
578 - power-of-2 values with IEC prefixes (e.g., KiB)
579 - power-of-10 values with SI prefixes (e.g., kB)
582 Compatibility mode (default). To avoid breaking old scripts:
584 - power-of-2 values with SI prefixes
585 - power-of-10 values with IEC prefixes
587 See :option:`bs` for more details on input parameters.
589 Outputs always use correct prefixes. Most outputs include both
592 bw=2383.3kB/s (2327.4KiB/s)
594 If only one value is reported, then kb_base selects the one to use:
596 **1000** -- SI prefixes
598 **1024** -- IEC prefixes
600 .. option:: unit_base=int
602 Base unit for reporting. Allowed values are:
605 Use auto-detection (default).
612 With the above in mind, here follows the complete list of fio job parameters.
620 ASCII name of the job. This may be used to override the name printed by fio
621 for this job. Otherwise the job name is used. On the command line this
622 parameter has the special purpose of also signaling the start of a new job.
624 .. option:: description=str
626 Text description of the job. Doesn't do anything except dump this text
627 description when this job is run. It's not parsed.
629 .. option:: loops=int
631 Run the specified number of iterations of this job. Used to repeat the same
632 workload a given number of times. Defaults to 1.
634 .. option:: numjobs=int
636 Create the specified number of clones of this job. Each clone of job
637 is spawned as an independent thread or process. May be used to setup a
638 larger number of threads/processes doing the same thing. Each thread is
639 reported separately; to see statistics for all clones as a whole, use
640 :option:`group_reporting` in conjunction with :option:`new_group`.
641 See :option:`--max-jobs`.
644 Time related parameters
645 ~~~~~~~~~~~~~~~~~~~~~~~
647 .. option:: runtime=time
649 Tell fio to terminate processing after the specified period of time. It
650 can be quite hard to determine for how long a specified job will run, so
651 this parameter is handy to cap the total runtime to a given time. When
652 the unit is omitted, the value is given in seconds.
654 .. option:: time_based
656 If set, fio will run for the duration of the :option:`runtime` specified
657 even if the file(s) are completely read or written. It will simply loop over
658 the same workload as many times as the :option:`runtime` allows.
660 .. option:: startdelay=irange(time)
662 Delay start of job for the specified number of seconds. Supports all time
663 suffixes to allow specification of hours, minutes, seconds and milliseconds
664 -- seconds are the default if a unit is omitted. Can be given as a range
665 which causes each thread to choose randomly out of the range.
667 .. option:: ramp_time=time
669 If set, fio will run the specified workload for this amount of time before
670 logging any performance numbers. Useful for letting performance settle
671 before logging results, thus minimizing the runtime required for stable
672 results. Note that the ``ramp_time`` is considered lead in time for a job,
673 thus it will increase the total runtime if a special timeout or
674 :option:`runtime` is specified. When the unit is omitted, the value is
677 .. option:: clocksource=str
679 Use the given clocksource as the base of timing. The supported options are:
682 :manpage:`gettimeofday(2)`
685 :manpage:`clock_gettime(2)`
688 Internal CPU clock source
690 cpu is the preferred clocksource if it is reliable, as it is very fast (and
691 fio is heavy on time calls). Fio will automatically use this clocksource if
692 it's supported and considered reliable on the system it is running on,
693 unless another clocksource is specifically set. For x86/x86-64 CPUs, this
694 means supporting TSC Invariant.
696 .. option:: gtod_reduce=bool
698 Enable all of the :manpage:`gettimeofday(2)` reducing options
699 (:option:`disable_clat`, :option:`disable_slat`, :option:`disable_bw_measurement`) plus
700 reduce precision of the timeout somewhat to really shrink the
701 :manpage:`gettimeofday(2)` call count. With this option enabled, we only do
702 about 0.4% of the :manpage:`gettimeofday(2)` calls we would have done if all
703 time keeping was enabled.
705 .. option:: gtod_cpu=int
707 Sometimes it's cheaper to dedicate a single thread of execution to just
708 getting the current time. Fio (and databases, for instance) are very
709 intensive on :manpage:`gettimeofday(2)` calls. With this option, you can set
710 one CPU aside for doing nothing but logging current time to a shared memory
711 location. Then the other threads/processes that run I/O workloads need only
712 copy that segment, instead of entering the kernel with a
713 :manpage:`gettimeofday(2)` call. The CPU set aside for doing these time
714 calls will be excluded from other uses. Fio will manually clear it from the
715 CPU mask of other jobs.
721 .. option:: directory=str
723 Prefix filenames with this directory. Used to place files in a different
724 location than :file:`./`. You can specify a number of directories by
725 separating the names with a ':' character. These directories will be
726 assigned equally distributed to job clones creates with :option:`numjobs` as
727 long as they are using generated filenames. If specific `filename(s)` are
728 set fio will use the first listed directory, and thereby matching the
729 `filename` semantic which generates a file each clone if not specified, but
730 let all clones use the same if set.
732 See the :option:`filename` option for escaping certain characters.
734 .. option:: filename=str
736 Fio normally makes up a `filename` based on the job name, thread number, and
737 file number. If you want to share files between threads in a job or several
738 jobs with fixed file paths, specify a `filename` for each of them to override
739 the default. If the ioengine is file based, you can specify a number of files
740 by separating the names with a ':' colon. So if you wanted a job to open
741 :file:`/dev/sda` and :file:`/dev/sdb` as the two working files, you would use
742 ``filename=/dev/sda:/dev/sdb``. This also means that whenever this option is
743 specified, :option:`nrfiles` is ignored. The size of regular files specified
744 by this option will be :option:`size` divided by number of files unless
745 explicit size is specified by :option:`filesize`.
747 On Windows, disk devices are accessed as :file:`\\\\.\\PhysicalDrive0` for
748 the first device, :file:`\\\\.\\PhysicalDrive1` for the second etc.
749 Note: Windows and FreeBSD prevent write access to areas
750 of the disk containing in-use data (e.g. filesystems). If the wanted
751 `filename` does need to include a colon, then escape that with a ``\``
752 character. For instance, if the `filename` is :file:`/dev/dsk/foo@3,0:c`,
753 then you would use ``filename="/dev/dsk/foo@3,0\:c"``. The
754 :file:`-` is a reserved name, meaning stdin or stdout. Which of the two
755 depends on the read/write direction set.
757 .. option:: filename_format=str
759 If sharing multiple files between jobs, it is usually necessary to have fio
760 generate the exact names that you want. By default, fio will name a file
761 based on the default file format specification of
762 :file:`jobname.jobnumber.filenumber`. With this option, that can be
763 customized. Fio will recognize and replace the following keywords in this
767 The name of the worker thread or process.
769 The incremental number of the worker thread or process.
771 The incremental number of the file for that worker thread or
774 To have dependent jobs share a set of files, this option can be set to have
775 fio generate filenames that are shared between the two. For instance, if
776 :file:`testfiles.$filenum` is specified, file number 4 for any job will be
777 named :file:`testfiles.4`. The default of :file:`$jobname.$jobnum.$filenum`
778 will be used if no other format specifier is given.
780 .. option:: unique_filename=bool
782 To avoid collisions between networked clients, fio defaults to prefixing any
783 generated filenames (with a directory specified) with the source of the
784 client connecting. To disable this behavior, set this option to 0.
786 .. option:: opendir=str
788 Recursively open any files below directory `str`.
790 .. option:: lockfile=str
792 Fio defaults to not locking any files before it does I/O to them. If a file
793 or file descriptor is shared, fio can serialize I/O to that file to make the
794 end result consistent. This is usual for emulating real workloads that share
795 files. The lock modes are:
798 No locking. The default.
800 Only one thread or process may do I/O at a time, excluding all
803 Read-write locking on the file. Many readers may
804 access the file at the same time, but writes get exclusive access.
806 .. option:: nrfiles=int
808 Number of files to use for this job. Defaults to 1. The size of files
809 will be :option:`size` divided by this unless explicit size is specified by
810 :option:`filesize`. Files are created for each thread separately, and each
811 file will have a file number within its name by default, as explained in
812 :option:`filename` section.
815 .. option:: openfiles=int
817 Number of files to keep open at the same time. Defaults to the same as
818 :option:`nrfiles`, can be set smaller to limit the number simultaneous
821 .. option:: file_service_type=str
823 Defines how fio decides which file from a job to service next. The following
827 Choose a file at random.
830 Round robin over opened files. This is the default.
833 Finish one file before moving on to the next. Multiple files can
834 still be open depending on 'openfiles'.
837 Use a *Zipf* distribution to decide what file to access.
840 Use a *Pareto* distribution to decide what file to access.
843 Use a *Gaussian* (normal) distribution to decide what file to
846 For *random*, *roundrobin*, and *sequential*, a postfix can be appended to
847 tell fio how many I/Os to issue before switching to a new file. For example,
848 specifying ``file_service_type=random:8`` would cause fio to issue
849 8 I/Os before selecting a new file at random. For the non-uniform
850 distributions, a floating point postfix can be given to influence how the
851 distribution is skewed. See :option:`random_distribution` for a description
852 of how that would work.
854 .. option:: ioscheduler=str
856 Attempt to switch the device hosting the file to the specified I/O scheduler
859 .. option:: create_serialize=bool
861 If true, serialize the file creation for the jobs. This may be handy to
862 avoid interleaving of data files, which may greatly depend on the filesystem
863 used and even the number of processors in the system.
865 .. option:: create_fsync=bool
867 fsync the data file after creation. This is the default.
869 .. option:: create_on_open=bool
871 Don't pre-setup the files for I/O, just create open() when it's time to do
874 .. option:: create_only=bool
876 If true, fio will only run the setup phase of the job. If files need to be
877 laid out or updated on disk, only that will be done. The actual job contents
880 .. option:: allow_file_create=bool
882 If true, fio is permitted to create files as part of its workload. This is
883 the default behavior. If this option is false, then fio will error out if
884 the files it needs to use don't already exist. Default: true.
886 .. option:: allow_mounted_write=bool
888 If this isn't set, fio will abort jobs that are destructive (e.g. that write)
889 to what appears to be a mounted device or partition. This should help catch
890 creating inadvertently destructive tests, not realizing that the test will
891 destroy data on the mounted file system. Note that some platforms don't allow
892 writing against a mounted device regardless of this option. Default: false.
894 .. option:: pre_read=bool
896 If this is given, files will be pre-read into memory before starting the
897 given I/O operation. This will also clear the :option:`invalidate` flag,
898 since it is pointless to pre-read and then drop the cache. This will only
899 work for I/O engines that are seek-able, since they allow you to read the
900 same data multiple times. Thus it will not work on e.g. network or splice I/O.
902 .. option:: unlink=bool
904 Unlink the job files when done. Not the default, as repeated runs of that
905 job would then waste time recreating the file set again and again.
907 .. option:: unlink_each_loop=bool
909 Unlink job files after each iteration or loop.
911 .. option:: zonesize=int
913 Divide a file into zones of the specified size. See :option:`zoneskip`.
915 .. option:: zonerange=int
917 Give size of an I/O zone. See :option:`zoneskip`.
919 .. option:: zoneskip=int
921 Skip the specified number of bytes when :option:`zonesize` data has been
922 read. The two zone options can be used to only do I/O on zones of a file.
928 .. option:: direct=bool
930 If value is true, use non-buffered I/O. This is usually O_DIRECT. Note that
931 ZFS on Solaris doesn't support direct I/O. On Windows the synchronous
932 ioengines don't support direct I/O. Default: false.
934 .. option:: atomic=bool
936 If value is true, attempt to use atomic direct I/O. Atomic writes are
937 guaranteed to be stable once acknowledged by the operating system. Only
938 Linux supports O_ATOMIC right now.
940 .. option:: buffered=bool
942 If value is true, use buffered I/O. This is the opposite of the
943 :option:`direct` option. Defaults to true.
945 .. option:: readwrite=str, rw=str
947 Type of I/O pattern. Accepted values are:
954 Sequential trims (Linux block devices only).
960 Random trims (Linux block devices only).
962 Sequential mixed reads and writes.
964 Random mixed reads and writes.
966 Sequential trim+write sequences. Blocks will be trimmed first,
967 then the same blocks will be written to.
969 Fio defaults to read if the option is not specified. For the mixed I/O
970 types, the default is to split them 50/50. For certain types of I/O the
971 result may still be skewed a bit, since the speed may be different. It is
972 possible to specify a number of I/O's to do before getting a new offset,
973 this is done by appending a ``:<nr>`` to the end of the string given. For a
974 random read, it would look like ``rw=randread:8`` for passing in an offset
975 modifier with a value of 8. If the suffix is used with a sequential I/O
976 pattern, then the value specified will be added to the generated offset for
977 each I/O. For instance, using ``rw=write:4k`` will skip 4k for every
978 write. It turns sequential I/O into sequential I/O with holes. See the
979 :option:`rw_sequencer` option.
