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, separated 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 or 5).
132 .. option:: --version
134 Print version info and exit.
138 Print a summary of the command line options and exit.
140 .. option:: --cpuclock-test
142 Perform test and validation of internal CPU clock.
144 .. option:: --crctest=[test]
146 Test the speed of the built-in checksumming functions. If no argument is
147 given all of them are tested. Alternatively, a comma separated list can 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. When the unit is omitted,
181 the value is interpreted in seconds.
183 .. option:: --status-interval=time
185 Force full status dump every `time` period passed. When the unit is
186 omitted, the value is interpreted in seconds.
188 .. option:: --section=name
190 Only run specified section in job file. Multiple sections can be specified.
191 The ``--section`` option allows one to combine related jobs into one file.
192 E.g. one job file could define light, moderate, and heavy sections. Tell
193 fio to run only the "heavy" section by giving ``--section=heavy``
194 command line option. One can also specify the "write" operations in one
195 section and "verify" operation in another section. The ``--section`` option
196 only applies to job sections. The reserved *global* section is always
199 .. option:: --alloc-size=kb
201 Set the internal smalloc pool to this size in KiB. The
202 ``--alloc-size`` switch allows one to use a larger pool size for smalloc.
203 If running large jobs with randommap enabled, fio can run out of memory.
204 Smalloc is an internal allocator for shared structures from a fixed size
205 memory pool and can grow to 16 pools. The pool size defaults to 16MiB.
207 NOTE: While running :file:`.fio_smalloc.*` backing store files are visible
210 .. option:: --warnings-fatal
212 All fio parser warnings are fatal, causing fio to exit with an
215 .. option:: --max-jobs=nr
217 Maximum number of threads/processes to support.
219 .. option:: --server=args
221 Start a backend server, with `args` specifying what to listen to.
222 See `Client/Server`_ section.
224 .. option:: --daemonize=pidfile
226 Background a fio server, writing the pid to the given `pidfile` file.
228 .. option:: --client=hostname
230 Instead of running the jobs locally, send and run them on the given host or
231 set of hosts. See `Client/Server`_ section.
233 .. option:: --remote-config=file
235 Tell fio server to load this local file.
237 .. option:: --idle-prof=option
239 Report CPU idleness. *option* is one of the following:
242 Run unit work calibration only and exit.
245 Show aggregate system idleness and unit work.
248 As **system** but also show per CPU idleness.
250 .. option:: --inflate-log=log
252 Inflate and output compressed log.
254 .. option:: --trigger-file=file
256 Execute trigger cmd when file exists.
258 .. option:: --trigger-timeout=t
260 Execute trigger at this time.
262 .. option:: --trigger=cmd
264 Set this command as local trigger.
266 .. option:: --trigger-remote=cmd
268 Set this command as remote trigger.
270 .. option:: --aux-path=path
272 Use this path for fio state generated files.
274 Any parameters following the options will be assumed to be job files, unless
275 they match a job file parameter. Multiple job files can be listed and each job
276 file will be regarded as a separate group. Fio will :option:`stonewall`
277 execution between each group.
283 As previously described, fio accepts one or more job files describing what it is
284 supposed to do. The job file format is the classic ini file, where the names
285 enclosed in [] brackets define the job name. You are free to use any ASCII name
286 you want, except *global* which has special meaning. Following the job name is
287 a sequence of zero or more parameters, one per line, that define the behavior of
288 the job. If the first character in a line is a ';' or a '#', the entire line is
289 discarded as a comment.
291 A *global* section sets defaults for the jobs described in that file. A job may
292 override a *global* section parameter, and a job file may even have several
293 *global* sections if so desired. A job is only affected by a *global* section
296 The :option:`--cmdhelp` option also lists all options. If used with an `option`
297 argument, :option:`--cmdhelp` will detail the given `option`.
299 See the `examples/` directory for inspiration on how to write job files. Note
300 the copyright and license requirements currently apply to `examples/` files.
302 So let's look at a really simple job file that defines two processes, each
303 randomly reading from a 128MiB file:
307 ; -- start job file --
318 As you can see, the job file sections themselves are empty as all the described
319 parameters are shared. As no :option:`filename` option is given, fio makes up a
320 `filename` for each of the jobs as it sees fit. On the command line, this job
321 would look as follows::
323 $ fio --name=global --rw=randread --size=128m --name=job1 --name=job2
326 Let's look at an example that has a number of processes writing randomly to
331 ; -- start job file --
342 Here we have no *global* section, as we only have one job defined anyway. We
343 want to use async I/O here, with a depth of 4 for each file. We also increased
344 the buffer size used to 32KiB and define numjobs to 4 to fork 4 identical
345 jobs. The result is 4 processes each randomly writing to their own 64MiB
346 file. Instead of using the above job file, you could have given the parameters
347 on the command line. For this case, you would specify::
349 $ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4
351 When fio is utilized as a basis of any reasonably large test suite, it might be
352 desirable to share a set of standardized settings across multiple job files.
353 Instead of copy/pasting such settings, any section may pull in an external
354 :file:`filename.fio` file with *include filename* directive, as in the following
357 ; -- start job file including.fio --
361 include glob-include.fio
368 include test-include.fio
369 ; -- end job file including.fio --
373 ; -- start job file glob-include.fio --
376 ; -- end job file glob-include.fio --
380 ; -- start job file test-include.fio --
383 ; -- end job file test-include.fio --
385 Settings pulled into a section apply to that section only (except *global*
386 section). Include directives may be nested in that any included file may contain
387 further include directive(s). Include files may not contain [] sections.
390 Environment variables
391 ~~~~~~~~~~~~~~~~~~~~~
393 Fio also supports environment variable expansion in job files. Any sub-string of
394 the form ``${VARNAME}`` as part of an option value (in other words, on the right
395 of the '='), will be expanded to the value of the environment variable called
396 `VARNAME`. If no such environment variable is defined, or `VARNAME` is the
397 empty string, the empty string will be substituted.
399 As an example, let's look at a sample fio invocation and job file::
401 $ SIZE=64m NUMJOBS=4 fio jobfile.fio
405 ; -- start job file --
412 This will expand to the following equivalent job file at runtime:
416 ; -- start job file --
423 Fio ships with a few example job files, you can also look there for inspiration.
428 Additionally, fio has a set of reserved keywords that will be replaced
429 internally with the appropriate value. Those keywords are:
433 The architecture page size of the running system.
437 Megabytes of total memory in the system.
441 Number of online available CPUs.
443 These can be used on the command line or in the job file, and will be
444 automatically substituted with the current system values when the job is
445 run. Simple math is also supported on these keywords, so you can perform actions
450 and get that properly expanded to 8 times the size of memory in the machine.
456 This section describes in details each parameter associated with a job. Some
457 parameters take an option of a given type, such as an integer or a
458 string. Anywhere a numeric value is required, an arithmetic expression may be
459 used, provided it is surrounded by parentheses. Supported operators are:
468 For time values in expressions, units are microseconds by default. This is
469 different than for time values not in expressions (not enclosed in
470 parentheses). The following types are used:
477 String. This is a sequence of alpha characters.
480 Integer with possible time suffix. Without a unit value is interpreted as
481 seconds unless otherwise specified. Accepts a suffix of 'd' for days, 'h' for
482 hours, 'm' for minutes, 's' for seconds, 'ms' (or 'msec') for milliseconds and
483 'us' (or 'usec') for microseconds. For example, use 10m for 10 minutes.
488 Integer. A whole number value, which may contain an integer prefix
489 and an integer suffix:
491 [*integer prefix*] **number** [*integer suffix*]
493 The optional *integer prefix* specifies the number's base. The default
494 is decimal. *0x* specifies hexadecimal.
496 The optional *integer suffix* specifies the number's units, and includes an
497 optional unit prefix and an optional unit. For quantities of data, the
498 default unit is bytes. For quantities of time, the default unit is seconds
499 unless otherwise specified.
501 With :option:`kb_base`\=1000, fio follows international standards for unit
502 prefixes. To specify power-of-10 decimal values defined in the
503 International System of Units (SI):
505 * *Ki* -- means kilo (K) or 1000
506 * *Mi* -- means mega (M) or 1000**2
507 * *Gi* -- means giga (G) or 1000**3
508 * *Ti* -- means tera (T) or 1000**4
509 * *Pi* -- means peta (P) or 1000**5
511 To specify power-of-2 binary values defined in IEC 80000-13:
513 * *k* -- means kibi (Ki) or 1024
514 * *M* -- means mebi (Mi) or 1024**2
515 * *G* -- means gibi (Gi) or 1024**3
516 * *T* -- means tebi (Ti) or 1024**4
517 * *P* -- means pebi (Pi) or 1024**5
519 With :option:`kb_base`\=1024 (the default), the unit prefixes are opposite
520 from those specified in the SI and IEC 80000-13 standards to provide
521 compatibility with old scripts. For example, 4k means 4096.
523 For quantities of data, an optional unit of 'B' may be included
524 (e.g., 'kB' is the same as 'k').
526 The *integer suffix* is not case sensitive (e.g., m/mi mean mebi/mega,
527 not milli). 'b' and 'B' both mean byte, not bit.
529 Examples with :option:`kb_base`\=1000:
531 * *4 KiB*: 4096, 4096b, 4096B, 4ki, 4kib, 4kiB, 4Ki, 4KiB
532 * *1 MiB*: 1048576, 1mi, 1024ki
533 * *1 MB*: 1000000, 1m, 1000k
534 * *1 TiB*: 1099511627776, 1ti, 1024gi, 1048576mi
535 * *1 TB*: 1000000000, 1t, 1000m, 1000000k
537 Examples with :option:`kb_base`\=1024 (default):
539 * *4 KiB*: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB
540 * *1 MiB*: 1048576, 1m, 1024k
541 * *1 MB*: 1000000, 1mi, 1000ki
542 * *1 TiB*: 1099511627776, 1t, 1024g, 1048576m
543 * *1 TB*: 1000000000, 1ti, 1000mi, 1000000ki
545 To specify times (units are not case sensitive):
549 * *M* -- means minutes
550 * *s* -- or sec means seconds (default)
551 * *ms* -- or *msec* means milliseconds
552 * *us* -- or *usec* means microseconds
554 If the option accepts an upper and lower range, use a colon ':' or
555 minus '-' to separate such values. See :ref:`irange <irange>`.
556 If the lower value specified happens to be larger than the upper value
557 the two values are swapped.
562 Boolean. Usually parsed as an integer, however only defined for
563 true and false (1 and 0).
568 Integer range with suffix. Allows value range to be given, such as
569 1024-4096. A colon may also be used as the separator, e.g. 1k:4k. If the
570 option allows two sets of ranges, they can be specified with a ',' or '/'
571 delimiter: 1k-4k/8k-32k. Also see :ref:`int <int>`.
574 A list of floating point numbers, separated by a ':' character.
580 .. option:: kb_base=int
582 Select the interpretation of unit prefixes in input parameters.
585 Inputs comply with IEC 80000-13 and the International
586 System of Units (SI). Use:
588 - power-of-2 values with IEC prefixes (e.g., KiB)
589 - power-of-10 values with SI prefixes (e.g., kB)
592 Compatibility mode (default). To avoid breaking old scripts:
594 - power-of-2 values with SI prefixes
595 - power-of-10 values with IEC prefixes
597 See :option:`bs` for more details on input parameters.
599 Outputs always use correct prefixes. Most outputs include both
602 bw=2383.3kB/s (2327.4KiB/s)
604 If only one value is reported, then kb_base selects the one to use:
606 **1000** -- SI prefixes
608 **1024** -- IEC prefixes
610 .. option:: unit_base=int
612 Base unit for reporting. Allowed values are:
615 Use auto-detection (default).
622 With the above in mind, here follows the complete list of fio job parameters.
630 ASCII name of the job. This may be used to override the name printed by fio
631 for this job. Otherwise the job name is used. On the command line this
632 parameter has the special purpose of also signaling the start of a new job.
634 .. option:: description=str
636 Text description of the job. Doesn't do anything except dump this text
637 description when this job is run. It's not parsed.
639 .. option:: loops=int
641 Run the specified number of iterations of this job. Used to repeat the same
642 workload a given number of times. Defaults to 1.
644 .. option:: numjobs=int
646 Create the specified number of clones of this job. Each clone of job
647 is spawned as an independent thread or process. May be used to setup a
648 larger number of threads/processes doing the same thing. Each thread is
649 reported separately; to see statistics for all clones as a whole, use
650 :option:`group_reporting` in conjunction with :option:`new_group`.
651 See :option:`--max-jobs`. Default: 1.
654 Time related parameters
655 ~~~~~~~~~~~~~~~~~~~~~~~
657 .. option:: runtime=time
659 Tell fio to terminate processing after the specified period of time. It
660 can be quite hard to determine for how long a specified job will run, so
661 this parameter is handy to cap the total runtime to a given time. When
662 the unit is omitted, the value is intepreted in seconds.
664 .. option:: time_based
666 If set, fio will run for the duration of the :option:`runtime` specified
667 even if the file(s) are completely read or written. It will simply loop over
668 the same workload as many times as the :option:`runtime` allows.
670 .. option:: startdelay=irange(time)
672 Delay the start of job for the specified amount of time. Can be a single
673 value or a range. When given as a range, each thread will choose a value
674 randomly from within the range. Value is in seconds if a unit is omitted.