981 .. option:: rw_sequencer=str
983 If an offset modifier is given by appending a number to the ``rw=<str>``
984 line, then this option controls how that number modifies the I/O offset
985 being generated. Accepted values are:
988 Generate sequential offset.
990 Generate the same offset.
992 ``sequential`` is only useful for random I/O, where fio would normally
993 generate a new random offset for every I/O. If you append e.g. 8 to randread,
994 you would get a new random offset for every 8 I/O's. The result would be a
995 seek for only every 8 I/O's, instead of for every I/O. Use ``rw=randread:8``
996 to specify that. As sequential I/O is already sequential, setting
997 ``sequential`` for that would not result in any differences. ``identical``
998 behaves in a similar fashion, except it sends the same offset 8 number of
999 times before generating a new offset.
1001 .. option:: unified_rw_reporting=bool
1003 Fio normally reports statistics on a per data direction basis, meaning that
1004 reads, writes, and trims are accounted and reported separately. If this
1005 option is set fio sums the results and report them as "mixed" instead.
1007 .. option:: randrepeat=bool
1009 Seed the random number generator used for random I/O patterns in a
1010 predictable way so the pattern is repeatable across runs. Default: true.
1012 .. option:: allrandrepeat=bool
1014 Seed all random number generators in a predictable way so results are
1015 repeatable across runs. Default: false.
1017 .. option:: randseed=int
1019 Seed the random number generators based on this seed value, to be able to
1020 control what sequence of output is being generated. If not set, the random
1021 sequence depends on the :option:`randrepeat` setting.
1023 .. option:: fallocate=str
1025 Whether pre-allocation is performed when laying down files.
1026 Accepted values are:
1029 Do not pre-allocate space.
1032 Pre-allocate via :manpage:`posix_fallocate(3)`.
1035 Pre-allocate via :manpage:`fallocate(2)` with
1036 FALLOC_FL_KEEP_SIZE set.
1039 Backward-compatible alias for **none**.
1042 Backward-compatible alias for **posix**.
1044 May not be available on all supported platforms. **keep** is only available
1045 on Linux. If using ZFS on Solaris this must be set to **none** because ZFS
1046 doesn't support it. Default: **posix**.
1048 .. option:: fadvise_hint=str
1050 Use :manpage:`posix_fadvise(2)` to advise the kernel on what I/O patterns
1051 are likely to be issued. Accepted values are:
1054 Backwards-compatible hint for "no hint".
1057 Backwards compatible hint for "advise with fio workload type". This
1058 uses **FADV_RANDOM** for a random workload, and **FADV_SEQUENTIAL**
1059 for a sequential workload.
1062 Advise using **FADV_SEQUENTIAL**.
1065 Advise using **FADV_RANDOM**.
1067 .. option:: fadvise_stream=int
1069 Use :manpage:`posix_fadvise(2)` to advise the kernel what stream ID the
1070 writes issued belong to. Only supported on Linux. Note, this option may
1071 change going forward.
1073 .. option:: offset=int
1075 Start I/O at the provided offset in the file, given as either a fixed size or
1076 a percentage. If a percentage is given, the next ``blockalign``-ed offset
1077 will be used. Data before the given offset will not be touched. This
1078 effectively caps the file size at `real_size - offset`. Can be combined with
1079 :option:`size` to constrain the start and end range of the I/O workload.
1081 .. option:: offset_increment=int
1083 If this is provided, then the real offset becomes `offset + offset_increment
1084 * thread_number`, where the thread number is a counter that starts at 0 and
1085 is incremented for each sub-job (i.e. when :option:`numjobs` option is
1086 specified). This option is useful if there are several jobs which are
1087 intended to operate on a file in parallel disjoint segments, with even
1088 spacing between the starting points.
1090 .. option:: number_ios=int
1092 Fio will normally perform I/Os until it has exhausted the size of the region
1093 set by :option:`size`, or if it exhaust the allocated time (or hits an error
1094 condition). With this setting, the range/size can be set independently of
1095 the number of I/Os to perform. When fio reaches this number, it will exit
1096 normally and report status. Note that this does not extend the amount of I/O
1097 that will be done, it will only stop fio if this condition is met before
1098 other end-of-job criteria.
1100 .. option:: fsync=int
1102 If writing to a file, issue a sync of the dirty data for every number of
1103 blocks given. For example, if you give 32 as a parameter, fio will sync the
1104 file for every 32 writes issued. If fio is using non-buffered I/O, we may
1105 not sync the file. The exception is the sg I/O engine, which synchronizes
1106 the disk cache anyway. Defaults to 0, which means no sync every certain
1109 .. option:: fdatasync=int
1111 Like :option:`fsync` but uses :manpage:`fdatasync(2)` to only sync data and
1112 not metadata blocks. In Windows, FreeBSD, and DragonFlyBSD there is no
1113 :manpage:`fdatasync(2)`, this falls back to using :manpage:`fsync(2)`.
1114 Defaults to 0, which means no sync data every certain number of writes.
1116 .. option:: write_barrier=int
1118 Make every `N-th` write a barrier write.
1120 .. option:: sync_file_range=str:val
1122 Use :manpage:`sync_file_range(2)` for every `val` number of write
1123 operations. Fio will track range of writes that have happened since the last
1124 :manpage:`sync_file_range(2)` call. `str` can currently be one or more of:
1127 SYNC_FILE_RANGE_WAIT_BEFORE
1129 SYNC_FILE_RANGE_WRITE
1131 SYNC_FILE_RANGE_WAIT_AFTER
1133 So if you do ``sync_file_range=wait_before,write:8``, fio would use
1134 ``SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE`` for every 8
1135 writes. Also see the :manpage:`sync_file_range(2)` man page. This option is
1138 .. option:: overwrite=bool
1140 If true, writes to a file will always overwrite existing data. If the file
1141 doesn't already exist, it will be created before the write phase begins. If
1142 the file exists and is large enough for the specified write phase, nothing
1145 .. option:: end_fsync=bool
1147 If true, fsync file contents when a write stage has completed.
1149 .. option:: fsync_on_close=bool
1151 If true, fio will :manpage:`fsync(2)` a dirty file on close. This differs
1152 from end_fsync in that it will happen on every file close, not just at the
1155 .. option:: rwmixread=int
1157 Percentage of a mixed workload that should be reads. Default: 50.
1159 .. option:: rwmixwrite=int
1161 Percentage of a mixed workload that should be writes. If both
1162 :option:`rwmixread` and :option:`rwmixwrite` is given and the values do not
1163 add up to 100%, the latter of the two will be used to override the
1164 first. This may interfere with a given rate setting, if fio is asked to
1165 limit reads or writes to a certain rate. If that is the case, then the
1166 distribution may be skewed. Default: 50.
1168 .. option:: random_distribution=str:float[,str:float][,str:float]
1170 By default, fio will use a completely uniform random distribution when asked
1171 to perform random I/O. Sometimes it is useful to skew the distribution in
1172 specific ways, ensuring that some parts of the data is more hot than others.
1173 fio includes the following distribution models:
1176 Uniform random distribution
1185 Normal (Gaussian) distribution
1188 Zoned random distribution
1190 When using a **zipf** or **pareto** distribution, an input value is also
1191 needed to define the access pattern. For **zipf**, this is the `zipf
1192 theta`. For **pareto**, it's the `Pareto power`. Fio includes a test
1193 program, :command:`genzipf`, that can be used visualize what the given input
1194 values will yield in terms of hit rates. If you wanted to use **zipf** with
1195 a `theta` of 1.2, you would use ``random_distribution=zipf:1.2`` as the
1196 option. If a non-uniform model is used, fio will disable use of the random
1197 map. For the **gauss** distribution, a normal deviation is supplied as a
1198 value between 0 and 100.
1200 For a **zoned** distribution, fio supports specifying percentages of I/O
1201 access that should fall within what range of the file or device. For
1202 example, given a criteria of:
1204 * 60% of accesses should be to the first 10%
1205 * 30% of accesses should be to the next 20%
1206 * 8% of accesses should be to to the next 30%
1207 * 2% of accesses should be to the next 40%
1209 we can define that through zoning of the random accesses. For the above
1210 example, the user would do::
1212 random_distribution=zoned:60/10:30/20:8/30:2/40
1214 similarly to how :option:`bssplit` works for setting ranges and percentages
1215 of block sizes. Like :option:`bssplit`, it's possible to specify separate
1216 zones for reads, writes, and trims. If just one set is given, it'll apply to
1219 .. option:: percentage_random=int[,int][,int]
1221 For a random workload, set how big a percentage should be random. This
1222 defaults to 100%, in which case the workload is fully random. It can be set
1223 from anywhere from 0 to 100. Setting it to 0 would make the workload fully
1224 sequential. Any setting in between will result in a random mix of sequential
1225 and random I/O, at the given percentages. Comma-separated values may be
1226 specified for reads, writes, and trims as described in :option:`blocksize`.
1228 .. option:: norandommap
1230 Normally fio will cover every block of the file when doing random I/O. If
1231 this option is given, fio will just get a new random offset without looking
1232 at past I/O history. This means that some blocks may not be read or written,
1233 and that some blocks may be read/written more than once. If this option is
1234 used with :option:`verify` and multiple blocksizes (via :option:`bsrange`),
1235 only intact blocks are verified, i.e., partially-overwritten blocks are
1238 .. option:: softrandommap=bool
1240 See :option:`norandommap`. If fio runs with the random block map enabled and
1241 it fails to allocate the map, if this option is set it will continue without
1242 a random block map. As coverage will not be as complete as with random maps,
1243 this option is disabled by default.
1245 .. option:: random_generator=str
1247 Fio supports the following engines for generating
1248 I/O offsets for random I/O:
1251 Strong 2^88 cycle random number generator
1253 Linear feedback shift register generator
1255 Strong 64-bit 2^258 cycle random number generator
1257 **tausworthe** is a strong random number generator, but it requires tracking
1258 on the side if we want to ensure that blocks are only read or written
1259 once. **LFSR** guarantees that we never generate the same offset twice, and
1260 it's also less computationally expensive. It's not a true random generator,
1261 however, though for I/O purposes it's typically good enough. **LFSR** only
1262 works with single block sizes, not with workloads that use multiple block
1263 sizes. If used with such a workload, fio may read or write some blocks
1264 multiple times. The default value is **tausworthe**, unless the required
1265 space exceeds 2^32 blocks. If it does, then **tausworthe64** is
1266 selected automatically.
1272 .. option:: blocksize=int[,int][,int], bs=int[,int][,int]
1274 The block size in bytes used for I/O units. Default: 4096. A single value
1275 applies to reads, writes, and trims. Comma-separated values may be
1276 specified for reads, writes, and trims. A value not terminated in a comma
1277 applies to subsequent types.
1282 means 256k for reads, writes and trims.
1285 means 8k for reads, 32k for writes and trims.
1288 means 8k for reads, 32k for writes, and default for trims.
1291 means default for reads, 8k for writes and trims.
1294 means default for reads, 8k for writes, and default for trims.
1296 .. option:: blocksize_range=irange[,irange][,irange], bsrange=irange[,irange][,irange]
1298 A range of block sizes in bytes for I/O units. The issued I/O unit will
1299 always be a multiple of the minimum size, unless
1300 :option:`blocksize_unaligned` is set.
1302 Comma-separated ranges may be specified for reads, writes, and trims as
1303 described in :option:`blocksize`.
1305 Example: ``bsrange=1k-4k,2k-8k``.
1307 .. option:: bssplit=str[,str][,str]
1309 Sometimes you want even finer grained control of the block sizes issued, not
1310 just an even split between them. This option allows you to weight various
1311 block sizes, so that you are able to define a specific amount of block sizes
1312 issued. The format for this option is::
1314 bssplit=blocksize/percentage:blocksize/percentage
1316 for as many block sizes as needed. So if you want to define a workload that
1317 has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would write::
1319 bssplit=4k/10:64k/50:32k/40
1321 Ordering does not matter. If the percentage is left blank, fio will fill in
1322 the remaining values evenly. So a bssplit option like this one::
1324 bssplit=4k/50:1k/:32k/
1326 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always add up
1327 to 100, if bssplit is given a range that adds up to more, it will error out.
1329 Comma-separated values may be specified for reads, writes, and trims as
1330 described in :option:`blocksize`.
1332 If you want a workload that has 50% 2k reads and 50% 4k reads, while having
1333 90% 4k writes and 10% 8k writes, you would specify::
1335 bssplit=2k/50:4k/50,4k/90,8k/10
1337 .. option:: blocksize_unaligned, bs_unaligned
1339 If set, fio will issue I/O units with any size within
1340 :option:`blocksize_range`, not just multiples of the minimum size. This
1341 typically won't work with direct I/O, as that normally requires sector
1344 .. option:: bs_is_seq_rand
1346 If this option is set, fio will use the normal read,write blocksize settings
1347 as sequential,random blocksize settings instead. Any random read or write
1348 will use the WRITE blocksize settings, and any sequential read or write will
1349 use the READ blocksize settings.