676 .. option:: ramp_time=time
678 If set, fio will run the specified workload for this amount of time before
679 logging any performance numbers. Useful for letting performance settle
680 before logging results, thus minimizing the runtime required for stable
681 results. Note that the ``ramp_time`` is considered lead in time for a job,
682 thus it will increase the total runtime if a special timeout or
683 :option:`runtime` is specified. When the unit is omitted, the value is
686 .. option:: clocksource=str
688 Use the given clocksource as the base of timing. The supported options are:
691 :manpage:`gettimeofday(2)`
694 :manpage:`clock_gettime(2)`
697 Internal CPU clock source
699 cpu is the preferred clocksource if it is reliable, as it is very fast (and
700 fio is heavy on time calls). Fio will automatically use this clocksource if
701 it's supported and considered reliable on the system it is running on,
702 unless another clocksource is specifically set. For x86/x86-64 CPUs, this
703 means supporting TSC Invariant.
705 .. option:: gtod_reduce=bool
707 Enable all of the :manpage:`gettimeofday(2)` reducing options
708 (:option:`disable_clat`, :option:`disable_slat`, :option:`disable_bw_measurement`) plus
709 reduce precision of the timeout somewhat to really shrink the
710 :manpage:`gettimeofday(2)` call count. With this option enabled, we only do
711 about 0.4% of the :manpage:`gettimeofday(2)` calls we would have done if all
712 time keeping was enabled.
714 .. option:: gtod_cpu=int
716 Sometimes it's cheaper to dedicate a single thread of execution to just
717 getting the current time. Fio (and databases, for instance) are very
718 intensive on :manpage:`gettimeofday(2)` calls. With this option, you can set
719 one CPU aside for doing nothing but logging current time to a shared memory
720 location. Then the other threads/processes that run I/O workloads need only
721 copy that segment, instead of entering the kernel with a
722 :manpage:`gettimeofday(2)` call. The CPU set aside for doing these time
723 calls will be excluded from other uses. Fio will manually clear it from the
724 CPU mask of other jobs.
730 .. option:: directory=str
732 Prefix filenames with this directory. Used to place files in a different
733 location than :file:`./`. You can specify a number of directories by
734 separating the names with a ':' character. These directories will be
735 assigned equally distributed to job clones created by :option:`numjobs` as
736 long as they are using generated filenames. If specific `filename(s)` are
737 set fio will use the first listed directory, and thereby matching the
738 `filename` semantic which generates a file each clone if not specified, but
739 let all clones use the same if set.
741 See the :option:`filename` option for information on how to escape "``:``" and
742 "``\``" characters within the directory path itself.
744 .. option:: filename=str
746 Fio normally makes up a `filename` based on the job name, thread number, and
747 file number (see :option:`filename_format`). If you want to share files
748 between threads in a job or several
749 jobs with fixed file paths, specify a `filename` for each of them to override
750 the default. If the ioengine is file based, you can specify a number of files
751 by separating the names with a ':' colon. So if you wanted a job to open
752 :file:`/dev/sda` and :file:`/dev/sdb` as the two working files, you would use
753 ``filename=/dev/sda:/dev/sdb``. This also means that whenever this option is
754 specified, :option:`nrfiles` is ignored. The size of regular files specified
755 by this option will be :option:`size` divided by number of files unless an
756 explicit size is specified by :option:`filesize`.
758 Each colon and backslash in the wanted path must be escaped with a ``\``
759 character. For instance, if the path is :file:`/dev/dsk/foo@3,0:c` then you
760 would use ``filename=/dev/dsk/foo@3,0\:c`` and if the path is
761 :file:`F:\\filename` then you would use ``filename=F\:\\filename``.
763 On Windows, disk devices are accessed as :file:`\\\\.\\PhysicalDrive0` for
764 the first device, :file:`\\\\.\\PhysicalDrive1` for the second etc.
765 Note: Windows and FreeBSD prevent write access to areas
766 of the disk containing in-use data (e.g. filesystems).
768 The filename "`-`" is a reserved name, meaning *stdin* or *stdout*. Which
769 of the two depends on the read/write direction set.
771 .. option:: filename_format=str
773 If sharing multiple files between jobs, it is usually necessary to have fio
774 generate the exact names that you want. By default, fio will name a file
775 based on the default file format specification of
776 :file:`jobname.jobnumber.filenumber`. With this option, that can be
777 customized. Fio will recognize and replace the following keywords in this
781 The name of the worker thread or process.
783 The incremental number of the worker thread or process.
785 The incremental number of the file for that worker thread or
788 To have dependent jobs share a set of files, this option can be set to have
789 fio generate filenames that are shared between the two. For instance, if
790 :file:`testfiles.$filenum` is specified, file number 4 for any job will be
791 named :file:`testfiles.4`. The default of :file:`$jobname.$jobnum.$filenum`
792 will be used if no other format specifier is given.
794 .. option:: unique_filename=bool
796 To avoid collisions between networked clients, fio defaults to prefixing any
797 generated filenames (with a directory specified) with the source of the
798 client connecting. To disable this behavior, set this option to 0.
800 .. option:: opendir=str
802 Recursively open any files below directory `str`.
804 .. option:: lockfile=str
806 Fio defaults to not locking any files before it does I/O to them. If a file
807 or file descriptor is shared, fio can serialize I/O to that file to make the
808 end result consistent. This is usual for emulating real workloads that share
809 files. The lock modes are:
812 No locking. The default.
814 Only one thread or process may do I/O at a time, excluding all
817 Read-write locking on the file. Many readers may
818 access the file at the same time, but writes get exclusive access.
820 .. option:: nrfiles=int
822 Number of files to use for this job. Defaults to 1. The size of files
823 will be :option:`size` divided by this unless explicit size is specified by
824 :option:`filesize`. Files are created for each thread separately, and each
825 file will have a file number within its name by default, as explained in
826 :option:`filename` section.
829 .. option:: openfiles=int
831 Number of files to keep open at the same time. Defaults to the same as
832 :option:`nrfiles`, can be set smaller to limit the number simultaneous
835 .. option:: file_service_type=str
837 Defines how fio decides which file from a job to service next. The following
841 Choose a file at random.
844 Round robin over opened files. This is the default.
847 Finish one file before moving on to the next. Multiple files can
848 still be open depending on 'openfiles'.
851 Use a *Zipf* distribution to decide what file to access.
854 Use a *Pareto* distribution to decide what file to access.
857 Use a *Gaussian* (normal) distribution to decide what file to
860 For *random*, *roundrobin*, and *sequential*, a postfix can be appended to
861 tell fio how many I/Os to issue before switching to a new file. For example,
862 specifying ``file_service_type=random:8`` would cause fio to issue
863 8 I/Os before selecting a new file at random. For the non-uniform
864 distributions, a floating point postfix can be given to influence how the
865 distribution is skewed. See :option:`random_distribution` for a description
866 of how that would work.
868 .. option:: ioscheduler=str
870 Attempt to switch the device hosting the file to the specified I/O scheduler
873 .. option:: create_serialize=bool
875 If true, serialize the file creation for the jobs. This may be handy to
876 avoid interleaving of data files, which may greatly depend on the filesystem
877 used and even the number of processors in the system. Default: true.
879 .. option:: create_fsync=bool
881 :manpage:`fsync(2)` the data file after creation. This is the default.
883 .. option:: create_on_open=bool
885 If true, don't pre-create files but allow the job's open() to create a file
886 when it's time to do I/O. Default: false -- pre-create all necessary files
889 .. option:: create_only=bool
891 If true, fio will only run the setup phase of the job. If files need to be
892 laid out or updated on disk, only that will be done -- the actual job contents
893 are not executed. Default: false.
895 .. option:: allow_file_create=bool
897 If true, fio is permitted to create files as part of its workload. If this
898 option is false, then fio will error out if
899 the files it needs to use don't already exist. Default: true.
901 .. option:: allow_mounted_write=bool
903 If this isn't set, fio will abort jobs that are destructive (e.g. that write)
904 to what appears to be a mounted device or partition. This should help catch
905 creating inadvertently destructive tests, not realizing that the test will
906 destroy data on the mounted file system. Note that some platforms don't allow
907 writing against a mounted device regardless of this option. Default: false.
909 .. option:: pre_read=bool
911 If this is given, files will be pre-read into memory before starting the
912 given I/O operation. This will also clear the :option:`invalidate` flag,
913 since it is pointless to pre-read and then drop the cache. This will only
914 work for I/O engines that are seek-able, since they allow you to read the
915 same data multiple times. Thus it will not work on non-seekable I/O engines
916 (e.g. network, splice). Default: false.
918 .. option:: unlink=bool
920 Unlink the job files when done. Not the default, as repeated runs of that
921 job would then waste time recreating the file set again and again. Default:
924 .. option:: unlink_each_loop=bool
926 Unlink job files after each iteration or loop. Default: false.
928 .. option:: zonesize=int
930 Divide a file into zones of the specified size. See :option:`zoneskip`.
932 .. option:: zonerange=int
934 Give size of an I/O zone. See :option:`zoneskip`.
936 .. option:: zoneskip=int
938 Skip the specified number of bytes when :option:`zonesize` data has been
939 read. The two zone options can be used to only do I/O on zones of a file.
945 .. option:: direct=bool
947 If value is true, use non-buffered I/O. This is usually O_DIRECT. Note that
948 ZFS on Solaris doesn't support direct I/O. On Windows the synchronous
949 ioengines don't support direct I/O. Default: false.
951 .. option:: atomic=bool
953 If value is true, attempt to use atomic direct I/O. Atomic writes are
954 guaranteed to be stable once acknowledged by the operating system. Only
955 Linux supports O_ATOMIC right now.
957 .. option:: buffered=bool
959 If value is true, use buffered I/O. This is the opposite of the
960 :option:`direct` option. Defaults to true.
962 .. option:: readwrite=str, rw=str
964 Type of I/O pattern. Accepted values are:
971 Sequential trims (Linux block devices only).
977 Random trims (Linux block devices only).
979 Sequential mixed reads and writes.
981 Random mixed reads and writes.
983 Sequential trim+write sequences. Blocks will be trimmed first,
984 then the same blocks will be written to.
986 Fio defaults to read if the option is not specified. For the mixed I/O
987 types, the default is to split them 50/50. For certain types of I/O the
988 result may still be skewed a bit, since the speed may be different.
990 It is possible to specify the number of I/Os to do before getting a new
991 offset by appending ``:<nr>`` to the end of the string given. For a
992 random read, it would look like ``rw=randread:8`` for passing in an offset
993 modifier with a value of 8. If the suffix is used with a sequential I/O
994 pattern, then the *<nr>* value specified will be **added** to the generated
995 offset for each I/O turning sequential I/O into sequential I/O with holes.
996 For instance, using ``rw=write:4k`` will skip 4k for every write. Also see
997 the :option:`rw_sequencer` option.
999 .. option:: rw_sequencer=str
1001 If an offset modifier is given by appending a number to the ``rw=<str>``
1002 line, then this option controls how that number modifies the I/O offset
1003 being generated. Accepted values are:
1006 Generate sequential offset.
1008 Generate the same offset.
1010 ``sequential`` is only useful for random I/O, where fio would normally
1011 generate a new random offset for every I/O. If you append e.g. 8 to randread,
1012 you would get a new random offset for every 8 I/O's. The result would be a
1013 seek for only every 8 I/O's, instead of for every I/O. Use ``rw=randread:8``
1014 to specify that. As sequential I/O is already sequential, setting
1015 ``sequential`` for that would not result in any differences. ``identical``
1016 behaves in a similar fashion, except it sends the same offset 8 number of
1017 times before generating a new offset.
1019 .. option:: unified_rw_reporting=bool
1021 Fio normally reports statistics on a per data direction basis, meaning that
1022 reads, writes, and trims are accounted and reported separately. If this
1023 option is set fio sums the results and report them as "mixed" instead.
1025 .. option:: randrepeat=bool
1027 Seed the random number generator used for random I/O patterns in a
1028 predictable way so the pattern is repeatable across runs. Default: true.
1030 .. option:: allrandrepeat=bool
1032 Seed all random number generators in a predictable way so results are
1033 repeatable across runs. Default: false.
1035 .. option:: randseed=int
1037 Seed the random number generators based on this seed value, to be able to
1038 control what sequence of output is being generated. If not set, the random
1039 sequence depends on the :option:`randrepeat` setting.
1041 .. option:: fallocate=str
1043 Whether pre-allocation is performed when laying down files.
1044 Accepted values are:
1047 Do not pre-allocate space.
1050 Pre-allocate via :manpage:`posix_fallocate(3)`.
1053 Pre-allocate via :manpage:`fallocate(2)` with
1054 FALLOC_FL_KEEP_SIZE set.
1057 Backward-compatible alias for **none**.
1060 Backward-compatible alias for **posix**.
1062 May not be available on all supported platforms. **keep** is only available
1063 on Linux. If using ZFS on Solaris this must be set to **none** because ZFS
1064 doesn't support it. Default: **posix**.
1066 .. option:: fadvise_hint=str
1068 Use :manpage:`posix_fadvise(2)` to advise the kernel on what I/O patterns
1069 are likely to be issued. Accepted values are:
1072 Backwards-compatible hint for "no hint".
1075 Backwards compatible hint for "advise with fio workload type". This
1076 uses **FADV_RANDOM** for a random workload, and **FADV_SEQUENTIAL**
1077 for a sequential workload.
1080 Advise using **FADV_SEQUENTIAL**.
1083 Advise using **FADV_RANDOM**.
1085 .. option:: fadvise_stream=int
1087 Use :manpage:`posix_fadvise(2)` to advise the kernel what stream ID the
1088 writes issued belong to. Only supported on Linux. Note, this option may
1089 change going forward.