1351 .. option:: blockalign=int[,int][,int], ba=int[,int][,int]
1353 Boundary to which fio will align random I/O units. Default:
1354 :option:`blocksize`. Minimum alignment is typically 512b for using direct
1355 I/O, though it usually depends on the hardware block size. This option is
1356 mutually exclusive with using a random map for files, so it will turn off
1357 that option. Comma-separated values may be specified for reads, writes, and
1358 trims as described in :option:`blocksize`.
1364 .. option:: zero_buffers
1366 Initialize buffers with all zeros. Default: fill buffers with random data.
1368 .. option:: refill_buffers
1370 If this option is given, fio will refill the I/O buffers on every
1371 submit. The default is to only fill it at init time and reuse that
1372 data. Only makes sense if zero_buffers isn't specified, naturally. If data
1373 verification is enabled, `refill_buffers` is also automatically enabled.
1375 .. option:: scramble_buffers=bool
1377 If :option:`refill_buffers` is too costly and the target is using data
1378 deduplication, then setting this option will slightly modify the I/O buffer
1379 contents to defeat normal de-dupe attempts. This is not enough to defeat
1380 more clever block compression attempts, but it will stop naive dedupe of
1381 blocks. Default: true.
1383 .. option:: buffer_compress_percentage=int
1385 If this is set, then fio will attempt to provide I/O buffer content (on
1386 WRITEs) that compress to the specified level. Fio does this by providing a
1387 mix of random data and a fixed pattern. The fixed pattern is either zeroes,
1388 or the pattern specified by :option:`buffer_pattern`. If the pattern option
1389 is used, it might skew the compression ratio slightly. Note that this is per
1390 block size unit, for file/disk wide compression level that matches this
1391 setting, you'll also want to set :option:`refill_buffers`.
1393 .. option:: buffer_compress_chunk=int
1395 See :option:`buffer_compress_percentage`. This setting allows fio to manage
1396 how big the ranges of random data and zeroed data is. Without this set, fio
1397 will provide :option:`buffer_compress_percentage` of blocksize random data,
1398 followed by the remaining zeroed. With this set to some chunk size smaller
1399 than the block size, fio can alternate random and zeroed data throughout the
1402 .. option:: buffer_pattern=str
1404 If set, fio will fill the I/O buffers with this pattern. If not set, the
1405 contents of I/O buffers is defined by the other options related to buffer
1406 contents. The setting can be any pattern of bytes, and can be prefixed with
1407 0x for hex values. It may also be a string, where the string must then be
1408 wrapped with ``""``, e.g.::
1410 buffer_pattern="abcd"
1418 buffer_pattern=0xdeadface
1420 Also you can combine everything together in any order::
1422 buffer_pattern=0xdeadface"abcd"-12
1424 .. option:: dedupe_percentage=int
1426 If set, fio will generate this percentage of identical buffers when
1427 writing. These buffers will be naturally dedupable. The contents of the
1428 buffers depend on what other buffer compression settings have been set. It's
1429 possible to have the individual buffers either fully compressible, or not at
1430 all. This option only controls the distribution of unique buffers.
1432 .. option:: invalidate=bool
1434 Invalidate the buffer/page cache parts for this file prior to starting
1435 I/O if the platform and file type support it. Defaults to true.
1436 This will be ignored if :option:`pre_read` is also specified for the
1439 .. option:: sync=bool
1441 Use synchronous I/O for buffered writes. For the majority of I/O engines,
1442 this means using O_SYNC. Default: false.
1444 .. option:: iomem=str, mem=str
1446 Fio can use various types of memory as the I/O unit buffer. The allowed
1450 Use memory from :manpage:`malloc(3)` as the buffers. Default memory
1454 Use shared memory as the buffers. Allocated through
1455 :manpage:`shmget(2)`.
1458 Same as shm, but use huge pages as backing.
1461 Use mmap to allocate buffers. May either be anonymous memory, or can
1462 be file backed if a filename is given after the option. The format
1463 is `mem=mmap:/path/to/file`.
1466 Use a memory mapped huge file as the buffer backing. Append filename
1467 after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file`.
1470 Same as mmap, but use a MMAP_SHARED mapping.
1473 Use GPU memory as the buffers for GPUDirect RDMA benchmark.
1475 The area allocated is a function of the maximum allowed bs size for the job,
1476 multiplied by the I/O depth given. Note that for **shmhuge** and
1477 **mmaphuge** to work, the system must have free huge pages allocated. This
1478 can normally be checked and set by reading/writing
1479 :file:`/proc/sys/vm/nr_hugepages` on a Linux system. Fio assumes a huge page
1480 is 4MiB in size. So to calculate the number of huge pages you need for a
1481 given job file, add up the I/O depth of all jobs (normally one unless
1482 :option:`iodepth` is used) and multiply by the maximum bs set. Then divide
1483 that number by the huge page size. You can see the size of the huge pages in
1484 :file:`/proc/meminfo`. If no huge pages are allocated by having a non-zero
1485 number in `nr_hugepages`, using **mmaphuge** or **shmhuge** will fail. Also
1486 see :option:`hugepage-size`.
1488 **mmaphuge** also needs to have hugetlbfs mounted and the file location
1489 should point there. So if it's mounted in :file:`/huge`, you would use
1490 `mem=mmaphuge:/huge/somefile`.
1492 .. option:: iomem_align=int
1494 This indicates the memory alignment of the I/O memory buffers. Note that
1495 the given alignment is applied to the first I/O unit buffer, if using
1496 :option:`iodepth` the alignment of the following buffers are given by the
1497 :option:`bs` used. In other words, if using a :option:`bs` that is a
1498 multiple of the page sized in the system, all buffers will be aligned to
1499 this value. If using a :option:`bs` that is not page aligned, the alignment
1500 of subsequent I/O memory buffers is the sum of the :option:`iomem_align` and
1503 .. option:: hugepage-size=int
1505 Defines the size of a huge page. Must at least be equal to the system
1506 setting, see :file:`/proc/meminfo`. Defaults to 4MiB. Should probably
1507 always be a multiple of megabytes, so using ``hugepage-size=Xm`` is the
1508 preferred way to set this to avoid setting a non-pow-2 bad value.
1510 .. option:: lockmem=int
1512 Pin the specified amount of memory with :manpage:`mlock(2)`. Can be used to
1513 simulate a smaller amount of memory. The amount specified is per worker.
1519 .. option:: size=int
1521 The total size of file I/O for each thread of this job. Fio will run until
1522 this many bytes has been transferred, unless runtime is limited by other options
1523 (such as :option:`runtime`, for instance, or increased/decreased by :option:`io_size`).
1524 Fio will divide this size between the available files determined by options
1525 such as :option:`nrfiles`, :option:`filename`, unless :option:`filesize` is
1526 specified by the job. If the result of division happens to be 0, the size is
1527 set to the physical size of the given files or devices if they exist.
1528 If this option is not specified, fio will use the full size of the given
1529 files or devices. If the files do not exist, size must be given. It is also
1530 possible to give size as a percentage between 1 and 100. If ``size=20%`` is
1531 given, fio will use 20% of the full size of the given files or devices.
1532 Can be combined with :option:`offset` to constrain the start and end range
1533 that I/O will be done within.
1535 .. option:: io_size=int, io_limit=int
1537 Normally fio operates within the region set by :option:`size`, which means
1538 that the :option:`size` option sets both the region and size of I/O to be
1539 performed. Sometimes that is not what you want. With this option, it is
1540 possible to define just the amount of I/O that fio should do. For instance,
1541 if :option:`size` is set to 20GiB and :option:`io_size` is set to 5GiB, fio
1542 will perform I/O within the first 20GiB but exit when 5GiB have been
1543 done. The opposite is also possible -- if :option:`size` is set to 20GiB,
1544 and :option:`io_size` is set to 40GiB, then fio will do 40GiB of I/O within
1545 the 0..20GiB region.
1547 .. option:: filesize=int
1549 Individual file sizes. May be a range, in which case fio will select sizes
1550 for files at random within the given range and limited to :option:`size` in
1551 total (if that is given). If not given, each created file is the same size.
1552 This option overrides :option:`size` in terms of file size, which means
1553 this value is used as a fixed size or possible range of each file.
1555 .. option:: file_append=bool
1557 Perform I/O after the end of the file. Normally fio will operate within the
1558 size of a file. If this option is set, then fio will append to the file
1559 instead. This has identical behavior to setting :option:`offset` to the size
1560 of a file. This option is ignored on non-regular files.
1562 .. option:: fill_device=bool, fill_fs=bool
1564 Sets size to something really large and waits for ENOSPC (no space left on
1565 device) as the terminating condition. Only makes sense with sequential
1566 write. For a read workload, the mount point will be filled first then I/O
1567 started on the result. This option doesn't make sense if operating on a raw
1568 device node, since the size of that is already known by the file system.
1569 Additionally, writing beyond end-of-device will not return ENOSPC there.
1575 .. option:: ioengine=str
1577 Defines how the job issues I/O to the file. The following types are defined:
1580 Basic :manpage:`read(2)` or :manpage:`write(2)`
1581 I/O. :manpage:`lseek(2)` is used to position the I/O location.
1582 See :option:`fsync` and :option:`fdatasync` for syncing write I/Os.
1585 Basic :manpage:`pread(2)` or :manpage:`pwrite(2)` I/O. Default on
1586 all supported operating systems except for Windows.
1589 Basic :manpage:`readv(2)` or :manpage:`writev(2)` I/O. Will emulate
1590 queuing by coalescing adjacent I/Os into a single submission.
1593 Basic :manpage:`preadv(2)` or :manpage:`pwritev(2)` I/O.
1596 Basic :manpage:`preadv2(2)` or :manpage:`pwritev2(2)` I/O.
1599 Linux native asynchronous I/O. Note that Linux may only support
1600 queued behaviour with non-buffered I/O (set ``direct=1`` or
1602 This engine defines engine specific options.
1605 POSIX asynchronous I/O using :manpage:`aio_read(3)` and
1606 :manpage:`aio_write(3)`.
1609 Solaris native asynchronous I/O.
1612 Windows native asynchronous I/O. Default on Windows.
1615 File is memory mapped with :manpage:`mmap(2)` and data copied
1616 to/from using :manpage:`memcpy(3)`.
1619 :manpage:`splice(2)` is used to transfer the data and
1620 :manpage:`vmsplice(2)` to transfer data from user space to the
1624 SCSI generic sg v3 I/O. May either be synchronous using the SG_IO
1625 ioctl, or if the target is an sg character device we use
1626 :manpage:`read(2)` and :manpage:`write(2)` for asynchronous
1627 I/O. Requires filename option to specify either block or character
1631 Doesn't transfer any data, just pretends to. This is mainly used to
1632 exercise fio itself and for debugging/testing purposes.
1635 Transfer over the network to given ``host:port``. Depending on the
1636 :option:`protocol` used, the :option:`hostname`, :option:`port`,
1637 :option:`listen` and :option:`filename` options are used to specify
1638 what sort of connection to make, while the :option:`protocol` option
1639 determines which protocol will be used. This engine defines engine
1643 Like **net**, but uses :manpage:`splice(2)` and
1644 :manpage:`vmsplice(2)` to map data and send/receive.
1645 This engine defines engine specific options.
1648 Doesn't transfer any data, but burns CPU cycles according to the
1649 :option:`cpuload` and :option:`cpuchunks` options. Setting
1650 :option:`cpuload` =85 will cause that job to do nothing but burn 85%
1651 of the CPU. In case of SMP machines, use :option:`numjobs`
1652 =<no_of_cpu> to get desired CPU usage, as the cpuload only loads a
1653 single CPU at the desired rate. A job never finishes unless there is
1654 at least one non-cpuio job.
1657 The GUASI I/O engine is the Generic Userspace Asyncronous Syscall
1658 Interface approach to async I/O. See
1660 http://www.xmailserver.org/guasi-lib.html
1662 for more info on GUASI.
1665 The RDMA I/O engine supports both RDMA memory semantics
1666 (RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the
1667 InfiniBand, RoCE and iWARP protocols.
1670 I/O engine that does regular fallocate to simulate data transfer as
1674 does fallocate(,mode = FALLOC_FL_KEEP_SIZE,).
1677 does fallocate(,mode = 0).
1680 does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE).
1683 I/O engine that sends :manpage:`ftruncate(2)` operations in response
1684 to write (DDIR_WRITE) events. Each ftruncate issued sets the file's
1685 size to the current block offset. Block size is ignored.
1688 I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate
1689 defragment activity in request to DDIR_WRITE event.
1692 I/O engine supporting direct access to Ceph Rados Block Devices
1693 (RBD) via librbd without the need to use the kernel rbd driver. This
1694 ioengine defines engine specific options.