1091 .. option:: offset=int
1093 Start I/O at the provided offset in the file, given as either a fixed size in
1094 bytes or a percentage. If a percentage is given, the next ``blockalign``-ed
1095 offset will be used. Data before the given offset will not be touched. This
1096 effectively caps the file size at `real_size - offset`. Can be combined with
1097 :option:`size` to constrain the start and end range of the I/O workload.
1098 A percentage can be specified by a number between 1 and 100 followed by '%',
1099 for example, ``offset=20%`` to specify 20%.
1101 .. option:: offset_increment=int
1103 If this is provided, then the real offset becomes `offset + offset_increment
1104 * thread_number`, where the thread number is a counter that starts at 0 and
1105 is incremented for each sub-job (i.e. when :option:`numjobs` option is
1106 specified). This option is useful if there are several jobs which are
1107 intended to operate on a file in parallel disjoint segments, with even
1108 spacing between the starting points.
1110 .. option:: number_ios=int
1112 Fio will normally perform I/Os until it has exhausted the size of the region
1113 set by :option:`size`, or if it exhaust the allocated time (or hits an error
1114 condition). With this setting, the range/size can be set independently of
1115 the number of I/Os to perform. When fio reaches this number, it will exit
1116 normally and report status. Note that this does not extend the amount of I/O
1117 that will be done, it will only stop fio if this condition is met before
1118 other end-of-job criteria.
1120 .. option:: fsync=int
1122 If writing to a file, issue an :manpage:`fsync(2)` (or its equivalent) of
1123 the dirty data for every number of blocks given. For example, if you give 32
1124 as a parameter, fio will sync the file after every 32 writes issued. If fio is
1125 using non-buffered I/O, we may not sync the file. The exception is the sg
1126 I/O engine, which synchronizes the disk cache anyway. Defaults to 0, which
1127 means fio does not periodically issue and wait for a sync to complete. Also
1128 see :option:`end_fsync` and :option:`fsync_on_close`.
1130 .. option:: fdatasync=int
1132 Like :option:`fsync` but uses :manpage:`fdatasync(2)` to only sync data and
1133 not metadata blocks. In Windows, FreeBSD, and DragonFlyBSD there is no
1134 :manpage:`fdatasync(2)` so this falls back to using :manpage:`fsync(2)`.
1135 Defaults to 0, which means fio does not periodically issue and wait for a
1136 data-only sync to complete.
1138 .. option:: write_barrier=int
1140 Make every `N-th` write a barrier write.
1142 .. option:: sync_file_range=str:val
1144 Use :manpage:`sync_file_range(2)` for every `val` number of write
1145 operations. Fio will track range of writes that have happened since the last
1146 :manpage:`sync_file_range(2)` call. `str` can currently be one or more of:
1149 SYNC_FILE_RANGE_WAIT_BEFORE
1151 SYNC_FILE_RANGE_WRITE
1153 SYNC_FILE_RANGE_WAIT_AFTER
1155 So if you do ``sync_file_range=wait_before,write:8``, fio would use
1156 ``SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE`` for every 8
1157 writes. Also see the :manpage:`sync_file_range(2)` man page. This option is
1160 .. option:: overwrite=bool
1162 If true, writes to a file will always overwrite existing data. If the file
1163 doesn't already exist, it will be created before the write phase begins. If
1164 the file exists and is large enough for the specified write phase, nothing
1165 will be done. Default: false.
1167 .. option:: end_fsync=bool
1169 If true, :manpage:`fsync(2)` file contents when a write stage has completed.
1172 .. option:: fsync_on_close=bool
1174 If true, fio will :manpage:`fsync(2)` a dirty file on close. This differs
1175 from :option:`end_fsync` in that it will happen on every file close, not
1176 just at the end of the job. Default: false.
1178 .. option:: rwmixread=int
1180 Percentage of a mixed workload that should be reads. Default: 50.
1182 .. option:: rwmixwrite=int
1184 Percentage of a mixed workload that should be writes. If both
1185 :option:`rwmixread` and :option:`rwmixwrite` is given and the values do not
1186 add up to 100%, the latter of the two will be used to override the
1187 first. This may interfere with a given rate setting, if fio is asked to
1188 limit reads or writes to a certain rate. If that is the case, then the
1189 distribution may be skewed. Default: 50.
1191 .. option:: random_distribution=str:float[,str:float][,str:float]
1193 By default, fio will use a completely uniform random distribution when asked
1194 to perform random I/O. Sometimes it is useful to skew the distribution in
1195 specific ways, ensuring that some parts of the data is more hot than others.
1196 fio includes the following distribution models:
1199 Uniform random distribution
1208 Normal (Gaussian) distribution
1211 Zoned random distribution
1213 When using a **zipf** or **pareto** distribution, an input value is also
1214 needed to define the access pattern. For **zipf**, this is the `zipf
1215 theta`. For **pareto**, it's the `Pareto power`. Fio includes a test
1216 program, :command:`genzipf`, that can be used visualize what the given input
1217 values will yield in terms of hit rates. If you wanted to use **zipf** with
1218 a `theta` of 1.2, you would use ``random_distribution=zipf:1.2`` as the
1219 option. If a non-uniform model is used, fio will disable use of the random
1220 map. For the **gauss** distribution, a normal deviation is supplied as a
1221 value between 0 and 100.
1223 For a **zoned** distribution, fio supports specifying percentages of I/O
1224 access that should fall within what range of the file or device. For
1225 example, given a criteria of:
1227 * 60% of accesses should be to the first 10%
1228 * 30% of accesses should be to the next 20%
1229 * 8% of accesses should be to to the next 30%
1230 * 2% of accesses should be to the next 40%
1232 we can define that through zoning of the random accesses. For the above
1233 example, the user would do::
1235 random_distribution=zoned:60/10:30/20:8/30:2/40
1237 similarly to how :option:`bssplit` works for setting ranges and percentages
1238 of block sizes. Like :option:`bssplit`, it's possible to specify separate
1239 zones for reads, writes, and trims. If just one set is given, it'll apply to
1242 .. option:: percentage_random=int[,int][,int]
1244 For a random workload, set how big a percentage should be random. This
1245 defaults to 100%, in which case the workload is fully random. It can be set
1246 from anywhere from 0 to 100. Setting it to 0 would make the workload fully
1247 sequential. Any setting in between will result in a random mix of sequential
1248 and random I/O, at the given percentages. Comma-separated values may be
1249 specified for reads, writes, and trims as described in :option:`blocksize`.
1251 .. option:: norandommap
1253 Normally fio will cover every block of the file when doing random I/O. If
1254 this option is given, fio will just get a new random offset without looking
1255 at past I/O history. This means that some blocks may not be read or written,
1256 and that some blocks may be read/written more than once. If this option is
1257 used with :option:`verify` and multiple blocksizes (via :option:`bsrange`),
1258 only intact blocks are verified, i.e., partially-overwritten blocks are
1261 .. option:: softrandommap=bool
1263 See :option:`norandommap`. If fio runs with the random block map enabled and
1264 it fails to allocate the map, if this option is set it will continue without
1265 a random block map. As coverage will not be as complete as with random maps,
1266 this option is disabled by default.
1268 .. option:: random_generator=str
1270 Fio supports the following engines for generating
1271 I/O offsets for random I/O:
1274 Strong 2^88 cycle random number generator
1276 Linear feedback shift register generator
1278 Strong 64-bit 2^258 cycle random number generator
1280 **tausworthe** is a strong random number generator, but it requires tracking
1281 on the side if we want to ensure that blocks are only read or written
1282 once. **LFSR** guarantees that we never generate the same offset twice, and
1283 it's also less computationally expensive. It's not a true random generator,
1284 however, though for I/O purposes it's typically good enough. **LFSR** only
1285 works with single block sizes, not with workloads that use multiple block
1286 sizes. If used with such a workload, fio may read or write some blocks
1287 multiple times. The default value is **tausworthe**, unless the required
1288 space exceeds 2^32 blocks. If it does, then **tausworthe64** is
1289 selected automatically.
1295 .. option:: blocksize=int[,int][,int], bs=int[,int][,int]
1297 The block size in bytes used for I/O units. Default: 4096. A single value
1298 applies to reads, writes, and trims. Comma-separated values may be
1299 specified for reads, writes, and trims. A value not terminated in a comma
1300 applies to subsequent types.
1305 means 256k for reads, writes and trims.
1308 means 8k for reads, 32k for writes and trims.
1311 means 8k for reads, 32k for writes, and default for trims.
1314 means default for reads, 8k for writes and trims.
1317 means default for reads, 8k for writes, and default for trims.
1319 .. option:: blocksize_range=irange[,irange][,irange], bsrange=irange[,irange][,irange]
1321 A range of block sizes in bytes for I/O units. The issued I/O unit will
1322 always be a multiple of the minimum size, unless
1323 :option:`blocksize_unaligned` is set.
1325 Comma-separated ranges may be specified for reads, writes, and trims as
1326 described in :option:`blocksize`.
1328 Example: ``bsrange=1k-4k,2k-8k``.
1330 .. option:: bssplit=str[,str][,str]
1332 Sometimes you want even finer grained control of the block sizes issued, not
1333 just an even split between them. This option allows you to weight various
1334 block sizes, so that you are able to define a specific amount of block sizes
1335 issued. The format for this option is::
1337 bssplit=blocksize/percentage:blocksize/percentage
1339 for as many block sizes as needed. So if you want to define a workload that
1340 has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would write::
1342 bssplit=4k/10:64k/50:32k/40
1344 Ordering does not matter. If the percentage is left blank, fio will fill in
1345 the remaining values evenly. So a bssplit option like this one::
1347 bssplit=4k/50:1k/:32k/
1349 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always add up
1350 to 100, if bssplit is given a range that adds up to more, it will error out.
1352 Comma-separated values may be specified for reads, writes, and trims as
1353 described in :option:`blocksize`.
1355 If you want a workload that has 50% 2k reads and 50% 4k reads, while having
1356 90% 4k writes and 10% 8k writes, you would specify::
1358 bssplit=2k/50:4k/50,4k/90,8k/10
1360 .. option:: blocksize_unaligned, bs_unaligned
1362 If set, fio will issue I/O units with any size within
1363 :option:`blocksize_range`, not just multiples of the minimum size. This
1364 typically won't work with direct I/O, as that normally requires sector
1367 .. option:: bs_is_seq_rand
1369 If this option is set, fio will use the normal read,write blocksize settings
1370 as sequential,random blocksize settings instead. Any random read or write
1371 will use the WRITE blocksize settings, and any sequential read or write will
1372 use the READ blocksize settings.
1374 .. option:: blockalign=int[,int][,int], ba=int[,int][,int]
1376 Boundary to which fio will align random I/O units. Default:
1377 :option:`blocksize`. Minimum alignment is typically 512b for using direct
1378 I/O, though it usually depends on the hardware block size. This option is
1379 mutually exclusive with using a random map for files, so it will turn off
1380 that option. Comma-separated values may be specified for reads, writes, and
1381 trims as described in :option:`blocksize`.
1387 .. option:: zero_buffers
1389 Initialize buffers with all zeros. Default: fill buffers with random data.
1391 .. option:: refill_buffers
1393 If this option is given, fio will refill the I/O buffers on every
1394 submit. The default is to only fill it at init time and reuse that
1395 data. Only makes sense if zero_buffers isn't specified, naturally. If data
1396 verification is enabled, `refill_buffers` is also automatically enabled.
1398 .. option:: scramble_buffers=bool
1400 If :option:`refill_buffers` is too costly and the target is using data
1401 deduplication, then setting this option will slightly modify the I/O buffer
1402 contents to defeat normal de-dupe attempts. This is not enough to defeat
1403 more clever block compression attempts, but it will stop naive dedupe of
1404 blocks. Default: true.
1406 .. option:: buffer_compress_percentage=int
1408 If this is set, then fio will attempt to provide I/O buffer content (on
1409 WRITEs) that compresses to the specified level. Fio does this by providing a
1410 mix of random data and a fixed pattern. The fixed pattern is either zeros,
1411 or the pattern specified by :option:`buffer_pattern`. If the pattern option
1412 is used, it might skew the compression ratio slightly. Note that this is per
1413 block size unit, for file/disk wide compression level that matches this
1414 setting, you'll also want to set :option:`refill_buffers`.
1416 .. option:: buffer_compress_chunk=int
1418 See :option:`buffer_compress_percentage`. This setting allows fio to manage
1419 how big the ranges of random data and zeroed data is. Without this set, fio
1420 will provide :option:`buffer_compress_percentage` of blocksize random data,
1421 followed by the remaining zeroed. With this set to some chunk size smaller
1422 than the block size, fio can alternate random and zeroed data throughout the
1425 .. option:: buffer_pattern=str
1427 If set, fio will fill the I/O buffers with this pattern or with the contents
1428 of a file. If not set, the contents of I/O buffers are defined by the other
1429 options related to buffer contents. The setting can be any pattern of bytes,
1430 and can be prefixed with 0x for hex values. It may also be a string, where
1431 the string must then be wrapped with ``""``. Or it may also be a filename,
1432 where the filename must be wrapped with ``''`` in which case the file is
1433 opened and read. Note that not all the file contents will be read if that
1434 would cause the buffers to overflow. So, for example::
1436 buffer_pattern='filename'
1440 buffer_pattern="abcd"
1448 buffer_pattern=0xdeadface
1450 Also you can combine everything together in any order::
1452 buffer_pattern=0xdeadface"abcd"-12'filename'
1454 .. option:: dedupe_percentage=int
1456 If set, fio will generate this percentage of identical buffers when
1457 writing. These buffers will be naturally dedupable. The contents of the
1458 buffers depend on what other buffer compression settings have been set. It's
1459 possible to have the individual buffers either fully compressible, or not at
1460 all. This option only controls the distribution of unique buffers.
1462 .. option:: invalidate=bool
1464 Invalidate the buffer/page cache parts of the files to be used prior to
1465 starting I/O if the platform and file type support it. Defaults to true.
1466 This will be ignored if :option:`pre_read` is also specified for the
1469 .. option:: sync=bool
1471 Use synchronous I/O for buffered writes. For the majority of I/O engines,
1472 this means using O_SYNC. Default: false.
1474 .. option:: iomem=str, mem=str
1476 Fio can use various types of memory as the I/O unit buffer. The allowed
1480 Use memory from :manpage:`malloc(3)` as the buffers. Default memory
1484 Use shared memory as the buffers. Allocated through
1485 :manpage:`shmget(2)`.
1488 Same as shm, but use huge pages as backing.
1491 Use :manpage:`mmap(2)` to allocate buffers. May either be anonymous memory, or can
1492 be file backed if a filename is given after the option. The format
1493 is `mem=mmap:/path/to/file`.
1496 Use a memory mapped huge file as the buffer backing. Append filename
1497 after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file`.
1500 Same as mmap, but use a MMAP_SHARED mapping.
1503 Use GPU memory as the buffers for GPUDirect RDMA benchmark.
1505 The area allocated is a function of the maximum allowed bs size for the job,
1506 multiplied by the I/O depth given. Note that for **shmhuge** and
1507 **mmaphuge** to work, the system must have free huge pages allocated. This
1508 can normally be checked and set by reading/writing
1509 :file:`/proc/sys/vm/nr_hugepages` on a Linux system. Fio assumes a huge page
1510 is 4MiB in size. So to calculate the number of huge pages you need for a
1511 given job file, add up the I/O depth of all jobs (normally one unless
1512 :option:`iodepth` is used) and multiply by the maximum bs set. Then divide
1513 that number by the huge page size. You can see the size of the huge pages in
1514 :file:`/proc/meminfo`. If no huge pages are allocated by having a non-zero
1515 number in `nr_hugepages`, using **mmaphuge** or **shmhuge** will fail. Also
1516 see :option:`hugepage-size`.
1518 **mmaphuge** also needs to have hugetlbfs mounted and the file location
1519 should point there. So if it's mounted in :file:`/huge`, you would use
1520 `mem=mmaphuge:/huge/somefile`.
1522 .. option:: iomem_align=int
1524 This indicates the memory alignment of the I/O memory buffers. Note that
1525 the given alignment is applied to the first I/O unit buffer, if using
1526 :option:`iodepth` the alignment of the following buffers are given by the
1527 :option:`bs` used. In other words, if using a :option:`bs` that is a
1528 multiple of the page sized in the system, all buffers will be aligned to
1529 this value. If using a :option:`bs` that is not page aligned, the alignment
1530 of subsequent I/O memory buffers is the sum of the :option:`iomem_align` and
1533 .. option:: hugepage-size=int
1535 Defines the size of a huge page. Must at least be equal to the system
1536 setting, see :file:`/proc/meminfo`. Defaults to 4MiB. Should probably
1537 always be a multiple of megabytes, so using ``hugepage-size=Xm`` is the
1538 preferred way to set this to avoid setting a non-pow-2 bad value.
1540 .. option:: lockmem=int
1542 Pin the specified amount of memory with :manpage:`mlock(2)`. Can be used to
1543 simulate a smaller amount of memory. The amount specified is per worker.
1549 .. option:: size=int
1551 The total size of file I/O for each thread of this job. Fio will run until
1552 this many bytes has been transferred, unless runtime is limited by other options
1553 (such as :option:`runtime`, for instance, or increased/decreased by :option:`io_size`).
1554 Fio will divide this size between the available files determined by options
1555 such as :option:`nrfiles`, :option:`filename`, unless :option:`filesize` is
1556 specified by the job. If the result of division happens to be 0, the size is
1557 set to the physical size of the given files or devices if they exist.
1558 If this option is not specified, fio will use the full size of the given
1559 files or devices. If the files do not exist, size must be given. It is also
1560 possible to give size as a percentage between 1 and 100. If ``size=20%`` is
1561 given, fio will use 20% of the full size of the given files or devices.
1562 Can be combined with :option:`offset` to constrain the start and end range
1563 that I/O will be done within.
1565 .. option:: io_size=int, io_limit=int
1567 Normally fio operates within the region set by :option:`size`, which means
1568 that the :option:`size` option sets both the region and size of I/O to be
1569 performed. Sometimes that is not what you want. With this option, it is
1570 possible to define just the amount of I/O that fio should do. For instance,
1571 if :option:`size` is set to 20GiB and :option:`io_size` is set to 5GiB, fio
1572 will perform I/O within the first 20GiB but exit when 5GiB have been
1573 done. The opposite is also possible -- if :option:`size` is set to 20GiB,
1574 and :option:`io_size` is set to 40GiB, then fio will do 40GiB of I/O within
1575 the 0..20GiB region.
1577 .. option:: filesize=irange(int)
1579 Individual file sizes. May be a range, in which case fio will select sizes
1580 for files at random within the given range and limited to :option:`size` in
1581 total (if that is given). If not given, each created file is the same size.
1582 This option overrides :option:`size` in terms of file size, which means
1583 this value is used as a fixed size or possible range of each file.
1585 .. option:: file_append=bool
1587 Perform I/O after the end of the file. Normally fio will operate within the
1588 size of a file. If this option is set, then fio will append to the file
1589 instead. This has identical behavior to setting :option:`offset` to the size
1590 of a file. This option is ignored on non-regular files.
1592 .. option:: fill_device=bool, fill_fs=bool
1594 Sets size to something really large and waits for ENOSPC (no space left on
1595 device) as the terminating condition. Only makes sense with sequential
1596 write. For a read workload, the mount point will be filled first then I/O
1597 started on the result. This option doesn't make sense if operating on a raw
1598 device node, since the size of that is already known by the file system.
1599 Additionally, writing beyond end-of-device will not return ENOSPC there.
1605 .. option:: ioengine=str
1607 Defines how the job issues I/O to the file. The following types are defined:
1610 Basic :manpage:`read(2)` or :manpage:`write(2)`
1611 I/O. :manpage:`lseek(2)` is used to position the I/O location.
1612 See :option:`fsync` and :option:`fdatasync` for syncing write I/Os.
1615 Basic :manpage:`pread(2)` or :manpage:`pwrite(2)` I/O. Default on
1616 all supported operating systems except for Windows.
1619 Basic :manpage:`readv(2)` or :manpage:`writev(2)` I/O. Will emulate
1620 queuing by coalescing adjacent I/Os into a single submission.
1623 Basic :manpage:`preadv(2)` or :manpage:`pwritev(2)` I/O.
1626 Basic :manpage:`preadv2(2)` or :manpage:`pwritev2(2)` I/O.
1629 Linux native asynchronous I/O. Note that Linux may only support
1630 queued behavior with non-buffered I/O (set ``direct=1`` or
1632 This engine defines engine specific options.
1635 POSIX asynchronous I/O using :manpage:`aio_read(3)` and
1636 :manpage:`aio_write(3)`.
1639 Solaris native asynchronous I/O.
1642 Windows native asynchronous I/O. Default on Windows.
1645 File is memory mapped with :manpage:`mmap(2)` and data copied
1646 to/from using :manpage:`memcpy(3)`.
1649 :manpage:`splice(2)` is used to transfer the data and
1650 :manpage:`vmsplice(2)` to transfer data from user space to the
1654 SCSI generic sg v3 I/O. May either be synchronous using the SG_IO
1655 ioctl, or if the target is an sg character device we use
1656 :manpage:`read(2)` and :manpage:`write(2)` for asynchronous
1657 I/O. Requires filename option to specify either block or character
1661 Doesn't transfer any data, just pretends to. This is mainly used to
1662 exercise fio itself and for debugging/testing purposes.
1665 Transfer over the network to given ``host:port``. Depending on the
1666 :option:`protocol` used, the :option:`hostname`, :option:`port`,
1667 :option:`listen` and :option:`filename` options are used to specify
1668 what sort of connection to make, while the :option:`protocol` option
1669 determines which protocol will be used. This engine defines engine
1673 Like **net**, but uses :manpage:`splice(2)` and
1674 :manpage:`vmsplice(2)` to map data and send/receive.
1675 This engine defines engine specific options.
1678 Doesn't transfer any data, but burns CPU cycles according to the
1679 :option:`cpuload` and :option:`cpuchunks` options. Setting
1680 :option:`cpuload`\=85 will cause that job to do nothing but burn 85%
1681 of the CPU. In case of SMP machines, use :option:`numjobs`
1682 =<no_of_cpu> to get desired CPU usage, as the cpuload only loads a
1683 single CPU at the desired rate. A job never finishes unless there is
1684 at least one non-cpuio job.
1687 The GUASI I/O engine is the Generic Userspace Asyncronous Syscall
1688 Interface approach to async I/O. See
1690 http://www.xmailserver.org/guasi-lib.html
1692 for more info on GUASI.
1695 The RDMA I/O engine supports both RDMA memory semantics
1696 (RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the
1697 InfiniBand, RoCE and iWARP protocols.
1700 I/O engine that does regular fallocate to simulate data transfer as
1704 does fallocate(,mode = FALLOC_FL_KEEP_SIZE,).
1707 does fallocate(,mode = 0).
1710 does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE).
1713 I/O engine that sends :manpage:`ftruncate(2)` operations in response
1714 to write (DDIR_WRITE) events. Each ftruncate issued sets the file's
1715 size to the current block offset. Block size is ignored.
1718 I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate
1719 defragment activity in request to DDIR_WRITE event.
1722 I/O engine supporting direct access to Ceph Rados Block Devices
1723 (RBD) via librbd without the need to use the kernel rbd driver. This
1724 ioengine defines engine specific options.
1727 Using GlusterFS libgfapi sync interface to direct access to
1728 GlusterFS volumes without having to go through FUSE. This ioengine
1729 defines engine specific options.
1732 Using GlusterFS libgfapi async interface to direct access to
1733 GlusterFS volumes without having to go through FUSE. This ioengine
1734 defines engine specific options.
1737 Read and write through Hadoop (HDFS). The :file:`filename` option
1738 is used to specify host,port of the hdfs name-node to connect. This
1739 engine interprets offsets a little differently. In HDFS, files once
1740 created cannot be modified so random writes are not possible. To
1741 imitate this the libhdfs engine expects a bunch of small files to be
1742 created over HDFS and will randomly pick a file from them
1743 based on the offset generated by fio backend (see the example
1744 job file to create such files, use ``rw=write`` option). Please
1745 note, it may be necessary to set environment variables to work
1746 with HDFS/libhdfs properly. Each job uses its own connection to
1750 Read, write and erase an MTD character device (e.g.,
1751 :file:`/dev/mtd0`). Discards are treated as erases. Depending on the
1752 underlying device type, the I/O may have to go in a certain pattern,
1753 e.g., on NAND, writing sequentially to erase blocks and discarding
1754 before overwriting. The `trimwrite` mode works well for this
1758 Read and write using filesystem DAX to a file on a filesystem
1759 mounted with DAX on a persistent memory device through the NVML
1763 Read and write using device DAX to a persistent memory device (e.g.,
1764 /dev/dax0.0) through the NVML libpmem library.
1767 Prefix to specify loading an external I/O engine object file. Append
1768 the engine filename, e.g. ``ioengine=external:/tmp/foo.o`` to load
1769 ioengine :file:`foo.o` in :file:`/tmp`.
1772 I/O engine specific parameters
1773 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1775 In addition, there are some parameters which are only valid when a specific
1776 ioengine is in use. These are used identically to normal parameters, with the
1777 caveat that when used on the command line, they must come after the
1778 :option:`ioengine` that defines them is selected.
1780 .. option:: userspace_reap : [libaio]
1782 Normally, with the libaio engine in use, fio will use the
1783 :manpage:`io_getevents(2)` system call to reap newly returned events. With
1784 this flag turned on, the AIO ring will be read directly from user-space to
1785 reap events. The reaping mode is only enabled when polling for a minimum of
1786 0 events (e.g. when :option:`iodepth_batch_complete` `=0`).
1788 .. option:: hipri : [pvsync2]
1790 Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority
1793 .. option:: cpuload=int : [cpuio]
1795 Attempt to use the specified percentage of CPU cycles. This is a mandatory
1796 option when using cpuio I/O engine.
1798 .. option:: cpuchunks=int : [cpuio]
1800 Split the load into cycles of the given time. In microseconds.
1802 .. option:: exit_on_io_done=bool : [cpuio]
1804 Detect when I/O threads are done, then exit.
1806 .. option:: hostname=str : [netsplice] [net]
1808 The hostname or IP address to use for TCP or UDP based I/O. If the job is
1809 a TCP listener or UDP reader, the hostname is not used and must be omitted
1810 unless it is a valid UDP multicast address.
1812 .. option:: namenode=str : [libhdfs]
1814 The hostname or IP address of a HDFS cluster namenode to contact.
1816 .. option:: port=int
1820 The TCP or UDP port to bind to or connect to. If this is used with
1821 :option:`numjobs` to spawn multiple instances of the same job type, then
1822 this will be the starting port number since fio will use a range of
1827 the listening port of the HFDS cluster namenode.