1697 Using Glusterfs libgfapi sync interface to direct access to
1698 Glusterfs volumes without having to go through FUSE. This ioengine
1699 defines engine specific options.
1702 Using Glusterfs libgfapi async interface to direct access to
1703 Glusterfs volumes without having to go through FUSE. This ioengine
1704 defines engine specific options.
1707 Read and write through Hadoop (HDFS). The :file:`filename` option
1708 is used to specify host,port of the hdfs name-node to connect. This
1709 engine interprets offsets a little differently. In HDFS, files once
1710 created cannot be modified. So random writes are not possible. To
1711 imitate this, libhdfs engine expects bunch of small files to be
1712 created over HDFS, and engine will randomly pick a file out of those
1713 files based on the offset generated by fio backend. (see the example
1714 job file to create such files, use ``rw=write`` option). Please
1715 note, you might want to set necessary environment variables to work
1716 with hdfs/libhdfs properly. Each job uses its own connection to
1720 Read, write and erase an MTD character device (e.g.,
1721 :file:`/dev/mtd0`). Discards are treated as erases. Depending on the
1722 underlying device type, the I/O may have to go in a certain pattern,
1723 e.g., on NAND, writing sequentially to erase blocks and discarding
1724 before overwriting. The writetrim mode works well for this
1728 Read and write using filesystem DAX to a file on a filesystem
1729 mounted with DAX on a persistent memory device through the NVML
1733 Read and write using device DAX to a persistent memory device (e.g.,
1734 /dev/dax0.0) through the NVML libpmem library.
1737 Prefix to specify loading an external I/O engine object file. Append
1738 the engine filename, e.g. ``ioengine=external:/tmp/foo.o`` to load
1739 ioengine :file:`foo.o` in :file:`/tmp`.
1742 I/O engine specific parameters
1743 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1745 In addition, there are some parameters which are only valid when a specific
1746 ioengine is in use. These are used identically to normal parameters, with the
1747 caveat that when used on the command line, they must come after the
1748 :option:`ioengine` that defines them is selected.
1750 .. option:: userspace_reap : [libaio]
1752 Normally, with the libaio engine in use, fio will use the
1753 :manpage:`io_getevents(2)` system call to reap newly returned events. With
1754 this flag turned on, the AIO ring will be read directly from user-space to
1755 reap events. The reaping mode is only enabled when polling for a minimum of
1756 0 events (e.g. when :option:`iodepth_batch_complete` `=0`).
1758 .. option:: hipri : [pvsync2]
1760 Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority
1763 .. option:: cpuload=int : [cpuio]
1765 Attempt to use the specified percentage of CPU cycles. This is a mandatory
1766 option when using cpuio I/O engine.
1768 .. option:: cpuchunks=int : [cpuio]
1770 Split the load into cycles of the given time. In microseconds.
1772 .. option:: exit_on_io_done=bool : [cpuio]
1774 Detect when I/O threads are done, then exit.
1776 .. option:: hostname=str : [netsplice] [net]
1778 The host name or IP address to use for TCP or UDP based I/O. If the job is
1779 a TCP listener or UDP reader, the host name is not used and must be omitted
1780 unless it is a valid UDP multicast address.
1782 .. option:: namenode=str : [libhdfs]
1784 The host name or IP address of a HDFS cluster namenode to contact.
1786 .. option:: port=int
1790 The TCP or UDP port to bind to or connect to. If this is used with
1791 :option:`numjobs` to spawn multiple instances of the same job type, then
1792 this will be the starting port number since fio will use a range of
1797 the listening port of the HFDS cluster namenode.
1799 .. option:: interface=str : [netsplice] [net]
1801 The IP address of the network interface used to send or receive UDP
1804 .. option:: ttl=int : [netsplice] [net]
1806 Time-to-live value for outgoing UDP multicast packets. Default: 1.
1808 .. option:: nodelay=bool : [netsplice] [net]
1810 Set TCP_NODELAY on TCP connections.
1812 .. option:: protocol=str : [netsplice] [net]
1814 .. option:: proto=str : [netsplice] [net]
1816 The network protocol to use. Accepted values are:
1819 Transmission control protocol.
1821 Transmission control protocol V6.
1823 User datagram protocol.
1825 User datagram protocol V6.
1829 When the protocol is TCP or UDP, the port must also be given, as well as the
1830 hostname if the job is a TCP listener or UDP reader. For unix sockets, the
1831 normal filename option should be used and the port is invalid.
1833 .. option:: listen : [net]
1835 For TCP network connections, tell fio to listen for incoming connections
1836 rather than initiating an outgoing connection. The :option:`hostname` must
1837 be omitted if this option is used.
1839 .. option:: pingpong : [net]
1841 Normally a network writer will just continue writing data, and a network
1842 reader will just consume packages. If ``pingpong=1`` is set, a writer will
1843 send its normal payload to the reader, then wait for the reader to send the
1844 same payload back. This allows fio to measure network latencies. The
1845 submission and completion latencies then measure local time spent sending or
1846 receiving, and the completion latency measures how long it took for the
1847 other end to receive and send back. For UDP multicast traffic
1848 ``pingpong=1`` should only be set for a single reader when multiple readers
1849 are listening to the same address.
1851 .. option:: window_size : [net]
1853 Set the desired socket buffer size for the connection.
1855 .. option:: mss : [net]
1857 Set the TCP maximum segment size (TCP_MAXSEG).
1859 .. option:: donorname=str : [e4defrag]
1861 File will be used as a block donor(swap extents between files).
1863 .. option:: inplace=int : [e4defrag]
1865 Configure donor file blocks allocation strategy:
1868 Default. Preallocate donor's file on init.
1870 Allocate space immediately inside defragment event, and free right
1873 .. option:: clustername=str : [rbd]
1875 Specifies the name of the Ceph cluster.
1877 .. option:: rbdname=str : [rbd]
1879 Specifies the name of the RBD.
1881 .. option:: pool=str : [rbd]
1883 Specifies the name of the Ceph pool containing RBD.
1885 .. option:: clientname=str : [rbd]
1887 Specifies the username (without the 'client.' prefix) used to access the
1888 Ceph cluster. If the *clustername* is specified, the *clientname* shall be
1889 the full *type.id* string. If no type. prefix is given, fio will add
1890 'client.' by default.
1892 .. option:: skip_bad=bool : [mtd]
1894 Skip operations against known bad blocks.
1896 .. option:: hdfsdirectory : [libhdfs]
1898 libhdfs will create chunk in this HDFS directory.
1900 .. option:: chunk_size : [libhdfs]
1902 the size of the chunk to use for each file.
1908 .. option:: iodepth=int
1910 Number of I/O units to keep in flight against the file. Note that
1911 increasing *iodepth* beyond 1 will not affect synchronous ioengines (except
1912 for small degrees when :option:`verify_async` is in use). Even async
1913 engines may impose OS restrictions causing the desired depth not to be
1914 achieved. This may happen on Linux when using libaio and not setting
1915 :option:`direct` =1, since buffered I/O is not async on that OS. Keep an
1916 eye on the I/O depth distribution in the fio output to verify that the
1917 achieved depth is as expected. Default: 1.
1919 .. option:: iodepth_batch_submit=int, iodepth_batch=int
1921 This defines how many pieces of I/O to submit at once. It defaults to 1
1922 which means that we submit each I/O as soon as it is available, but can be
1923 raised to submit bigger batches of I/O at the time. If it is set to 0 the
1924 :option:`iodepth` value will be used.
1926 .. option:: iodepth_batch_complete_min=int, iodepth_batch_complete=int
1928 This defines how many pieces of I/O to retrieve at once. It defaults to 1
1929 which means that we'll ask for a minimum of 1 I/O in the retrieval process
1930 from the kernel. The I/O retrieval will go on until we hit the limit set by
1931 :option:`iodepth_low`. If this variable is set to 0, then fio will always
1932 check for completed events before queuing more I/O. This helps reduce I/O
1933 latency, at the cost of more retrieval system calls.
1935 .. option:: iodepth_batch_complete_max=int
1937 This defines maximum pieces of I/O to retrieve at once. This variable should
1938 be used along with :option:`iodepth_batch_complete_min` =int variable,
1939 specifying the range of min and max amount of I/O which should be
1940 retrieved. By default it is equal to :option:`iodepth_batch_complete_min`
1945 iodepth_batch_complete_min=1
1946 iodepth_batch_complete_max=<iodepth>
1948 which means that we will retrieve at least 1 I/O and up to the whole
1949 submitted queue depth. If none of I/O has been completed yet, we will wait.
1953 iodepth_batch_complete_min=0
1954 iodepth_batch_complete_max=<iodepth>
1956 which means that we can retrieve up to the whole submitted queue depth, but
1957 if none of I/O has been completed yet, we will NOT wait and immediately exit
1958 the system call. In this example we simply do polling.
1960 .. option:: iodepth_low=int
1962 The low water mark indicating when to start filling the queue
1963 again. Defaults to the same as :option:`iodepth`, meaning that fio will
1964 attempt to keep the queue full at all times. If :option:`iodepth` is set to
1965 e.g. 16 and *iodepth_low* is set to 4, then after fio has filled the queue of
1966 16 requests, it will let the depth drain down to 4 before starting to fill
1969 .. option:: io_submit_mode=str
1971 This option controls how fio submits the I/O to the I/O engine. The default
1972 is `inline`, which means that the fio job threads submit and reap I/O
1973 directly. If set to `offload`, the job threads will offload I/O submission
1974 to a dedicated pool of I/O threads. This requires some coordination and thus
1975 has a bit of extra overhead, especially for lower queue depth I/O where it
1976 can increase latencies. The benefit is that fio can manage submission rates
1977 independently of the device completion rates. This avoids skewed latency
1978 reporting if I/O gets back up on the device side (the coordinated omission
1985 .. option:: thinktime=time
1987 Stall the job for the specified period of time after an I/O has completed before issuing the
1988 next. May be used to simulate processing being done by an application.
1989 When the unit is omitted, the value is given in microseconds. See
1990 :option:`thinktime_blocks` and :option:`thinktime_spin`.
1992 .. option:: thinktime_spin=time
1994 Only valid if :option:`thinktime` is set - pretend to spend CPU time doing
1995 something with the data received, before falling back to sleeping for the
1996 rest of the period specified by :option:`thinktime`. When the unit is
1997 omitted, the value is given in microseconds.
1999 .. option:: thinktime_blocks=int
2001 Only valid if :option:`thinktime` is set - control how many blocks to issue,
2002 before waiting `thinktime` usecs. If not set, defaults to 1 which will make
2003 fio wait `thinktime` usecs after every block. This effectively makes any
2004 queue depth setting redundant, since no more than 1 I/O will be queued
2005 before we have to complete it and do our thinktime. In other words, this
2006 setting effectively caps the queue depth if the latter is larger.
2008 .. option:: rate=int[,int][,int]
2010 Cap the bandwidth used by this job. The number is in bytes/sec, the normal
2011 suffix rules apply. Comma-separated values may be specified for reads,
2012 writes, and trims as described in :option:`blocksize`.
2014 .. option:: rate_min=int[,int][,int]
2016 Tell fio to do whatever it can to maintain at least this bandwidth. Failing
2017 to meet this requirement will cause the job to exit. Comma-separated values
2018 may be specified for reads, writes, and trims as described in
2019 :option:`blocksize`.
2021 .. option:: rate_iops=int[,int][,int]
2023 Cap the bandwidth to this number of IOPS. Basically the same as
2024 :option:`rate`, just specified independently of bandwidth. If the job is
2025 given a block size range instead of a fixed value, the smallest block size
2026 is used as the metric. Comma-separated values may be specified for reads,
2027 writes, and trims as described in :option:`blocksize`.
2029 .. option:: rate_iops_min=int[,int][,int]
2031 If fio doesn't meet this rate of I/O, it will cause the job to exit.
2032 Comma-separated values may be specified for reads, writes, and trims as
2033 described in :option:`blocksize`.
2035 .. option:: rate_process=str
2037 This option controls how fio manages rated I/O submissions. The default is
2038 `linear`, which submits I/O in a linear fashion with fixed delays between
2039 I/Os that gets adjusted based on I/O completion rates. If this is set to
2040 `poisson`, fio will submit I/O based on a more real world random request
2041 flow, known as the Poisson process
2042 (https://en.wikipedia.org/wiki/Poisson_point_process). The lambda will be
2043 10^6 / IOPS for the given workload.
2049 .. option:: latency_target=time
2051 If set, fio will attempt to find the max performance point that the given
2052 workload will run at while maintaining a latency below this target. When
2053 the unit is omitted, the value is given in microseconds. See
2054 :option:`latency_window` and :option:`latency_percentile`.
2056 .. option:: latency_window=time
2058 Used with :option:`latency_target` to specify the sample window that the job
2059 is run at varying queue depths to test the performance. When the unit is
2060 omitted, the value is given in microseconds.