1829 .. option:: interface=str : [netsplice] [net]
1831 The IP address of the network interface used to send or receive UDP
1834 .. option:: ttl=int : [netsplice] [net]
1836 Time-to-live value for outgoing UDP multicast packets. Default: 1.
1838 .. option:: nodelay=bool : [netsplice] [net]
1840 Set TCP_NODELAY on TCP connections.
1842 .. option:: protocol=str : [netsplice] [net]
1844 .. option:: proto=str : [netsplice] [net]
1846 The network protocol to use. Accepted values are:
1849 Transmission control protocol.
1851 Transmission control protocol V6.
1853 User datagram protocol.
1855 User datagram protocol V6.
1859 When the protocol is TCP or UDP, the port must also be given, as well as the
1860 hostname if the job is a TCP listener or UDP reader. For unix sockets, the
1861 normal filename option should be used and the port is invalid.
1863 .. option:: listen : [net]
1865 For TCP network connections, tell fio to listen for incoming connections
1866 rather than initiating an outgoing connection. The :option:`hostname` must
1867 be omitted if this option is used.
1869 .. option:: pingpong : [net]
1871 Normally a network writer will just continue writing data, and a network
1872 reader will just consume packages. If ``pingpong=1`` is set, a writer will
1873 send its normal payload to the reader, then wait for the reader to send the
1874 same payload back. This allows fio to measure network latencies. The
1875 submission and completion latencies then measure local time spent sending or
1876 receiving, and the completion latency measures how long it took for the
1877 other end to receive and send back. For UDP multicast traffic
1878 ``pingpong=1`` should only be set for a single reader when multiple readers
1879 are listening to the same address.
1881 .. option:: window_size : [net]
1883 Set the desired socket buffer size for the connection.
1885 .. option:: mss : [net]
1887 Set the TCP maximum segment size (TCP_MAXSEG).
1889 .. option:: donorname=str : [e4defrag]
1891 File will be used as a block donor (swap extents between files).
1893 .. option:: inplace=int : [e4defrag]
1895 Configure donor file blocks allocation strategy:
1898 Default. Preallocate donor's file on init.
1900 Allocate space immediately inside defragment event, and free right
1903 .. option:: clustername=str : [rbd]
1905 Specifies the name of the Ceph cluster.
1907 .. option:: rbdname=str : [rbd]
1909 Specifies the name of the RBD.
1911 .. option:: pool=str : [rbd]
1913 Specifies the name of the Ceph pool containing RBD.
1915 .. option:: clientname=str : [rbd]
1917 Specifies the username (without the 'client.' prefix) used to access the
1918 Ceph cluster. If the *clustername* is specified, the *clientname* shall be
1919 the full *type.id* string. If no type. prefix is given, fio will add
1920 'client.' by default.
1922 .. option:: skip_bad=bool : [mtd]
1924 Skip operations against known bad blocks.
1926 .. option:: hdfsdirectory : [libhdfs]
1928 libhdfs will create chunk in this HDFS directory.
1930 .. option:: chunk_size : [libhdfs]
1932 the size of the chunk to use for each file.
1938 .. option:: iodepth=int
1940 Number of I/O units to keep in flight against the file. Note that
1941 increasing *iodepth* beyond 1 will not affect synchronous ioengines (except
1942 for small degrees when :option:`verify_async` is in use). Even async
1943 engines may impose OS restrictions causing the desired depth not to be
1944 achieved. This may happen on Linux when using libaio and not setting
1945 :option:`direct`\=1, since buffered I/O is not async on that OS. Keep an
1946 eye on the I/O depth distribution in the fio output to verify that the
1947 achieved depth is as expected. Default: 1.
1949 .. option:: iodepth_batch_submit=int, iodepth_batch=int
1951 This defines how many pieces of I/O to submit at once. It defaults to 1
1952 which means that we submit each I/O as soon as it is available, but can be
1953 raised to submit bigger batches of I/O at the time. If it is set to 0 the
1954 :option:`iodepth` value will be used.
1956 .. option:: iodepth_batch_complete_min=int, iodepth_batch_complete=int
1958 This defines how many pieces of I/O to retrieve at once. It defaults to 1
1959 which means that we'll ask for a minimum of 1 I/O in the retrieval process
1960 from the kernel. The I/O retrieval will go on until we hit the limit set by
1961 :option:`iodepth_low`. If this variable is set to 0, then fio will always
1962 check for completed events before queuing more I/O. This helps reduce I/O
1963 latency, at the cost of more retrieval system calls.
1965 .. option:: iodepth_batch_complete_max=int
1967 This defines maximum pieces of I/O to retrieve at once. This variable should
1968 be used along with :option:`iodepth_batch_complete_min`\=int variable,
1969 specifying the range of min and max amount of I/O which should be
1970 retrieved. By default it is equal to the :option:`iodepth_batch_complete_min`
1975 iodepth_batch_complete_min=1
1976 iodepth_batch_complete_max=<iodepth>
1978 which means that we will retrieve at least 1 I/O and up to the whole
1979 submitted queue depth. If none of I/O has been completed yet, we will wait.
1983 iodepth_batch_complete_min=0
1984 iodepth_batch_complete_max=<iodepth>
1986 which means that we can retrieve up to the whole submitted queue depth, but
1987 if none of I/O has been completed yet, we will NOT wait and immediately exit
1988 the system call. In this example we simply do polling.
1990 .. option:: iodepth_low=int
1992 The low water mark indicating when to start filling the queue
1993 again. Defaults to the same as :option:`iodepth`, meaning that fio will
1994 attempt to keep the queue full at all times. If :option:`iodepth` is set to
1995 e.g. 16 and *iodepth_low* is set to 4, then after fio has filled the queue of
1996 16 requests, it will let the depth drain down to 4 before starting to fill
1999 .. option:: io_submit_mode=str
2001 This option controls how fio submits the I/O to the I/O engine. The default
2002 is `inline`, which means that the fio job threads submit and reap I/O
2003 directly. If set to `offload`, the job threads will offload I/O submission
2004 to a dedicated pool of I/O threads. This requires some coordination and thus
2005 has a bit of extra overhead, especially for lower queue depth I/O where it
2006 can increase latencies. The benefit is that fio can manage submission rates
2007 independently of the device completion rates. This avoids skewed latency
2008 reporting if I/O gets backed up on the device side (the coordinated omission
2015 .. option:: thinktime=time
2017 Stall the job for the specified period of time after an I/O has completed before issuing the
2018 next. May be used to simulate processing being done by an application.
2019 When the unit is omitted, the value is interpreted in microseconds. See
2020 :option:`thinktime_blocks` and :option:`thinktime_spin`.
2022 .. option:: thinktime_spin=time
2024 Only valid if :option:`thinktime` is set - pretend to spend CPU time doing
2025 something with the data received, before falling back to sleeping for the
2026 rest of the period specified by :option:`thinktime`. When the unit is
2027 omitted, the value is interpreted in microseconds.
2029 .. option:: thinktime_blocks=int
2031 Only valid if :option:`thinktime` is set - control how many blocks to issue,
2032 before waiting `thinktime` usecs. If not set, defaults to 1 which will make
2033 fio wait `thinktime` usecs after every block. This effectively makes any
2034 queue depth setting redundant, since no more than 1 I/O will be queued
2035 before we have to complete it and do our thinktime. In other words, this
2036 setting effectively caps the queue depth if the latter is larger.
2038 .. option:: rate=int[,int][,int]
2040 Cap the bandwidth used by this job. The number is in bytes/sec, the normal
2041 suffix rules apply. Comma-separated values may be specified for reads,
2042 writes, and trims as described in :option:`blocksize`.
2044 For example, using `rate=1m,500k` would limit reads to 1MiB/sec and writes to
2045 500KiB/sec. Capping only reads or writes can be done with `rate=,500k` or
2046 `rate=500k,` where the former will only limit writes (to 500KiB/sec) and the
2047 latter will only limit reads.
2049 .. option:: rate_min=int[,int][,int]
2051 Tell fio to do whatever it can to maintain at least this bandwidth. Failing
2052 to meet this requirement will cause the job to exit. Comma-separated values
2053 may be specified for reads, writes, and trims as described in
2054 :option:`blocksize`.
2056 .. option:: rate_iops=int[,int][,int]
2058 Cap the bandwidth to this number of IOPS. Basically the same as
2059 :option:`rate`, just specified independently of bandwidth. If the job is
2060 given a block size range instead of a fixed value, the smallest block size
2061 is used as the metric. Comma-separated values may be specified for reads,
2062 writes, and trims as described in :option:`blocksize`.
2064 .. option:: rate_iops_min=int[,int][,int]
2066 If fio doesn't meet this rate of I/O, it will cause the job to exit.
2067 Comma-separated values may be specified for reads, writes, and trims as
2068 described in :option:`blocksize`.
2070 .. option:: rate_process=str
2072 This option controls how fio manages rated I/O submissions. The default is
2073 `linear`, which submits I/O in a linear fashion with fixed delays between
2074 I/Os that gets adjusted based on I/O completion rates. If this is set to
2075 `poisson`, fio will submit I/O based on a more real world random request
2076 flow, known as the Poisson process
2077 (https://en.wikipedia.org/wiki/Poisson_point_process). The lambda will be
2078 10^6 / IOPS for the given workload.
2084 .. option:: latency_target=time
2086 If set, fio will attempt to find the max performance point that the given
2087 workload will run at while maintaining a latency below this target. When
2088 the unit is omitted, the value is interpreted in microseconds. See
2089 :option:`latency_window` and :option:`latency_percentile`.
2091 .. option:: latency_window=time
2093 Used with :option:`latency_target` to specify the sample window that the job
2094 is run at varying queue depths to test the performance. When the unit is
2095 omitted, the value is interpreted in microseconds.
2097 .. option:: latency_percentile=float
2099 The percentage of I/Os that must fall within the criteria specified by
2100 :option:`latency_target` and :option:`latency_window`. If not set, this
2101 defaults to 100.0, meaning that all I/Os must be equal or below to the value
2102 set by :option:`latency_target`.
2104 .. option:: max_latency=time
2106 If set, fio will exit the job with an ETIMEDOUT error if it exceeds this
2107 maximum latency. When the unit is omitted, the value is interpreted in
2110 .. option:: rate_cycle=int
2112 Average bandwidth for :option:`rate` and :option:`rate_min` over this number
2113 of milliseconds. Defaults to 1000.
2119 .. option:: write_iolog=str
2121 Write the issued I/O patterns to the specified file. See
2122 :option:`read_iolog`. Specify a separate file for each job, otherwise the
2123 iologs will be interspersed and the file may be corrupt.
2125 .. option:: read_iolog=str
2127 Open an iolog with the specified filename and replay the I/O patterns it
2128 contains. This can be used to store a workload and replay it sometime
2129 later. The iolog given may also be a blktrace binary file, which allows fio
2130 to replay a workload captured by :command:`blktrace`. See
2131 :manpage:`blktrace(8)` for how to capture such logging data. For blktrace
2132 replay, the file needs to be turned into a blkparse binary data file first
2133 (``blkparse <device> -o /dev/null -d file_for_fio.bin``).
2135 .. option:: replay_no_stall=int
2137 When replaying I/O with :option:`read_iolog` the default behavior is to
2138 attempt to respect the timestamps within the log and replay them with the
2139 appropriate delay between IOPS. By setting this variable fio will not
2140 respect the timestamps and attempt to replay them as fast as possible while
2141 still respecting ordering. The result is the same I/O pattern to a given
2142 device, but different timings.
2144 .. option:: replay_redirect=str
2146 While replaying I/O patterns using :option:`read_iolog` the default behavior
2147 is to replay the IOPS onto the major/minor device that each IOP was recorded
2148 from. This is sometimes undesirable because on a different machine those
2149 major/minor numbers can map to a different device. Changing hardware on the
2150 same system can also result in a different major/minor mapping.
2151 ``replay_redirect`` causes all I/Os to be replayed onto the single specified
2152 device regardless of the device it was recorded
2153 from. i.e. :option:`replay_redirect`\= :file:`/dev/sdc` would cause all I/O
2154 in the blktrace or iolog to be replayed onto :file:`/dev/sdc`. This means
2155 multiple devices will be replayed onto a single device, if the trace
2156 contains multiple devices. If you want multiple devices to be replayed
2157 concurrently to multiple redirected devices you must blkparse your trace
2158 into separate traces and replay them with independent fio invocations.
2159 Unfortunately this also breaks the strict time ordering between multiple
2162 .. option:: replay_align=int
2164 Force alignment of I/O offsets and lengths in a trace to this power of 2
2167 .. option:: replay_scale=int
2169 Scale sector offsets down by this factor when replaying traces.
2172 Threads, processes and job synchronization
2173 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2177 Fio defaults to creating jobs by using fork, however if this option is
2178 given, fio will create jobs by using POSIX Threads' function
2179 :manpage:`pthread_create(3)` to create threads instead.
2181 .. option:: wait_for=str
2183 If set, the current job won't be started until all workers of the specified
2184 waitee job are done.
2186 ``wait_for`` operates on the job name basis, so there are a few
2187 limitations. First, the waitee must be defined prior to the waiter job
2188 (meaning no forward references). Second, if a job is being referenced as a
2189 waitee, it must have a unique name (no duplicate waitees).
2191 .. option:: nice=int
2193 Run the job with the given nice value. See man :manpage:`nice(2)`.
2195 On Windows, values less than -15 set the process class to "High"; -1 through
2196 -15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle"
2199 .. option:: prio=int
2201 Set the I/O priority value of this job. Linux limits us to a positive value
2202 between 0 and 7, with 0 being the highest. See man
2203 :manpage:`ionice(1)`. Refer to an appropriate manpage for other operating
2204 systems since meaning of priority may differ.
2206 .. option:: prioclass=int
2208 Set the I/O priority class. See man :manpage:`ionice(1)`.
2210 .. option:: cpumask=int
2212 Set the CPU affinity of this job. The parameter given is a bit mask of
2213 allowed CPUs the job may run on. So if you want the allowed CPUs to be 1
2214 and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man
2215 :manpage:`sched_setaffinity(2)`. This may not work on all supported
2216 operating systems or kernel versions. This option doesn't work well for a
2217 higher CPU count than what you can store in an integer mask, so it can only
2218 control cpus 1-32. For boxes with larger CPU counts, use
2219 :option:`cpus_allowed`.
2221 .. option:: cpus_allowed=str
2223 Controls the same options as :option:`cpumask`, but accepts a textual
2224 specification of the permitted CPUs instead. So to use CPUs 1 and 5 you
2225 would specify ``cpus_allowed=1,5``. This option also allows a range of CPUs
2226 to be specified -- say you wanted a binding to CPUs 1, 5, and 8 to 15, you
2227 would set ``cpus_allowed=1,5,8-15``.
2229 .. option:: cpus_allowed_policy=str
2231 Set the policy of how fio distributes the CPUs specified by
2232 :option:`cpus_allowed` or :option:`cpumask`. Two policies are supported:
2235 All jobs will share the CPU set specified.
2237 Each job will get a unique CPU from the CPU set.
2239 **shared** is the default behavior, if the option isn't specified. If
2240 **split** is specified, then fio will will assign one cpu per job. If not
2241 enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs
2244 .. option:: numa_cpu_nodes=str
2246 Set this job running on specified NUMA nodes' CPUs. The arguments allow
2247 comma delimited list of cpu numbers, A-B ranges, or `all`. Note, to enable
2248 NUMA options support, fio must be built on a system with libnuma-dev(el)
2251 .. option:: numa_mem_policy=str
2253 Set this job's memory policy and corresponding NUMA nodes. Format of the
2258 ``mode`` is one of the following memory poicies: ``default``, ``prefer``,
2259 ``bind``, ``interleave`` or ``local``. For ``default`` and ``local`` memory
2260 policies, no node needs to be specified. For ``prefer``, only one node is
2261 allowed. For ``bind`` and ``interleave`` the ``nodelist`` may be as
2262 follows: a comma delimited list of numbers, A-B ranges, or `all`.
2264 .. option:: cgroup=str
2266 Add job to this control group. If it doesn't exist, it will be created. The
2267 system must have a mounted cgroup blkio mount point for this to work. If
2268 your system doesn't have it mounted, you can do so with::
2270 # mount -t cgroup -o blkio none /cgroup
2272 .. option:: cgroup_weight=int
2274 Set the weight of the cgroup to this value. See the documentation that comes
2275 with the kernel, allowed values are in the range of 100..1000.
2277 .. option:: cgroup_nodelete=bool
2279 Normally fio will delete the cgroups it has created after the job
2280 completion. To override this behavior and to leave cgroups around after the
2281 job completion, set ``cgroup_nodelete=1``. This can be useful if one wants
2282 to inspect various cgroup files after job completion. Default: false.
2284 .. option:: flow_id=int
2286 The ID of the flow. If not specified, it defaults to being a global
2287 flow. See :option:`flow`.
2289 .. option:: flow=int
2291 Weight in token-based flow control. If this value is used, then there is a
2292 'flow counter' which is used to regulate the proportion of activity between
2293 two or more jobs. Fio attempts to keep this flow counter near zero. The
2294 ``flow`` parameter stands for how much should be added or subtracted to the
2295 flow counter on each iteration of the main I/O loop. That is, if one job has
2296 ``flow=8`` and another job has ``flow=-1``, then there will be a roughly 1:8
2297 ratio in how much one runs vs the other.
2299 .. option:: flow_watermark=int
2301 The maximum value that the absolute value of the flow counter is allowed to
2302 reach before the job must wait for a lower value of the counter.
2304 .. option:: flow_sleep=int
2306 The period of time, in microseconds, to wait after the flow watermark has
2307 been exceeded before retrying operations.
2309 .. option:: stonewall, wait_for_previous
2311 Wait for preceding jobs in the job file to exit, before starting this
2312 one. Can be used to insert serialization points in the job file. A stone
2313 wall also implies starting a new reporting group, see
2314 :option:`group_reporting`.
2318 By default, fio will continue running all other jobs when one job finishes
2319 but sometimes this is not the desired action. Setting ``exitall`` will
2320 instead make fio terminate all other jobs when one job finishes.
2322 .. option:: exec_prerun=str
2324 Before running this job, issue the command specified through
2325 :manpage:`system(3)`. Output is redirected in a file called
2326 :file:`jobname.prerun.txt`.
2328 .. option:: exec_postrun=str
2330 After the job completes, issue the command specified though
2331 :manpage:`system(3)`. Output is redirected in a file called
2332 :file:`jobname.postrun.txt`.
2336 Instead of running as the invoking user, set the user ID to this value
2337 before the thread/process does any work.
2341 Set group ID, see :option:`uid`.
2347 .. option:: verify_only
2349 Do not perform specified workload, only verify data still matches previous
2350 invocation of this workload. This option allows one to check data multiple
2351 times at a later date without overwriting it. This option makes sense only
2352 for workloads that write data, and does not support workloads with the
2353 :option:`time_based` option set.
2355 .. option:: do_verify=bool
2357 Run the verify phase after a write phase. Only valid if :option:`verify` is
2360 .. option:: verify=str
2362 If writing to a file, fio can verify the file contents after each iteration
2363 of the job. Each verification method also implies verification of special
2364 header, which is written to the beginning of each block. This header also
2365 includes meta information, like offset of the block, block number, timestamp
2366 when block was written, etc. :option:`verify` can be combined with
2367 :option:`verify_pattern` option. The allowed values are:
2370 Use an md5 sum of the data area and store it in the header of
2374 Use an experimental crc64 sum of the data area and store it in the
2375 header of each block.
2378 Use a crc32c sum of the data area and store it in the header of
2379 each block. This will automatically use hardware acceleration
2380 (e.g. SSE4.2 on an x86 or CRC crypto extensions on ARM64) but will
2381 fall back to software crc32c if none is found. Generally the
2382 fatest checksum fio supports when hardware accelerated.
2388 Use a crc32 sum of the data area and store it in the header of each
2392 Use a crc16 sum of the data area and store it in the header of each
2396 Use a crc7 sum of the data area and store it in the header of each
2400 Use xxhash as the checksum function. Generally the fastest software
2401 checksum that fio supports.
2404 Use sha512 as the checksum function.
2407 Use sha256 as the checksum function.
2410 Use optimized sha1 as the checksum function.
2413 Use optimized sha3-224 as the checksum function.
2416 Use optimized sha3-256 as the checksum function.
2419 Use optimized sha3-384 as the checksum function.
2422 Use optimized sha3-512 as the checksum function.
2425 This option is deprecated, since now meta information is included in
2426 generic verification header and meta verification happens by
2427 default. For detailed information see the description of the
2428 :option:`verify` setting. This option is kept because of
2429 compatibility's sake with old configurations. Do not use it.
2432 Verify a strict pattern. Normally fio includes a header with some
2433 basic information and checksumming, but if this option is set, only
2434 the specific pattern set with :option:`verify_pattern` is verified.
2437 Only pretend to verify. Useful for testing internals with
2438 :option:`ioengine`\=null, not for much else.
2440 This option can be used for repeated burn-in tests of a system to make sure
2441 that the written data is also correctly read back. If the data direction
2442 given is a read or random read, fio will assume that it should verify a
2443 previously written file. If the data direction includes any form of write,
2444 the verify will be of the newly written data.
2446 .. option:: verifysort=bool
2448 If true, fio will sort written verify blocks when it deems it faster to read
2449 them back in a sorted manner. This is often the case when overwriting an
2450 existing file, since the blocks are already laid out in the file system. You
2451 can ignore this option unless doing huge amounts of really fast I/O where
2452 the red-black tree sorting CPU time becomes significant. Default: true.
2454 .. option:: verifysort_nr=int
2456 Pre-load and sort verify blocks for a read workload.
2458 .. option:: verify_offset=int
2460 Swap the verification header with data somewhere else in the block before
2461 writing. It is swapped back before verifying.
2463 .. option:: verify_interval=int
2465 Write the verification header at a finer granularity than the
2466 :option:`blocksize`. It will be written for chunks the size of
2467 ``verify_interval``. :option:`blocksize` should divide this evenly.
2469 .. option:: verify_pattern=str
2471 If set, fio will fill the I/O buffers with this pattern. Fio defaults to
2472 filling with totally random bytes, but sometimes it's interesting to fill
2473 with a known pattern for I/O verification purposes. Depending on the width
2474 of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time (it can
2475 be either a decimal or a hex number). The ``verify_pattern`` if larger than
2476 a 32-bit quantity has to be a hex number that starts with either "0x" or
2477 "0X". Use with :option:`verify`. Also, ``verify_pattern`` supports %o
2478 format, which means that for each block offset will be written and then
2479 verified back, e.g.::
2483 Or use combination of everything::
2485 verify_pattern=0xff%o"abcd"-12
2487 .. option:: verify_fatal=bool
2489 Normally fio will keep checking the entire contents before quitting on a
2490 block verification failure. If this option is set, fio will exit the job on
2491 the first observed failure. Default: false.
2493 .. option:: verify_dump=bool
2495 If set, dump the contents of both the original data block and the data block
2496 we read off disk to files. This allows later analysis to inspect just what
2497 kind of data corruption occurred. Off by default.
2499 .. option:: verify_async=int
2501 Fio will normally verify I/O inline from the submitting thread. This option
2502 takes an integer describing how many async offload threads to create for I/O
2503 verification instead, causing fio to offload the duty of verifying I/O
2504 contents to one or more separate threads. If using this offload option, even
2505 sync I/O engines can benefit from using an :option:`iodepth` setting higher
2506 than 1, as it allows them to have I/O in flight while verifies are running.
2507 Defaults to 0 async threads, i.e. verification is not asynchronous.
2509 .. option:: verify_async_cpus=str
2511 Tell fio to set the given CPU affinity on the async I/O verification
2512 threads. See :option:`cpus_allowed` for the format used.
2514 .. option:: verify_backlog=int
2516 Fio will normally verify the written contents of a job that utilizes verify
2517 once that job has completed. In other words, everything is written then
2518 everything is read back and verified. You may want to verify continually
2519 instead for a variety of reasons. Fio stores the meta data associated with
2520 an I/O block in memory, so for large verify workloads, quite a bit of memory
2521 would be used up holding this meta data. If this option is enabled, fio will
2522 write only N blocks before verifying these blocks.
2524 .. option:: verify_backlog_batch=int
2526 Control how many blocks fio will verify if :option:`verify_backlog` is
2527 set. If not set, will default to the value of :option:`verify_backlog`
2528 (meaning the entire queue is read back and verified). If
2529 ``verify_backlog_batch`` is less than :option:`verify_backlog` then not all
2530 blocks will be verified, if ``verify_backlog_batch`` is larger than
2531 :option:`verify_backlog`, some blocks will be verified more than once.
2533 .. option:: verify_state_save=bool
2535 When a job exits during the write phase of a verify workload, save its
2536 current state. This allows fio to replay up until that point, if the verify
2537 state is loaded for the verify read phase. The format of the filename is,
2540 <type>-<jobname>-<jobindex>-verify.state.
2542 <type> is "local" for a local run, "sock" for a client/server socket
2543 connection, and "ip" (192.168.0.1, for instance) for a networked
2544 client/server connection. Defaults to true.
2546 .. option:: verify_state_load=bool
2548 If a verify termination trigger was used, fio stores the current write state
2549 of each thread. This can be used at verification time so that fio knows how
2550 far it should verify. Without this information, fio will run a full
2551 verification pass, according to the settings in the job file used. Default
2554 .. option:: trim_percentage=int
2556 Number of verify blocks to discard/trim.
2558 .. option:: trim_verify_zero=bool
2560 Verify that trim/discarded blocks are returned as zeros.
2562 .. option:: trim_backlog=int
2564 Verify that trim/discarded blocks are returned as zeros.
2566 .. option:: trim_backlog_batch=int
2568 Trim this number of I/O blocks.
2570 .. option:: experimental_verify=bool
2572 Enable experimental verification.
2578 .. option:: steadystate=str:float, ss=str:float
2580 Define the criterion and limit for assessing steady state performance. The
2581 first parameter designates the criterion whereas the second parameter sets
2582 the threshold. When the criterion falls below the threshold for the
2583 specified duration, the job will stop. For example, `iops_slope:0.1%` will
2584 direct fio to terminate the job when the least squares regression slope
2585 falls below 0.1% of the mean IOPS. If :option:`group_reporting` is enabled
2586 this will apply to all jobs in the group. Below is the list of available
2587 steady state assessment criteria. All assessments are carried out using only
2588 data from the rolling collection window. Threshold limits can be expressed
2589 as a fixed value or as a percentage of the mean in the collection window.
2592 Collect IOPS data. Stop the job if all individual IOPS measurements
2593 are within the specified limit of the mean IOPS (e.g., ``iops:2``
2594 means that all individual IOPS values must be within 2 of the mean,
2595 whereas ``iops:0.2%`` means that all individual IOPS values must be
2596 within 0.2% of the mean IOPS to terminate the job).