2062 .. option:: latency_percentile=float
2064 The percentage of I/Os that must fall within the criteria specified by
2065 :option:`latency_target` and :option:`latency_window`. If not set, this
2066 defaults to 100.0, meaning that all I/Os must be equal or below to the value
2067 set by :option:`latency_target`.
2069 .. option:: max_latency=time
2071 If set, fio will exit the job with an ETIMEDOUT error if it exceeds this
2072 maximum latency. When the unit is omitted, the value is given in
2075 .. option:: rate_cycle=int
2077 Average bandwidth for :option:`rate` and :option:`rate_min` over this number
2084 .. option:: write_iolog=str
2086 Write the issued I/O patterns to the specified file. See
2087 :option:`read_iolog`. Specify a separate file for each job, otherwise the
2088 iologs will be interspersed and the file may be corrupt.
2090 .. option:: read_iolog=str
2092 Open an iolog with the specified file name and replay the I/O patterns it
2093 contains. This can be used to store a workload and replay it sometime
2094 later. The iolog given may also be a blktrace binary file, which allows fio
2095 to replay a workload captured by :command:`blktrace`. See
2096 :manpage:`blktrace(8)` for how to capture such logging data. For blktrace
2097 replay, the file needs to be turned into a blkparse binary data file first
2098 (``blkparse <device> -o /dev/null -d file_for_fio.bin``).
2100 .. option:: replay_no_stall=int
2102 When replaying I/O with :option:`read_iolog` the default behavior is to
2103 attempt to respect the time stamps within the log and replay them with the
2104 appropriate delay between IOPS. By setting this variable fio will not
2105 respect the timestamps and attempt to replay them as fast as possible while
2106 still respecting ordering. The result is the same I/O pattern to a given
2107 device, but different timings.
2109 .. option:: replay_redirect=str
2111 While replaying I/O patterns using :option:`read_iolog` the default behavior
2112 is to replay the IOPS onto the major/minor device that each IOP was recorded
2113 from. This is sometimes undesirable because on a different machine those
2114 major/minor numbers can map to a different device. Changing hardware on the
2115 same system can also result in a different major/minor mapping.
2116 ``replay_redirect`` causes all IOPS to be replayed onto the single specified
2117 device regardless of the device it was recorded
2118 from. i.e. :option:`replay_redirect` = :file:`/dev/sdc` would cause all I/O
2119 in the blktrace or iolog to be replayed onto :file:`/dev/sdc`. This means
2120 multiple devices will be replayed onto a single device, if the trace
2121 contains multiple devices. If you want multiple devices to be replayed
2122 concurrently to multiple redirected devices you must blkparse your trace
2123 into separate traces and replay them with independent fio invocations.
2124 Unfortunately this also breaks the strict time ordering between multiple
2127 .. option:: replay_align=int
2129 Force alignment of I/O offsets and lengths in a trace to this power of 2
2132 .. option:: replay_scale=int
2134 Scale sector offsets down by this factor when replaying traces.
2137 Threads, processes and job synchronization
2138 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2142 Fio defaults to forking jobs, however if this option is given, fio will use
2143 POSIX Threads function :manpage:`pthread_create(3)` to create threads instead
2144 of forking processes.
2146 .. option:: wait_for=str
2148 Specifies the name of the already defined job to wait for. Single waitee
2149 name only may be specified. If set, the job won't be started until all
2150 workers of the waitee job are done.
2152 ``wait_for`` operates on the job name basis, so there are a few
2153 limitations. First, the waitee must be defined prior to the waiter job
2154 (meaning no forward references). Second, if a job is being referenced as a
2155 waitee, it must have a unique name (no duplicate waitees).
2157 .. option:: nice=int
2159 Run the job with the given nice value. See man :manpage:`nice(2)`.
2161 On Windows, values less than -15 set the process class to "High"; -1 through
2162 -15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle"
2165 .. option:: prio=int
2167 Set the I/O priority value of this job. Linux limits us to a positive value
2168 between 0 and 7, with 0 being the highest. See man
2169 :manpage:`ionice(1)`. Refer to an appropriate manpage for other operating
2170 systems since meaning of priority may differ.
2172 .. option:: prioclass=int
2174 Set the I/O priority class. See man :manpage:`ionice(1)`.
2176 .. option:: cpumask=int
2178 Set the CPU affinity of this job. The parameter given is a bitmask of
2179 allowed CPU's the job may run on. So if you want the allowed CPUs to be 1
2180 and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man
2181 :manpage:`sched_setaffinity(2)`. This may not work on all supported
2182 operating systems or kernel versions. This option doesn't work well for a
2183 higher CPU count than what you can store in an integer mask, so it can only
2184 control cpus 1-32. For boxes with larger CPU counts, use
2185 :option:`cpus_allowed`.
2187 .. option:: cpus_allowed=str
2189 Controls the same options as :option:`cpumask`, but it allows a text setting
2190 of the permitted CPUs instead. So to use CPUs 1 and 5, you would specify
2191 ``cpus_allowed=1,5``. This options also allows a range of CPUs. Say you
2192 wanted a binding to CPUs 1, 5, and 8-15, you would set
2193 ``cpus_allowed=1,5,8-15``.
2195 .. option:: cpus_allowed_policy=str
2197 Set the policy of how fio distributes the CPUs specified by
2198 :option:`cpus_allowed` or cpumask. Two policies are supported:
2201 All jobs will share the CPU set specified.
2203 Each job will get a unique CPU from the CPU set.
2205 **shared** is the default behaviour, if the option isn't specified. If
2206 **split** is specified, then fio will will assign one cpu per job. If not
2207 enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs
2210 .. option:: numa_cpu_nodes=str
2212 Set this job running on specified NUMA nodes' CPUs. The arguments allow
2213 comma delimited list of cpu numbers, A-B ranges, or `all`. Note, to enable
2214 numa options support, fio must be built on a system with libnuma-dev(el)
2217 .. option:: numa_mem_policy=str
2219 Set this job's memory policy and corresponding NUMA nodes. Format of the
2224 ``mode`` is one of the following memory policy: ``default``, ``prefer``,
2225 ``bind``, ``interleave``, ``local`` For ``default`` and ``local`` memory
2226 policy, no node is needed to be specified. For ``prefer``, only one node is
2227 allowed. For ``bind`` and ``interleave``, it allow comma delimited list of
2228 numbers, A-B ranges, or `all`.
2230 .. option:: cgroup=str
2232 Add job to this control group. If it doesn't exist, it will be created. The
2233 system must have a mounted cgroup blkio mount point for this to work. If
2234 your system doesn't have it mounted, you can do so with::
2236 # mount -t cgroup -o blkio none /cgroup
2238 .. option:: cgroup_weight=int
2240 Set the weight of the cgroup to this value. See the documentation that comes
2241 with the kernel, allowed values are in the range of 100..1000.
2243 .. option:: cgroup_nodelete=bool
2245 Normally fio will delete the cgroups it has created after the job
2246 completion. To override this behavior and to leave cgroups around after the
2247 job completion, set ``cgroup_nodelete=1``. This can be useful if one wants
2248 to inspect various cgroup files after job completion. Default: false.
2250 .. option:: flow_id=int
2252 The ID of the flow. If not specified, it defaults to being a global
2253 flow. See :option:`flow`.
2255 .. option:: flow=int
2257 Weight in token-based flow control. If this value is used, then there is a
2258 'flow counter' which is used to regulate the proportion of activity between
2259 two or more jobs. Fio attempts to keep this flow counter near zero. The
2260 ``flow`` parameter stands for how much should be added or subtracted to the
2261 flow counter on each iteration of the main I/O loop. That is, if one job has
2262 ``flow=8`` and another job has ``flow=-1``, then there will be a roughly 1:8
2263 ratio in how much one runs vs the other.
2265 .. option:: flow_watermark=int
2267 The maximum value that the absolute value of the flow counter is allowed to
2268 reach before the job must wait for a lower value of the counter.
2270 .. option:: flow_sleep=int
2272 The period of time, in microseconds, to wait after the flow watermark has
2273 been exceeded before retrying operations.
2275 .. option:: stonewall, wait_for_previous
2277 Wait for preceding jobs in the job file to exit, before starting this
2278 one. Can be used to insert serialization points in the job file. A stone
2279 wall also implies starting a new reporting group, see
2280 :option:`group_reporting`.
2284 When one job finishes, terminate the rest. The default is to wait for each
2285 job to finish, sometimes that is not the desired action.
2287 .. option:: exec_prerun=str
2289 Before running this job, issue the command specified through
2290 :manpage:`system(3)`. Output is redirected in a file called
2291 :file:`jobname.prerun.txt`.
2293 .. option:: exec_postrun=str
2295 After the job completes, issue the command specified though
2296 :manpage:`system(3)`. Output is redirected in a file called
2297 :file:`jobname.postrun.txt`.
2301 Instead of running as the invoking user, set the user ID to this value
2302 before the thread/process does any work.
2306 Set group ID, see :option:`uid`.
2312 .. option:: verify_only
2314 Do not perform specified workload, only verify data still matches previous
2315 invocation of this workload. This option allows one to check data multiple
2316 times at a later date without overwriting it. This option makes sense only
2317 for workloads that write data, and does not support workloads with the
2318 :option:`time_based` option set.
2320 .. option:: do_verify=bool
2322 Run the verify phase after a write phase. Only valid if :option:`verify` is
2325 .. option:: verify=str
2327 If writing to a file, fio can verify the file contents after each iteration
2328 of the job. Each verification method also implies verification of special
2329 header, which is written to the beginning of each block. This header also
2330 includes meta information, like offset of the block, block number, timestamp
2331 when block was written, etc. :option:`verify` can be combined with
2332 :option:`verify_pattern` option. The allowed values are:
2335 Use an md5 sum of the data area and store it in the header of
2339 Use an experimental crc64 sum of the data area and store it in the
2340 header of each block.
2343 Use a crc32c sum of the data area and store it in the header of each
2347 Use hardware assisted crc32c calculation provided on SSE4.2 enabled
2348 processors. Falls back to regular software crc32c, if not supported
2352 Use a crc32 sum of the data area and store it in the header of each
2356 Use a crc16 sum of the data area and store it in the header of each
2360 Use a crc7 sum of the data area and store it in the header of each
2364 Use xxhash as the checksum function. Generally the fastest software
2365 checksum that fio supports.
2368 Use sha512 as the checksum function.
2371 Use sha256 as the checksum function.
2374 Use optimized sha1 as the checksum function.
2377 Use optimized sha3-224 as the checksum function.
2380 Use optimized sha3-256 as the checksum function.
2383 Use optimized sha3-384 as the checksum function.
2386 Use optimized sha3-512 as the checksum function.
2389 This option is deprecated, since now meta information is included in
2390 generic verification header and meta verification happens by
2391 default. For detailed information see the description of the
2392 :option:`verify` setting. This option is kept because of
2393 compatibility's sake with old configurations. Do not use it.
2396 Verify a strict pattern. Normally fio includes a header with some
2397 basic information and checksumming, but if this option is set, only
2398 the specific pattern set with :option:`verify_pattern` is verified.
2401 Only pretend to verify. Useful for testing internals with
2402 :option:`ioengine` `=null`, not for much else.
2404 This option can be used for repeated burn-in tests of a system to make sure
2405 that the written data is also correctly read back. If the data direction
2406 given is a read or random read, fio will assume that it should verify a
2407 previously written file. If the data direction includes any form of write,
2408 the verify will be of the newly written data.
2410 .. option:: verifysort=bool
2412 If true, fio will sort written verify blocks when it deems it faster to read
2413 them back in a sorted manner. This is often the case when overwriting an
2414 existing file, since the blocks are already laid out in the file system. You
2415 can ignore this option unless doing huge amounts of really fast I/O where
2416 the red-black tree sorting CPU time becomes significant. Default: true.
2418 .. option:: verifysort_nr=int
2420 Pre-load and sort verify blocks for a read workload.
2422 .. option:: verify_offset=int
2424 Swap the verification header with data somewhere else in the block before
2425 writing. It is swapped back before verifying.
2427 .. option:: verify_interval=int
2429 Write the verification header at a finer granularity than the
2430 :option:`blocksize`. It will be written for chunks the size of
2431 ``verify_interval``. :option:`blocksize` should divide this evenly.
2433 .. option:: verify_pattern=str
2435 If set, fio will fill the I/O buffers with this pattern. Fio defaults to
2436 filling with totally random bytes, but sometimes it's interesting to fill
2437 with a known pattern for I/O verification purposes. Depending on the width
2438 of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time(it can
2439 be either a decimal or a hex number). The ``verify_pattern`` if larger than
2440 a 32-bit quantity has to be a hex number that starts with either "0x" or
2441 "0X". Use with :option:`verify`. Also, ``verify_pattern`` supports %o
2442 format, which means that for each block offset will be written and then
2443 verified back, e.g.::
2447 Or use combination of everything::
2449 verify_pattern=0xff%o"abcd"-12
2451 .. option:: verify_fatal=bool
2453 Normally fio will keep checking the entire contents before quitting on a
2454 block verification failure. If this option is set, fio will exit the job on
2455 the first observed failure. Default: false.