2599 Collect IOPS data and calculate the least squares regression
2600 slope. Stop the job if the slope falls below the specified limit.
2603 Collect bandwidth data. Stop the job if all individual bandwidth
2604 measurements are within the specified limit of the mean bandwidth.
2607 Collect bandwidth data and calculate the least squares regression
2608 slope. Stop the job if the slope falls below the specified limit.
2610 .. option:: steadystate_duration=time, ss_dur=time
2612 A rolling window of this duration will be used to judge whether steady state
2613 has been reached. Data will be collected once per second. The default is 0
2614 which disables steady state detection. When the unit is omitted, the
2615 value is interpreted in seconds.
2617 .. option:: steadystate_ramp_time=time, ss_ramp=time
2619 Allow the job to run for the specified duration before beginning data
2620 collection for checking the steady state job termination criterion. The
2621 default is 0. When the unit is omitted, the value is interpreted in seconds.
2624 Measurements and reporting
2625 ~~~~~~~~~~~~~~~~~~~~~~~~~~
2627 .. option:: per_job_logs=bool
2629 If set, this generates bw/clat/iops log with per file private filenames. If
2630 not set, jobs with identical names will share the log filename. Default:
2633 .. option:: group_reporting
2635 It may sometimes be interesting to display statistics for groups of jobs as
2636 a whole instead of for each individual job. This is especially true if
2637 :option:`numjobs` is used; looking at individual thread/process output
2638 quickly becomes unwieldy. To see the final report per-group instead of
2639 per-job, use :option:`group_reporting`. Jobs in a file will be part of the
2640 same reporting group, unless if separated by a :option:`stonewall`, or by
2641 using :option:`new_group`.
2643 .. option:: new_group
2645 Start a new reporting group. See: :option:`group_reporting`. If not given,
2646 all jobs in a file will be part of the same reporting group, unless
2647 separated by a :option:`stonewall`.
2651 By default, fio collects and shows final output results for all jobs
2652 that run. If this option is set to 0, then fio will ignore it in
2653 the final stat output.
2655 .. option:: write_bw_log=str
2657 If given, write a bandwidth log for this job. Can be used to store data of
2658 the bandwidth of the jobs in their lifetime. The included
2659 :command:`fio_generate_plots` script uses :command:`gnuplot` to turn these
2660 text files into nice graphs. See :option:`write_lat_log` for behavior of
2661 given filename. For this option, the postfix is :file:`_bw.x.log`, where `x`
2662 is the index of the job (`1..N`, where `N` is the number of jobs). If
2663 :option:`per_job_logs` is false, then the filename will not include the job
2664 index. See `Log File Formats`_.
2666 .. option:: write_lat_log=str
2668 Same as :option:`write_bw_log`, except that this option stores I/O
2669 submission, completion, and total latencies instead. If no filename is given
2670 with this option, the default filename of :file:`jobname_type.log` is
2671 used. Even if the filename is given, fio will still append the type of
2672 log. So if one specifies::
2676 The actual log names will be :file:`foo_slat.x.log`, :file:`foo_clat.x.log`,
2677 and :file:`foo_lat.x.log`, where `x` is the index of the job (1..N, where N
2678 is the number of jobs). This helps :command:`fio_generate_plot` find the
2679 logs automatically. If :option:`per_job_logs` is false, then the filename
2680 will not include the job index. See `Log File Formats`_.
2682 .. option:: write_hist_log=str
2684 Same as :option:`write_lat_log`, but writes I/O completion latency
2685 histograms. If no filename is given with this option, the default filename
2686 of :file:`jobname_clat_hist.x.log` is used, where `x` is the index of the
2687 job (1..N, where `N` is the number of jobs). Even if the filename is given,
2688 fio will still append the type of log. If :option:`per_job_logs` is false,
2689 then the filename will not include the job index. See `Log File Formats`_.
2691 .. option:: write_iops_log=str
2693 Same as :option:`write_bw_log`, but writes IOPS. If no filename is given
2694 with this option, the default filename of :file:`jobname_type.x.log` is
2695 used,where `x` is the index of the job (1..N, where `N` is the number of
2696 jobs). Even if the filename is given, fio will still append the type of
2697 log. If :option:`per_job_logs` is false, then the filename will not include
2698 the job index. See `Log File Formats`_.
2700 .. option:: log_avg_msec=int
2702 By default, fio will log an entry in the iops, latency, or bw log for every
2703 I/O that completes. When writing to the disk log, that can quickly grow to a
2704 very large size. Setting this option makes fio average the each log entry
2705 over the specified period of time, reducing the resolution of the log. See
2706 :option:`log_max_value` as well. Defaults to 0, logging all entries.
2707 Also see `Log File Formats`_.
2709 .. option:: log_hist_msec=int
2711 Same as :option:`log_avg_msec`, but logs entries for completion latency
2712 histograms. Computing latency percentiles from averages of intervals using
2713 :option:`log_avg_msec` is inaccurate. Setting this option makes fio log
2714 histogram entries over the specified period of time, reducing log sizes for
2715 high IOPS devices while retaining percentile accuracy. See
2716 :option:`log_hist_coarseness` as well. Defaults to 0, meaning histogram
2717 logging is disabled.
2719 .. option:: log_hist_coarseness=int
2721 Integer ranging from 0 to 6, defining the coarseness of the resolution of
2722 the histogram logs enabled with :option:`log_hist_msec`. For each increment
2723 in coarseness, fio outputs half as many bins. Defaults to 0, for which
2724 histogram logs contain 1216 latency bins. See `Log File Formats`_.
2726 .. option:: log_max_value=bool
2728 If :option:`log_avg_msec` is set, fio logs the average over that window. If
2729 you instead want to log the maximum value, set this option to 1. Defaults to
2730 0, meaning that averaged values are logged.
2732 .. option:: log_offset=int
2734 If this is set, the iolog options will include the byte offset for the I/O
2735 entry as well as the other data values.
2737 .. option:: log_compression=int
2739 If this is set, fio will compress the I/O logs as it goes, to keep the
2740 memory footprint lower. When a log reaches the specified size, that chunk is
2741 removed and compressed in the background. Given that I/O logs are fairly
2742 highly compressible, this yields a nice memory savings for longer runs. The
2743 downside is that the compression will consume some background CPU cycles, so
2744 it may impact the run. This, however, is also true if the logging ends up
2745 consuming most of the system memory. So pick your poison. The I/O logs are
2746 saved normally at the end of a run, by decompressing the chunks and storing
2747 them in the specified log file. This feature depends on the availability of
2750 .. option:: log_compression_cpus=str
2752 Define the set of CPUs that are allowed to handle online log compression for
2753 the I/O jobs. This can provide better isolation between performance
2754 sensitive jobs, and background compression work.
2756 .. option:: log_store_compressed=bool
2758 If set, fio will store the log files in a compressed format. They can be
2759 decompressed with fio, using the :option:`--inflate-log` command line
2760 parameter. The files will be stored with a :file:`.fz` suffix.
2762 .. option:: log_unix_epoch=bool
2764 If set, fio will log Unix timestamps to the log files produced by enabling
2765 write_type_log for each log type, instead of the default zero-based
2768 .. option:: block_error_percentiles=bool
2770 If set, record errors in trim block-sized units from writes and trims and
2771 output a histogram of how many trims it took to get to errors, and what kind
2772 of error was encountered.
2774 .. option:: bwavgtime=int
2776 Average the calculated bandwidth over the given time. Value is specified in
2777 milliseconds. If the job also does bandwidth logging through
2778 :option:`write_bw_log`, then the minimum of this option and
2779 :option:`log_avg_msec` will be used. Default: 500ms.
2781 .. option:: iopsavgtime=int
2783 Average the calculated IOPS over the given time. Value is specified in
2784 milliseconds. If the job also does IOPS logging through
2785 :option:`write_iops_log`, then the minimum of this option and
2786 :option:`log_avg_msec` will be used. Default: 500ms.
2788 .. option:: disk_util=bool
2790 Generate disk utilization statistics, if the platform supports it.
2793 .. option:: disable_lat=bool
2795 Disable measurements of total latency numbers. Useful only for cutting back
2796 the number of calls to :manpage:`gettimeofday(2)`, as that does impact
2797 performance at really high IOPS rates. Note that to really get rid of a
2798 large amount of these calls, this option must be used with
2799 :option:`disable_slat` and :option:`disable_bw_measurement` as well.
2801 .. option:: disable_clat=bool
2803 Disable measurements of completion latency numbers. See
2804 :option:`disable_lat`.
2806 .. option:: disable_slat=bool
2808 Disable measurements of submission latency numbers. See
2809 :option:`disable_slat`.
2811 .. option:: disable_bw_measurement=bool, disable_bw=bool
2813 Disable measurements of throughput/bandwidth numbers. See
2814 :option:`disable_lat`.
2816 .. option:: clat_percentiles=bool
2818 Enable the reporting of percentiles of completion latencies.
2820 .. option:: percentile_list=float_list
2822 Overwrite the default list of percentiles for completion latencies and the
2823 block error histogram. Each number is a floating number in the range
2824 (0,100], and the maximum length of the list is 20. Use ``:`` to separate the
2825 numbers, and list the numbers in ascending order. For example,
2826 ``--percentile_list=99.5:99.9`` will cause fio to report the values of
2827 completion latency below which 99.5% and 99.9% of the observed latencies
2834 .. option:: exitall_on_error
2836 When one job finishes in error, terminate the rest. The default is to wait
2837 for each job to finish.
2839 .. option:: continue_on_error=str
2841 Normally fio will exit the job on the first observed failure. If this option
2842 is set, fio will continue the job when there is a 'non-fatal error' (EIO or
2843 EILSEQ) until the runtime is exceeded or the I/O size specified is
2844 completed. If this option is used, there are two more stats that are
2845 appended, the total error count and the first error. The error field given
2846 in the stats is the first error that was hit during the run.
2848 The allowed values are:
2851 Exit on any I/O or verify errors.
2854 Continue on read errors, exit on all others.
2857 Continue on write errors, exit on all others.
2860 Continue on any I/O error, exit on all others.
2863 Continue on verify errors, exit on all others.
2866 Continue on all errors.
2869 Backward-compatible alias for 'none'.
2872 Backward-compatible alias for 'all'.
2874 .. option:: ignore_error=str
2876 Sometimes you want to ignore some errors during test in that case you can
2877 specify error list for each error type, instead of only being able to
2878 ignore the default 'non-fatal error' using :option:`continue_on_error`.
2879 ``ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST`` errors for
2880 given error type is separated with ':'. Error may be symbol ('ENOSPC',
2881 'ENOMEM') or integer. Example::
2883 ignore_error=EAGAIN,ENOSPC:122
2885 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from
2886 WRITE. This option works by overriding :option:`continue_on_error` with
2887 the list of errors for each error type if any.
2889 .. option:: error_dump=bool
2891 If set dump every error even if it is non fatal, true by default. If
2892 disabled only fatal error will be dumped.
2894 Running predefined workloads
2895 ----------------------------
2897 Fio includes predefined profiles that mimic the I/O workloads generated by
2900 .. option:: profile=str
2902 The predefined workload to run. Current profiles are:
2905 Threaded I/O bench (tiotest/tiobench) like workload.
2908 Aerospike Certification Tool (ACT) like workload.
2910 To view a profile's additional options use :option:`--cmdhelp` after specifying
2911 the profile. For example::
2913 $ fio --profile=act --cmdhelp
2918 .. option:: device-names=str
2923 .. option:: load=int
2926 ACT load multiplier. Default: 1.
2928 .. option:: test-duration=time
2931 How long the entire test takes to run. When the unit is omitted, the value
2932 is given in seconds. Default: 24h.