2457 .. option:: verify_dump=bool
2459 If set, dump the contents of both the original data block and the data block
2460 we read off disk to files. This allows later analysis to inspect just what
2461 kind of data corruption occurred. Off by default.
2463 .. option:: verify_async=int
2465 Fio will normally verify I/O inline from the submitting thread. This option
2466 takes an integer describing how many async offload threads to create for I/O
2467 verification instead, causing fio to offload the duty of verifying I/O
2468 contents to one or more separate threads. If using this offload option, even
2469 sync I/O engines can benefit from using an :option:`iodepth` setting higher
2470 than 1, as it allows them to have I/O in flight while verifies are running.
2471 Defaults to 0 async threads, i.e. verification is not asynchronous.
2473 .. option:: verify_async_cpus=str
2475 Tell fio to set the given CPU affinity on the async I/O verification
2476 threads. See :option:`cpus_allowed` for the format used.
2478 .. option:: verify_backlog=int
2480 Fio will normally verify the written contents of a job that utilizes verify
2481 once that job has completed. In other words, everything is written then
2482 everything is read back and verified. You may want to verify continually
2483 instead for a variety of reasons. Fio stores the meta data associated with
2484 an I/O block in memory, so for large verify workloads, quite a bit of memory
2485 would be used up holding this meta data. If this option is enabled, fio will
2486 write only N blocks before verifying these blocks.
2488 .. option:: verify_backlog_batch=int
2490 Control how many blocks fio will verify if :option:`verify_backlog` is
2491 set. If not set, will default to the value of :option:`verify_backlog`
2492 (meaning the entire queue is read back and verified). If
2493 ``verify_backlog_batch`` is less than :option:`verify_backlog` then not all
2494 blocks will be verified, if ``verify_backlog_batch`` is larger than
2495 :option:`verify_backlog`, some blocks will be verified more than once.
2497 .. option:: verify_state_save=bool
2499 When a job exits during the write phase of a verify workload, save its
2500 current state. This allows fio to replay up until that point, if the verify
2501 state is loaded for the verify read phase. The format of the filename is,
2504 <type>-<jobname>-<jobindex>-verify.state.
2506 <type> is "local" for a local run, "sock" for a client/server socket
2507 connection, and "ip" (192.168.0.1, for instance) for a networked
2508 client/server connection. Defaults to true.
2510 .. option:: verify_state_load=bool
2512 If a verify termination trigger was used, fio stores the current write state
2513 of each thread. This can be used at verification time so that fio knows how
2514 far it should verify. Without this information, fio will run a full
2515 verification pass, according to the settings in the job file used.
2517 .. option:: trim_percentage=int
2519 Number of verify blocks to discard/trim.
2521 .. option:: trim_verify_zero=bool
2523 Verify that trim/discarded blocks are returned as zeroes.
2525 .. option:: trim_backlog=int
2527 Verify that trim/discarded blocks are returned as zeroes.
2529 .. option:: trim_backlog_batch=int
2531 Trim this number of I/O blocks.
2533 .. option:: experimental_verify=bool
2535 Enable experimental verification.
2541 .. option:: steadystate=str:float, ss=str:float
2543 Define the criterion and limit for assessing steady state performance. The
2544 first parameter designates the criterion whereas the second parameter sets
2545 the threshold. When the criterion falls below the threshold for the
2546 specified duration, the job will stop. For example, `iops_slope:0.1%` will
2547 direct fio to terminate the job when the least squares regression slope
2548 falls below 0.1% of the mean IOPS. If :option:`group_reporting` is enabled
2549 this will apply to all jobs in the group. Below is the list of available
2550 steady state assessment criteria. All assessments are carried out using only
2551 data from the rolling collection window. Threshold limits can be expressed
2552 as a fixed value or as a percentage of the mean in the collection window.
2555 Collect IOPS data. Stop the job if all individual IOPS measurements
2556 are within the specified limit of the mean IOPS (e.g., ``iops:2``
2557 means that all individual IOPS values must be within 2 of the mean,
2558 whereas ``iops:0.2%`` means that all individual IOPS values must be
2559 within 0.2% of the mean IOPS to terminate the job).
2562 Collect IOPS data and calculate the least squares regression
2563 slope. Stop the job if the slope falls below the specified limit.
2566 Collect bandwidth data. Stop the job if all individual bandwidth
2567 measurements are within the specified limit of the mean bandwidth.
2570 Collect bandwidth data and calculate the least squares regression
2571 slope. Stop the job if the slope falls below the specified limit.
2573 .. option:: steadystate_duration=time, ss_dur=time
2575 A rolling window of this duration will be used to judge whether steady state
2576 has been reached. Data will be collected once per second. The default is 0
2577 which disables steady state detection. When the unit is omitted, the
2578 value is given in seconds.
2580 .. option:: steadystate_ramp_time=time, ss_ramp=time
2582 Allow the job to run for the specified duration before beginning data
2583 collection for checking the steady state job termination criterion. The
2584 default is 0. When the unit is omitted, the value is given in seconds.
2587 Measurements and reporting
2588 ~~~~~~~~~~~~~~~~~~~~~~~~~~
2590 .. option:: per_job_logs=bool
2592 If set, this generates bw/clat/iops log with per file private filenames. If
2593 not set, jobs with identical names will share the log filename. Default:
2596 .. option:: group_reporting
2598 It may sometimes be interesting to display statistics for groups of jobs as
2599 a whole instead of for each individual job. This is especially true if
2600 :option:`numjobs` is used; looking at individual thread/process output
2601 quickly becomes unwieldy. To see the final report per-group instead of
2602 per-job, use :option:`group_reporting`. Jobs in a file will be part of the
2603 same reporting group, unless if separated by a :option:`stonewall`, or by
2604 using :option:`new_group`.
2606 .. option:: new_group
2608 Start a new reporting group. See: :option:`group_reporting`. If not given,
2609 all jobs in a file will be part of the same reporting group, unless
2610 separated by a :option:`stonewall`.
2614 By default, fio collects and shows final output results for all jobs
2615 that run. If this option is set to 0, then fio will ignore it in
2616 the final stat output.
2618 .. option:: write_bw_log=str
2620 If given, write a bandwidth log for this job. Can be used to store data of
2621 the bandwidth of the jobs in their lifetime. The included
2622 :command:`fio_generate_plots` script uses :command:`gnuplot` to turn these
2623 text files into nice graphs. See :option:`write_lat_log` for behaviour of
2624 given filename. For this option, the postfix is :file:`_bw.x.log`, where `x`
2625 is the index of the job (`1..N`, where `N` is the number of jobs). If
2626 :option:`per_job_logs` is false, then the filename will not include the job
2627 index. See `Log File Formats`_.
2629 .. option:: write_lat_log=str
2631 Same as :option:`write_bw_log`, except that this option stores I/O
2632 submission, completion, and total latencies instead. If no filename is given
2633 with this option, the default filename of :file:`jobname_type.log` is
2634 used. Even if the filename is given, fio will still append the type of
2635 log. So if one specifies::
2639 The actual log names will be :file:`foo_slat.x.log`, :file:`foo_clat.x.log`,
2640 and :file:`foo_lat.x.log`, where `x` is the index of the job (1..N, where N
2641 is the number of jobs). This helps :command:`fio_generate_plot` find the
2642 logs automatically. If :option:`per_job_logs` is false, then the filename
2643 will not include the job index. See `Log File Formats`_.
2645 .. option:: write_hist_log=str
2647 Same as :option:`write_lat_log`, but writes I/O completion latency
2648 histograms. If no filename is given with this option, the default filename
2649 of :file:`jobname_clat_hist.x.log` is used, where `x` is the index of the
2650 job (1..N, where `N` is the number of jobs). Even if the filename is given,
2651 fio will still append the type of log. If :option:`per_job_logs` is false,
2652 then the filename will not include the job index. See `Log File Formats`_.
2654 .. option:: write_iops_log=str
2656 Same as :option:`write_bw_log`, but writes IOPS. If no filename is given
2657 with this option, the default filename of :file:`jobname_type.x.log` is
2658 used,where `x` is the index of the job (1..N, where `N` is the number of
2659 jobs). Even if the filename is given, fio will still append the type of
2660 log. If :option:`per_job_logs` is false, then the filename will not include
2661 the job index. See `Log File Formats`_.
2663 .. option:: log_avg_msec=int
2665 By default, fio will log an entry in the iops, latency, or bw log for every
2666 I/O that completes. When writing to the disk log, that can quickly grow to a
2667 very large size. Setting this option makes fio average the each log entry
2668 over the specified period of time, reducing the resolution of the log. See
2669 :option:`log_max_value` as well. Defaults to 0, logging all entries.
2671 .. option:: log_hist_msec=int
2673 Same as :option:`log_avg_msec`, but logs entries for completion latency
2674 histograms. Computing latency percentiles from averages of intervals using
2675 :option:`log_avg_msec` is inaccurate. Setting this option makes fio log
2676 histogram entries over the specified period of time, reducing log sizes for
2677 high IOPS devices while retaining percentile accuracy. See
2678 :option:`log_hist_coarseness` as well. Defaults to 0, meaning histogram
2679 logging is disabled.
2681 .. option:: log_hist_coarseness=int
2683 Integer ranging from 0 to 6, defining the coarseness of the resolution of
2684 the histogram logs enabled with :option:`log_hist_msec`. For each increment
2685 in coarseness, fio outputs half as many bins. Defaults to 0, for which
2686 histogram logs contain 1216 latency bins. See `Log File Formats`_.
2688 .. option:: log_max_value=bool
2690 If :option:`log_avg_msec` is set, fio logs the average over that window. If
2691 you instead want to log the maximum value, set this option to 1. Defaults to
2692 0, meaning that averaged values are logged.
2694 .. option:: log_offset=int
2696 If this is set, the iolog options will include the byte offset for the I/O
2697 entry as well as the other data values.
2699 .. option:: log_compression=int
2701 If this is set, fio will compress the I/O logs as it goes, to keep the
2702 memory footprint lower. When a log reaches the specified size, that chunk is
2703 removed and compressed in the background. Given that I/O logs are fairly
2704 highly compressible, this yields a nice memory savings for longer runs. The
2705 downside is that the compression will consume some background CPU cycles, so
2706 it may impact the run. This, however, is also true if the logging ends up
2707 consuming most of the system memory. So pick your poison. The I/O logs are
2708 saved normally at the end of a run, by decompressing the chunks and storing
2709 them in the specified log file. This feature depends on the availability of
2712 .. option:: log_compression_cpus=str
2714 Define the set of CPUs that are allowed to handle online log compression for
2715 the I/O jobs. This can provide better isolation between performance
2716 sensitive jobs, and background compression work.
2718 .. option:: log_store_compressed=bool
2720 If set, fio will store the log files in a compressed format. They can be
2721 decompressed with fio, using the :option:`--inflate-log` command line
2722 parameter. The files will be stored with a :file:`.fz` suffix.
2724 .. option:: log_unix_epoch=bool
2726 If set, fio will log Unix timestamps to the log files produced by enabling
2727 write_type_log for each log type, instead of the default zero-based
2730 .. option:: block_error_percentiles=bool
2732 If set, record errors in trim block-sized units from writes and trims and
2733 output a histogram of how many trims it took to get to errors, and what kind
2734 of error was encountered.
2736 .. option:: bwavgtime=int
2738 Average the calculated bandwidth over the given time. Value is specified in
2739 milliseconds. If the job also does bandwidth logging through
2740 :option:`write_bw_log`, then the minimum of this option and
2741 :option:`log_avg_msec` will be used. Default: 500ms.
2743 .. option:: iopsavgtime=int
2745 Average the calculated IOPS over the given time. Value is specified in
2746 milliseconds. If the job also does IOPS logging through
2747 :option:`write_iops_log`, then the minimum of this option and
2748 :option:`log_avg_msec` will be used. Default: 500ms.
2750 .. option:: disk_util=bool
2752 Generate disk utilization statistics, if the platform supports it.
2755 .. option:: disable_lat=bool
2757 Disable measurements of total latency numbers. Useful only for cutting back
2758 the number of calls to :manpage:`gettimeofday(2)`, as that does impact
2759 performance at really high IOPS rates. Note that to really get rid of a
2760 large amount of these calls, this option must be used with
2761 :option:`disable_slat` and :option:`disable_bw_measurement` as well.
2763 .. option:: disable_clat=bool
2765 Disable measurements of completion latency numbers. See
2766 :option:`disable_lat`.
2768 .. option:: disable_slat=bool
2770 Disable measurements of submission latency numbers. See
2771 :option:`disable_slat`.
2773 .. option:: disable_bw_measurement=bool, disable_bw=bool
2775 Disable measurements of throughput/bandwidth numbers. See
2776 :option:`disable_lat`.
2778 .. option:: clat_percentiles=bool
2780 Enable the reporting of percentiles of completion latencies.
2782 .. option:: percentile_list=float_list
2784 Overwrite the default list of percentiles for completion latencies and the
2785 block error histogram. Each number is a floating number in the range
2786 (0,100], and the maximum length of the list is 20. Use ``:`` to separate the
2787 numbers, and list the numbers in ascending order. For example,
2788 ``--percentile_list=99.5:99.9`` will cause fio to report the values of
2789 completion latency below which 99.5% and 99.9% of the observed latencies
2796 .. option:: exitall_on_error
2798 When one job finishes in error, terminate the rest. The default is to wait
2799 for each job to finish.
2801 .. option:: continue_on_error=str
2803 Normally fio will exit the job on the first observed failure. If this option
2804 is set, fio will continue the job when there is a 'non-fatal error' (EIO or
2805 EILSEQ) until the runtime is exceeded or the I/O size specified is
2806 completed. If this option is used, there are two more stats that are
2807 appended, the total error count and the first error. The error field given
2808 in the stats is the first error that was hit during the run.
2810 The allowed values are:
2813 Exit on any I/O or verify errors.
2816 Continue on read errors, exit on all others.
2819 Continue on write errors, exit on all others.
2822 Continue on any I/O error, exit on all others.
2825 Continue on verify errors, exit on all others.
2828 Continue on all errors.
2831 Backward-compatible alias for 'none'.
2834 Backward-compatible alias for 'all'.
2836 .. option:: ignore_error=str
2838 Sometimes you want to ignore some errors during test in that case you can
2839 specify error list for each error type.
2840 ``ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST`` errors for
2841 given error type is separated with ':'. Error may be symbol ('ENOSPC',
2842 'ENOMEM') or integer. Example::
2844 ignore_error=EAGAIN,ENOSPC:122
2846 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from
2849 .. option:: error_dump=bool
2851 If set dump every error even if it is non fatal, true by default. If
2852 disabled only fatal error will be dumped.
2854 Running predefined workloads
2855 ----------------------------
2857 Fio includes predefined profiles that mimic the I/O workloads generated by
2860 .. option:: profile=str
2862 The predefined workload to run. Current profiles are:
2865 Threaded I/O bench (tiotest/tiobench) like workload.
2868 Aerospike Certification Tool (ACT) like workload.
2870 To view a profile's additional options use :option:`--cmdhelp` after specifying
2871 the profile. For example::
2873 $ fio --profile=act --cmdhelp
2878 .. option:: device-names=str
2883 .. option:: load=int
2886 ACT load multiplier. Default: 1.
2888 .. option:: test-duration=time
2891 How long the entire test takes to run. Default: 24h.
2893 .. option:: threads-per-queue=int
2896 Number of read IO threads per device. Default: 8.
2898 .. option:: read-req-num-512-blocks=int
2901 Number of 512B blocks to read at the time. Default: 3.
2903 .. option:: large-block-op-kbytes=int
2906 Size of large block ops in KiB (writes). Default: 131072.
2911 Set to run ACT prep phase.
2913 Tiobench profile options
2914 ~~~~~~~~~~~~~~~~~~~~~~~~
2916 .. option:: size=str
2921 .. option:: block=int
2924 Block size in bytes. Default: 4096.
2926 .. option:: numruns=int
2936 .. option:: threads=int
2941 Interpreting the output
2942 -----------------------
2944 Fio spits out a lot of output. While running, fio will display the status of the
2945 jobs created. An example of that would be::
2947 Jobs: 1 (f=1): [_(1),M(1)][24.8%][r=20.5MiB/s,w=23.5MiB/s][r=82,w=94 IOPS][eta 01m:31s]
2949 The characters inside the square brackets denote the current status of each
2950 thread. The possible values (in typical life cycle order) are:
2952 +------+-----+-----------------------------------------------------------+
2954 +======+=====+===========================================================+
2955 | P | | Thread setup, but not started. |
2956 +------+-----+-----------------------------------------------------------+
2957 | C | | Thread created. |
2958 +------+-----+-----------------------------------------------------------+
2959 | I | | Thread initialized, waiting or generating necessary data. |
2960 +------+-----+-----------------------------------------------------------+
2961 | | p | Thread running pre-reading file(s). |
2962 +------+-----+-----------------------------------------------------------+
2963 | | R | Running, doing sequential reads. |
2964 +------+-----+-----------------------------------------------------------+
2965 | | r | Running, doing random reads. |
2966 +------+-----+-----------------------------------------------------------+
2967 | | W | Running, doing sequential writes. |
2968 +------+-----+-----------------------------------------------------------+
2969 | | w | Running, doing random writes. |
2970 +------+-----+-----------------------------------------------------------+
2971 | | M | Running, doing mixed sequential reads/writes. |
2972 +------+-----+-----------------------------------------------------------+
2973 | | m | Running, doing mixed random reads/writes. |
2974 +------+-----+-----------------------------------------------------------+
2975 | | F | Running, currently waiting for :manpage:`fsync(2)` |
2976 +------+-----+-----------------------------------------------------------+
2977 | | V | Running, doing verification of written data. |
2978 +------+-----+-----------------------------------------------------------+
2979 | E | | Thread exited, not reaped by main thread yet. |
2980 +------+-----+-----------------------------------------------------------+
2981 | _ | | Thread reaped, or |
2982 +------+-----+-----------------------------------------------------------+
2983 | X | | Thread reaped, exited with an error. |
2984 +------+-----+-----------------------------------------------------------+
2985 | K | | Thread reaped, exited due to signal. |
2986 +------+-----+-----------------------------------------------------------+
2988 Fio will condense the thread string as not to take up more space on the command
2989 line as is needed. For instance, if you have 10 readers and 10 writers running,
2990 the output would look like this::
2992 Jobs: 20 (f=20): [R(10),W(10)][4.0%][r=20.5MiB/s,w=23.5MiB/s][r=82,w=94 IOPS][eta 57m:36s]
2994 Fio will still maintain the ordering, though. So the above means that jobs 1..10
2995 are readers, and 11..20 are writers.
2997 The other values are fairly self explanatory -- number of threads currently
2998 running and doing I/O, the number of currently open files (f=), the rate of I/O
2999 since last check (read speed listed first, then write speed and optionally trim
3000 speed), and the estimated completion percentage and time for the current
3001 running group. It's impossible to estimate runtime of the following groups (if
3002 any). Note that the string is displayed in order, so it's possible to tell which
3003 of the jobs are currently doing what. The first character is the first job
3004 defined in the job file, and so forth.
3006 When fio is done (or interrupted by :kbd:`ctrl-c`), it will show the data for
3007 each thread, group of threads, and disks in that order. For each data direction,
3008 the output looks like::
3010 Client1 (g=0): err= 0:
3011 write: io= 32MiB, bw= 666KiB/s, iops=89 , runt= 50320msec
3012 slat (msec): min= 0, max= 136, avg= 0.03, stdev= 1.92
3013 clat (msec): min= 0, max= 631, avg=48.50, stdev=86.82
3014 bw (KiB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, stdev=681.68
3015 cpu : usr=1.49%, sys=0.25%, ctx=7969, majf=0, minf=17
3016 IO depths : 1=0.1%, 2=0.3%, 4=0.5%, 8=99.0%, 16=0.0%, 32=0.0%, >32=0.0%
3017 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
3018 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
3019 issued r/w: total=0/32768, short=0/0
3020 lat (msec): 2=1.6%, 4=0.0%, 10=3.2%, 20=12.8%, 50=38.4%, 100=24.8%,
3021 lat (msec): 250=15.2%, 500=0.0%, 750=0.0%, 1000=0.0%, >=2048=0.0%
3023 The client number is printed, along with the group id and error of that
3024 thread. Below is the I/O statistics, here for writes. In the order listed, they
3028 Number of megabytes I/O performed.
3031 Average bandwidth rate.
3034 Average I/Os performed per second.
3037 The runtime of that thread.
3040 Submission latency (avg being the average, stdev being the standard
3041 deviation). This is the time it took to submit the I/O. For sync I/O,
3042 the slat is really the completion latency, since queue/complete is one
3043 operation there. This value can be in milliseconds or microseconds, fio
3044 will choose the most appropriate base and print that. In the example
3045 above, milliseconds is the best scale. Note: in :option:`--minimal` mode
3046 latencies are always expressed in microseconds.
3049 Completion latency. Same names as slat, this denotes the time from
3050 submission to completion of the I/O pieces. For sync I/O, clat will
3051 usually be equal (or very close) to 0, as the time from submit to
3052 complete is basically just CPU time (I/O has already been done, see slat
3056 Bandwidth. Same names as the xlat stats, but also includes an
3057 approximate percentage of total aggregate bandwidth this thread received
3058 in this group. This last value is only really useful if the threads in
3059 this group are on the same disk, since they are then competing for disk
3063 CPU usage. User and system time, along with the number of context
3064 switches this thread went through, usage of system and user time, and
3065 finally the number of major and minor page faults. The CPU utilization
3066 numbers are averages for the jobs in that reporting group, while the
3067 context and fault counters are summed.
3070 The distribution of I/O depths over the job life time. The numbers are
3071 divided into powers of 2, so for example the 16= entries includes depths
3072 up to that value but higher than the previous entry. In other words, it
3073 covers the range from 16 to 31.
3076 How many pieces of I/O were submitting in a single submit call. Each
3077 entry denotes that amount and below, until the previous entry -- e.g.,
3078 8=100% mean that we submitted anywhere in between 5-8 I/Os per submit
3082 Like the above submit number, but for completions instead.
3085 The number of read/write requests issued, and how many of them were
3089 The distribution of I/O completion latencies. This is the time from when
3090 I/O leaves fio and when it gets completed. The numbers follow the same
3091 pattern as the I/O depths, meaning that 2=1.6% means that 1.6% of the
3092 I/O completed within 2 msecs, 20=12.8% means that 12.8% of the I/O took
3093 more than 10 msecs, but less than (or equal to) 20 msecs.
3095 After each client has been listed, the group statistics are printed. They
3096 will look like this::
3098 Run status group 0 (all jobs):
3099 READ: io=64MB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
3100 WRITE: io=64MB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
3102 For each data direction, it prints:
3105 Number of megabytes I/O performed.
3107 Aggregate bandwidth of threads in this group.
3109 The minimum average bandwidth a thread saw.
3111 The maximum average bandwidth a thread saw.
3113 The smallest runtime of the threads in that group.
3115 The longest runtime of the threads in that group.
3117 And finally, the disk statistics are printed. They will look like this::
3119 Disk stats (read/write):
3120 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
3122 Each value is printed for both reads and writes, with reads first. The
3126 Number of I/Os performed by all groups.
3128 Number of merges I/O the I/O scheduler.
3130 Number of ticks we kept the disk busy.
3132 Total time spent in the disk queue.
3134 The disk utilization. A value of 100% means we kept the disk
3135 busy constantly, 50% would be a disk idling half of the time.
3137 It is also possible to get fio to dump the current output while it is running,
3138 without terminating the job. To do that, send fio the **USR1** signal. You can
3139 also get regularly timed dumps by using the :option:`--status-interval`
3140 parameter, or by creating a file in :file:`/tmp` named
3141 :file:`fio-dump-status`. If fio sees this file, it will unlink it and dump the
3142 current output status.
3148 For scripted usage where you typically want to generate tables or graphs of the
3149 results, fio can output the results in a semicolon separated format. The format
3150 is one long line of values, such as::
3152 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%
3153 A description of this job goes here.
3155 The job description (if provided) follows on a second line.
3157 To enable terse output, use the :option:`--minimal` command line option. The
3158 first value is the version of the terse output format. If the output has to be
3159 changed for some reason, this number will be incremented by 1 to signify that
3162 Split up, the format is as follows:
3166 terse version, fio version, jobname, groupid, error
3170 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3171 Submission latency: min, max, mean, stdev (usec)
3172 Completion latency: min, max, mean, stdev (usec)
3173 Completion latency percentiles: 20 fields (see below)
3174 Total latency: min, max, mean, stdev (usec)
3175 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev
3181 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3182 Submission latency: min, max, mean, stdev (usec)
3183 Completion latency: min, max, mean, stdev(usec)
3184 Completion latency percentiles: 20 fields (see below)
3185 Total latency: min, max, mean, stdev (usec)
3186 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev
3190 user, system, context switches, major faults, minor faults
3194 <=1, 2, 4, 8, 16, 32, >=64
3196 I/O latencies microseconds::
3198 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
3200 I/O latencies milliseconds::
3202 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
3206 Disk name, Read ios, write ios,
3207 Read merges, write merges,
3208 Read ticks, write ticks,
3209 Time spent in queue, disk utilization percentage
3211 Additional Info (dependent on continue_on_error, default off)::
3213 total # errors, first error code
3215 Additional Info (dependent on description being set)::
3219 Completion latency percentiles can be a grouping of up to 20 sets, so for the
3220 terse output fio writes all of them. Each field will look like this::
3224 which is the Xth percentile, and the `usec` latency associated with it.
3226 For disk utilization, all disks used by fio are shown. So for each disk there
3227 will be a disk utilization section.
3229 Below is a single line containing short names for each of the fields in the
3230 minimal output v3, separated by semicolons:
3232 terse_version_3;fio_version;jobname;groupid;error;read_kb;read_bandwidth;read_iops;read_runtime_ms;read_slat_min;read_slat_max;read_slat_mean;read_slat_dev;read_clat_max;read_clat_min;read_clat_mean;read_clat_dev;read_clat_pct01;read_clat_pct02;read_clat_pct03;read_clat_pct04;read_clat_pct05;read_clat_pct06;read_clat_pct07;read_clat_pct08;read_clat_pct09;read_clat_pct10;read_clat_pct11;read_clat_pct12;read_clat_pct13;read_clat_pct14;read_clat_pct15;read_clat_pct16;read_clat_pct17;read_clat_pct18;read_clat_pct19;read_clat_pct20;read_tlat_min;read_lat_max;read_lat_mean;read_lat_dev;read_bw_min;read_bw_max;read_bw_agg_pct;read_bw_mean;read_bw_dev;write_kb;write_bandwidth;write_iops;write_runtime_ms;write_slat_min;write_slat_max;write_slat_mean;write_slat_dev;write_clat_max;write_clat_min;write_clat_mean;write_clat_dev;write_clat_pct01;write_clat_pct02;write_clat_pct03;write_clat_pct04;write_clat_pct05;write_clat_pct06;write_clat_pct07;write_clat_pct08;write_clat_pct09;write_clat_pct10;write_clat_pct11;write_clat_pct12;write_clat_pct13;write_clat_pct14;write_clat_pct15;write_clat_pct16;write_clat_pct17;write_clat_pct18;write_clat_pct19;write_clat_pct20;write_tlat_min;write_lat_max;write_lat_mean;write_lat_dev;write_bw_min;write_bw_max;write_bw_agg_pct;write_bw_mean;write_bw_dev;cpu_user;cpu_sys;cpu_csw;cpu_mjf;pu_minf;iodepth_1;iodepth_2;iodepth_4;iodepth_8;iodepth_16;iodepth_32;iodepth_64;lat_2us;lat_4us;lat_10us;lat_20us;lat_50us;lat_100us;lat_250us;lat_500us;lat_750us;lat_1000us;lat_2ms;lat_4ms;lat_10ms;lat_20ms;lat_50ms;lat_100ms;lat_250ms;lat_500ms;lat_750ms;lat_1000ms;lat_2000ms;lat_over_2000ms;disk_name;disk_read_iops;disk_write_iops;disk_read_merges;disk_write_merges;disk_read_ticks;write_ticks;disk_queue_time;disk_util
3238 There are two trace file format that you can encounter. The older (v1) format is
3239 unsupported since version 1.20-rc3 (March 2008). It will still be described
3240 below in case that you get an old trace and want to understand it.
3242 In any case the trace is a simple text file with a single action per line.
3245 Trace file format v1
3246 ~~~~~~~~~~~~~~~~~~~~
3248 Each line represents a single I/O action in the following format::
3252 where `rw=0/1` for read/write, and the offset and length entries being in bytes.
3254 This format is not supported in fio versions => 1.20-rc3.
3257 Trace file format v2
3258 ~~~~~~~~~~~~~~~~~~~~
3260 The second version of the trace file format was added in fio version 1.17. It
3261 allows to access more then one file per trace and has a bigger set of possible
3264 The first line of the trace file has to be::
3268 Following this can be lines in two different formats, which are described below.
3270 The file management format::
3274 The filename is given as an absolute path. The action can be one of these:
3277 Add the given filename to the trace.
3279 Open the file with the given filename. The filename has to have
3280 been added with the **add** action before.
3282 Close the file with the given filename. The file has to have been
3286 The file I/O action format::
3288 filename action offset length
3290 The `filename` is given as an absolute path, and has to have been added and
3291 opened before it can be used with this format. The `offset` and `length` are
3292 given in bytes. The `action` can be one of these:
3295 Wait for `offset` microseconds. Everything below 100 is discarded.
3296 The time is relative to the previous `wait` statement.
3298 Read `length` bytes beginning from `offset`.
3300 Write `length` bytes beginning from `offset`.
3302 :manpage:`fsync(2)` the file.
3304 :manpage:`fdatasync(2)` the file.
3306 Trim the given file from the given `offset` for `length` bytes.
3308 CPU idleness profiling
3309 ----------------------
3311 In some cases, we want to understand CPU overhead in a test. For example, we
3312 test patches for the specific goodness of whether they reduce CPU usage.
3313 Fio implements a balloon approach to create a thread per CPU that runs at idle
3314 priority, meaning that it only runs when nobody else needs the cpu.
3315 By measuring the amount of work completed by the thread, idleness of each CPU
3316 can be derived accordingly.
3318 An unit work is defined as touching a full page of unsigned characters. Mean and
3319 standard deviation of time to complete an unit work is reported in "unit work"
3320 section. Options can be chosen to report detailed percpu idleness or overall
3321 system idleness by aggregating percpu stats.
3324 Verification and triggers
3325 -------------------------
3327 Fio is usually run in one of two ways, when data verification is done. The first
3328 is a normal write job of some sort with verify enabled. When the write phase has
3329 completed, fio switches to reads and verifies everything it wrote. The second
3330 model is running just the write phase, and then later on running the same job
3331 (but with reads instead of writes) to repeat the same I/O patterns and verify
3332 the contents. Both of these methods depend on the write phase being completed,
3333 as fio otherwise has no idea how much data was written.
3335 With verification triggers, fio supports dumping the current write state to
3336 local files. Then a subsequent read verify workload can load this state and know
3337 exactly where to stop. This is useful for testing cases where power is cut to a
3338 server in a managed fashion, for instance.
3340 A verification trigger consists of two things:
3342 1) Storing the write state of each job.
3343 2) Executing a trigger command.
3345 The write state is relatively small, on the order of hundreds of bytes to single
3346 kilobytes. It contains information on the number of completions done, the last X
3349 A trigger is invoked either through creation ('touch') of a specified file in
3350 the system, or through a timeout setting. If fio is run with
3351 :option:`--trigger-file` = :file:`/tmp/trigger-file`, then it will continually
3352 check for the existence of :file:`/tmp/trigger-file`. When it sees this file, it
3353 will fire off the trigger (thus saving state, and executing the trigger
3356 For client/server runs, there's both a local and remote trigger. If fio is
3357 running as a server backend, it will send the job states back to the client for
3358 safe storage, then execute the remote trigger, if specified. If a local trigger
3359 is specified, the server will still send back the write state, but the client
3360 will then execute the trigger.
3362 Verification trigger example
3363 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3365 Lets say we want to run a powercut test on the remote machine 'server'. Our
3366 write workload is in :file:`write-test.fio`. We want to cut power to 'server' at
3367 some point during the run, and we'll run this test from the safety or our local
3368 machine, 'localbox'. On the server, we'll start the fio backend normally::
3370 server# fio --server
3372 and on the client, we'll fire off the workload::
3374 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger-remote="bash -c \"echo b > /proc/sysrq-triger\""
3376 We set :file:`/tmp/my-trigger` as the trigger file, and we tell fio to execute::
3378 echo b > /proc/sysrq-trigger
3380 on the server once it has received the trigger and sent us the write state. This
3381 will work, but it's not **really** cutting power to the server, it's merely
3382 abruptly rebooting it. If we have a remote way of cutting power to the server
3383 through IPMI or similar, we could do that through a local trigger command
3384 instead. Lets assume we have a script that does IPMI reboot of a given hostname,
3385 ipmi-reboot. On localbox, we could then have run fio with a local trigger
3388 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger="ipmi-reboot server"
3390 For this case, fio would wait for the server to send us the write state, then
3391 execute ``ipmi-reboot server`` when that happened.
3393 Loading verify state
3394 ~~~~~~~~~~~~~~~~~~~~
3396 To load store write state, read verification job file must contain the
3397 :option:`verify_state_load` option. If that is set, fio will load the previously
3398 stored state. For a local fio run this is done by loading the files directly,
3399 and on a client/server run, the server backend will ask the client to send the
3400 files over and load them from there.
3406 Fio supports a variety of log file formats, for logging latencies, bandwidth,
3407 and IOPS. The logs share a common format, which looks like this:
3409 *time* (`msec`), *value*, *data direction*, *offset*
3411 Time for the log entry is always in milliseconds. The *value* logged depends
3412 on the type of log, it will be one of the following:
3415 Value is latency in usecs
3421 *Data direction* is one of the following:
3430 The *offset* is the offset, in bytes, from the start of the file, for that
3431 particular I/O. The logging of the offset can be toggled with
3432 :option:`log_offset`.
3434 If windowed logging is enabled through :option:`log_avg_msec` then fio doesn't
3435 log individual I/Os. Instead of logs the average values over the specified period
3436 of time. Since 'data direction' and 'offset' are per-I/O values, they aren't
3437 applicable if windowed logging is enabled. If windowed logging is enabled and
3438 :option:`log_max_value` is set, then fio logs maximum values in that window
3439 instead of averages.
3445 Normally fio is invoked as a stand-alone application on the machine where the
3446 I/O workload should be generated. However, the frontend and backend of fio can
3447 be run separately. Ie the fio server can generate an I/O workload on the "Device
3448 Under Test" while being controlled from another machine.
3450 Start the server on the machine which has access to the storage DUT::
3454 where args defines what fio listens to. The arguments are of the form
3455 ``type,hostname`` or ``IP,port``. *type* is either ``ip`` (or ip4) for TCP/IP
3456 v4, ``ip6`` for TCP/IP v6, or ``sock`` for a local unix domain socket.
3457 *hostname* is either a hostname or IP address, and *port* is the port to listen
3458 to (only valid for TCP/IP, not a local socket). Some examples:
3462 Start a fio server, listening on all interfaces on the default port (8765).
3464 2) ``fio --server=ip:hostname,4444``
3466 Start a fio server, listening on IP belonging to hostname and on port 4444.
3468 3) ``fio --server=ip6:::1,4444``
3470 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
3472 4) ``fio --server=,4444``
3474 Start a fio server, listening on all interfaces on port 4444.
3476 5) ``fio --server=1.2.3.4``
3478 Start a fio server, listening on IP 1.2.3.4 on the default port.
3480 6) ``fio --server=sock:/tmp/fio.sock``
3482 Start a fio server, listening on the local socket /tmp/fio.sock.
3484 Once a server is running, a "client" can connect to the fio server with::
3486 fio <local-args> --client=<server> <remote-args> <job file(s)>
3488 where `local-args` are arguments for the client where it is running, `server`
3489 is the connect string, and `remote-args` and `job file(s)` are sent to the
3490 server. The `server` string follows the same format as it does on the server
3491 side, to allow IP/hostname/socket and port strings.
3493 Fio can connect to multiple servers this way::
3495 fio --client=<server1> <job file(s)> --client=<server2> <job file(s)>
3497 If the job file is located on the fio server, then you can tell the server to
3498 load a local file as well. This is done by using :option:`--remote-config` ::
3500 fio --client=server --remote-config /path/to/file.fio
3502 Then fio will open this local (to the server) job file instead of being passed
3503 one from the client.
3505 If you have many servers (example: 100 VMs/containers), you can input a pathname
3506 of a file containing host IPs/names as the parameter value for the
3507 :option:`--client` option. For example, here is an example :file:`host.list`
3508 file containing 2 hostnames::
3510 host1.your.dns.domain
3511 host2.your.dns.domain
3513 The fio command would then be::
3515 fio --client=host.list <job file(s)>
3517 In this mode, you cannot input server-specific parameters or job files -- all
3518 servers receive the same job file.
3520 In order to let ``fio --client`` runs use a shared filesystem from multiple
3521 hosts, ``fio --client`` now prepends the IP address of the server to the
3522 filename. For example, if fio is using directory :file:`/mnt/nfs/fio` and is
3523 writing filename :file:`fileio.tmp`, with a :option:`--client` `hostfile`
3524 containing two hostnames ``h1`` and ``h2`` with IP addresses 192.168.10.120 and
3525 192.168.10.121, then fio will create two files::
3527 /mnt/nfs/fio/192.168.10.120.fileio.tmp
3528 /mnt/nfs/fio/192.168.10.121.fileio.tmp