2934 .. option:: threads-per-queue=int
2937 Number of read IO threads per device. Default: 8.
2939 .. option:: read-req-num-512-blocks=int
2942 Number of 512B blocks to read at the time. Default: 3.
2944 .. option:: large-block-op-kbytes=int
2947 Size of large block ops in KiB (writes). Default: 131072.
2952 Set to run ACT prep phase.
2954 Tiobench profile options
2955 ~~~~~~~~~~~~~~~~~~~~~~~~
2957 .. option:: size=str
2962 .. option:: block=int
2965 Block size in bytes. Default: 4096.
2967 .. option:: numruns=int
2977 .. option:: threads=int
2982 Interpreting the output
2983 -----------------------
2986 Example output was based on the following:
2987 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --time_based \
2988 --rate=1256k --bs=14K --name=quick --runtime=1s --name=mixed \
2989 --runtime=2m --rw=rw
2991 Fio spits out a lot of output. While running, fio will display the status of the
2992 jobs created. An example of that would be::
2994 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]
2996 The characters inside the first set of square brackets denote the current status of
2997 each thread. The first character is the first job defined in the job file, and so
2998 forth. The possible values (in typical life cycle order) are:
3000 +------+-----+-----------------------------------------------------------+
3002 +======+=====+===========================================================+
3003 | P | | Thread setup, but not started. |
3004 +------+-----+-----------------------------------------------------------+
3005 | C | | Thread created. |
3006 +------+-----+-----------------------------------------------------------+
3007 | I | | Thread initialized, waiting or generating necessary data. |
3008 +------+-----+-----------------------------------------------------------+
3009 | | p | Thread running pre-reading file(s). |
3010 +------+-----+-----------------------------------------------------------+
3011 | | / | Thread is in ramp period. |
3012 +------+-----+-----------------------------------------------------------+
3013 | | R | Running, doing sequential reads. |
3014 +------+-----+-----------------------------------------------------------+
3015 | | r | Running, doing random reads. |
3016 +------+-----+-----------------------------------------------------------+
3017 | | W | Running, doing sequential writes. |
3018 +------+-----+-----------------------------------------------------------+
3019 | | w | Running, doing random writes. |
3020 +------+-----+-----------------------------------------------------------+
3021 | | M | Running, doing mixed sequential reads/writes. |
3022 +------+-----+-----------------------------------------------------------+
3023 | | m | Running, doing mixed random reads/writes. |
3024 +------+-----+-----------------------------------------------------------+
3025 | | D | Running, doing sequential trims. |
3026 +------+-----+-----------------------------------------------------------+
3027 | | d | Running, doing random trims. |
3028 +------+-----+-----------------------------------------------------------+
3029 | | F | Running, currently waiting for :manpage:`fsync(2)`. |
3030 +------+-----+-----------------------------------------------------------+
3031 | | V | Running, doing verification of written data. |
3032 +------+-----+-----------------------------------------------------------+
3033 | f | | Thread finishing. |
3034 +------+-----+-----------------------------------------------------------+
3035 | E | | Thread exited, not reaped by main thread yet. |
3036 +------+-----+-----------------------------------------------------------+
3037 | _ | | Thread reaped. |
3038 +------+-----+-----------------------------------------------------------+
3039 | X | | Thread reaped, exited with an error. |
3040 +------+-----+-----------------------------------------------------------+
3041 | K | | Thread reaped, exited due to signal. |
3042 +------+-----+-----------------------------------------------------------+
3045 Example output was based on the following:
3046 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --runtime=58m \
3047 --time_based --rate=2512k --bs=256K --numjobs=10 \
3048 --name=readers --rw=read --name=writers --rw=write
3050 Fio will condense the thread string as not to take up more space on the command
3051 line than needed. For instance, if you have 10 readers and 10 writers running,
3052 the output would look like this::
3054 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]
3056 Note that the status string is displayed in order, so it's possible to tell which of
3057 the jobs are currently doing what. In the example above this means that jobs 1--10
3058 are readers and 11--20 are writers.
3060 The other values are fairly self explanatory -- number of threads currently
3061 running and doing I/O, the number of currently open files (f=), the estimated
3062 completion percentage, the rate of I/O since last check (read speed listed first,
3063 then write speed and optionally trim speed) in terms of bandwidth and IOPS, and time to completion for the current
3064 running group. It's impossible to estimate runtime of the following groups (if
3068 Example output was based on the following:
3069 TZ=UTC fio --iodepth=16 --ioengine=posixaio --filename=/tmp/fiofile \
3070 --direct=1 --size=100M --time_based --runtime=50s --rate_iops=89 \
3071 --bs=7K --name=Client1 --rw=write
3073 When fio is done (or interrupted by :kbd:`Ctrl-C`), it will show the data for
3074 each thread, group of threads, and disks in that order. For each overall thread (or
3075 group) the output looks like::
3077 Client1: (groupid=0, jobs=1): err= 0: pid=16109: Sat Jun 24 12:07:54 2017
3078 write: IOPS=88, BW=623KiB/s (638kB/s)(30.4MiB/50032msec)
3079 slat (nsec): min=500, max=145500, avg=8318.00, stdev=4781.50
3080 clat (usec): min=170, max=78367, avg=4019.02, stdev=8293.31
3081 lat (usec): min=174, max=78375, avg=4027.34, stdev=8291.79
3082 clat percentiles (usec):
3083 | 1.00th=[ 302], 5.00th=[ 326], 10.00th=[ 343], 20.00th=[ 363],
3084 | 30.00th=[ 392], 40.00th=[ 404], 50.00th=[ 416], 60.00th=[ 445],
3085 | 70.00th=[ 816], 80.00th=[ 6718], 90.00th=[12911], 95.00th=[21627],
3086 | 99.00th=[43779], 99.50th=[51643], 99.90th=[68682], 99.95th=[72877],
3088 bw ( KiB/s): min= 532, max= 686, per=0.10%, avg=622.87, stdev=24.82, samples= 100
3089 iops : min= 76, max= 98, avg=88.98, stdev= 3.54, samples= 100
3090 lat (usec) : 250=0.04%, 500=64.11%, 750=4.81%, 1000=2.79%
3091 lat (msec) : 2=4.16%, 4=1.84%, 10=4.90%, 20=11.33%, 50=5.37%
3092 lat (msec) : 100=0.65%
3093 cpu : usr=0.27%, sys=0.18%, ctx=12072, majf=0, minf=21
3094 IO depths : 1=85.0%, 2=13.1%, 4=1.8%, 8=0.1%, 16=0.0%, 32=0.0%, >=64=0.0%
3095 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
3096 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
3097 issued rwt: total=0,4450,0, short=0,0,0, dropped=0,0,0
3098 latency : target=0, window=0, percentile=100.00%, depth=8
3100 The job name (or first job's name when using :option:`group_reporting`) is printed,
3101 along with the group id, count of jobs being aggregated, last error id seen (which
3102 is 0 when there are no errors), pid/tid of that thread and the time the job/group
3103 completed. Below are the I/O statistics for each data direction performed (showing
3104 writes in the example above). In the order listed, they denote:
3107 The string before the colon shows the I/O direction the statistics
3108 are for. **IOPS** is the average I/Os performed per second. **BW**
3109 is the average bandwidth rate shown as: value in power of 2 format
3110 (value in power of 10 format). The last two values show: (**total
3111 I/O performed** in power of 2 format / **runtime** of that thread).
3114 Submission latency (**min** being the minimum, **max** being the
3115 maximum, **avg** being the average, **stdev** being the standard
3116 deviation). This is the time it took to submit the I/O. For
3117 sync I/O this row is not displayed as the slat is really the
3118 completion latency (since queue/complete is one operation there).
3119 This value can be in nanoseconds, microseconds or milliseconds ---
3120 fio will choose the most appropriate base and print that (in the
3121 example above nanoseconds was the best scale). Note: in :option:`--minimal` mode
3122 latencies are always expressed in microseconds.
3125 Completion latency. Same names as slat, this denotes the time from
3126 submission to completion of the I/O pieces. For sync I/O, clat will
3127 usually be equal (or very close) to 0, as the time from submit to
3128 complete is basically just CPU time (I/O has already been done, see slat
3132 Bandwidth statistics based on samples. Same names as the xlat stats,
3133 but also includes the number of samples taken (**samples**) and an
3134 approximate percentage of total aggregate bandwidth this thread
3135 received in its group (**per**). This last value is only really
3136 useful if the threads in this group are on the same disk, since they
3137 are then competing for disk access.
3140 IOPS statistics based on samples. Same names as bw.
3143 CPU usage. User and system time, along with the number of context
3144 switches this thread went through, usage of system and user time, and
3145 finally the number of major and minor page faults. The CPU utilization
3146 numbers are averages for the jobs in that reporting group, while the
3147 context and fault counters are summed.
3150 The distribution of I/O depths over the job lifetime. The numbers are
3151 divided into powers of 2 and each entry covers depths from that value
3152 up to those that are lower than the next entry -- e.g., 16= covers
3153 depths from 16 to 31. Note that the range covered by a depth
3154 distribution entry can be different to the range covered by the
3155 equivalent submit/complete distribution entry.
3158 How many pieces of I/O were submitting in a single submit call. Each
3159 entry denotes that amount and below, until the previous entry -- e.g.,
3160 16=100% means that we submitted anywhere between 9 to 16 I/Os per submit
3161 call. Note that the range covered by a submit distribution entry can
3162 be different to the range covered by the equivalent depth distribution
3166 Like the above submit number, but for completions instead.
3169 The number of read/write/trim requests issued, and how many of them were
3173 The distribution of I/O completion latencies. This is the time from when
3174 I/O leaves fio and when it gets completed. The numbers follow the same
3175 pattern as the I/O depths, meaning that 2=1.6% means that 1.6% of the
3176 I/O completed within 2 msecs, 20=12.8% means that 12.8% of the I/O took
3177 more than 10 msecs, but less than (or equal to) 20 msecs.
3180 Example output was based on the following:
3181 TZ=UTC fio --ioengine=null --iodepth=2 --size=100M --numjobs=2 \
3182 --rate_process=poisson --io_limit=32M --name=read --bs=128k \
3183 --rate=11M --name=write --rw=write --bs=2k --rate=700k
3185 After each client has been listed, the group statistics are printed. They
3186 will look like this::
3188 Run status group 0 (all jobs):
3189 READ: bw=20.9MiB/s (21.9MB/s), 10.4MiB/s-10.8MiB/s (10.9MB/s-11.3MB/s), io=64.0MiB (67.1MB), run=2973-3069msec
3190 WRITE: bw=1231KiB/s (1261kB/s), 616KiB/s-621KiB/s (630kB/s-636kB/s), io=64.0MiB (67.1MB), run=52747-53223msec
3192 For each data direction it prints:
3195 Aggregate bandwidth of threads in this group followed by the
3196 minimum and maximum bandwidth of all the threads in this group.
3197 Values outside of brackets are power-of-2 format and those
3198 within are the equivalent value in a power-of-10 format.
3200 Aggregate I/O performed of all threads in this group. The
3201 format is the same as bw.
3203 The smallest and longest runtimes of the threads in this group.
3205 And finally, the disk statistics are printed. They will look like this::
3207 Disk stats (read/write):
3208 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
3210 Each value is printed for both reads and writes, with reads first. The
3214 Number of I/Os performed by all groups.
3216 Number of merges I/O the I/O scheduler.
3218 Number of ticks we kept the disk busy.
3220 Total time spent in the disk queue.
3222 The disk utilization. A value of 100% means we kept the disk
3223 busy constantly, 50% would be a disk idling half of the time.
3225 It is also possible to get fio to dump the current output while it is running,
3226 without terminating the job. To do that, send fio the **USR1** signal. You can
3227 also get regularly timed dumps by using the :option:`--status-interval`
3228 parameter, or by creating a file in :file:`/tmp` named
3229 :file:`fio-dump-status`. If fio sees this file, it will unlink it and dump the
3230 current output status.
3236 For scripted usage where you typically want to generate tables or graphs of the
3237 results, fio can output the results in a semicolon separated format. The format
3238 is one long line of values, such as::
3240 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%
3241 A description of this job goes here.
3243 The job description (if provided) follows on a second line.
3245 To enable terse output, use the :option:`--minimal` or
3246 :option:`--output-format`\=terse command line options. The
3247 first value is the version of the terse output format. If the output has to be
3248 changed for some reason, this number will be incremented by 1 to signify that
3251 Split up, the format is as follows (comments in brackets denote when a
3252 field was introduced or whether its specific to some terse version):
3256 terse version, fio version [v3], jobname, groupid, error
3260 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3261 Submission latency: min, max, mean, stdev (usec)
3262 Completion latency: min, max, mean, stdev (usec)
3263 Completion latency percentiles: 20 fields (see below)
3264 Total latency: min, max, mean, stdev (usec)
3265 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
3266 IOPS [v5]: min, max, mean, stdev, number of samples
3272 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3273 Submission latency: min, max, mean, stdev (usec)
3274 Completion latency: min, max, mean, stdev (usec)
3275 Completion latency percentiles: 20 fields (see below)
3276 Total latency: min, max, mean, stdev (usec)
3277 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
3278 IOPS [v5]: min, max, mean, stdev, number of samples
3280 TRIM status [all but version 3]:
3282 Fields are similar to READ/WRITE status.
3286 user, system, context switches, major faults, minor faults
3290 <=1, 2, 4, 8, 16, 32, >=64
3292 I/O latencies microseconds::
3294 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
3296 I/O latencies milliseconds::
3298 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
3300 Disk utilization [v3]::
3302 Disk name, Read ios, write ios,
3303 Read merges, write merges,
3304 Read ticks, write ticks,
3305 Time spent in queue, disk utilization percentage
3307 Additional Info (dependent on continue_on_error, default off)::
3309 total # errors, first error code
3311 Additional Info (dependent on description being set)::
3315 Completion latency percentiles can be a grouping of up to 20 sets, so for the
3316 terse output fio writes all of them. Each field will look like this::
3320 which is the Xth percentile, and the `usec` latency associated with it.
3322 For disk utilization, all disks used by fio are shown. So for each disk there
3323 will be a disk utilization section.
3325 Below is a single line containing short names for each of the fields in the
3326 minimal output v3, separated by semicolons::
3328 terse_version_3;fio_version;jobname;groupid;error;read_kb;read_bandwidth;read_iops;read_runtime_ms;read_slat_min;read_slat_max;read_slat_mean;read_slat_dev;read_clat_min;read_clat_max;read_clat_mean;read_clat_dev;read_clat_pct01;read_clat_pct02;read_clat_pct03;read_clat_pct04;read_clat_pct05;read_clat_pct06;read_clat_pct07;read_clat_pct08;read_clat_pct09;read_clat_pct10;read_clat_pct11;read_clat_pct12;read_clat_pct13;read_clat_pct14;read_clat_pct15;read_clat_pct16;read_clat_pct17;read_clat_pct18;read_clat_pct19;read_clat_pct20;read_tlat_min;read_lat_max;read_lat_mean;read_lat_dev;read_bw_min;read_bw_max;read_bw_agg_pct;read_bw_mean;read_bw_dev;write_kb;write_bandwidth;write_iops;write_runtime_ms;write_slat_min;write_slat_max;write_slat_mean;write_slat_dev;write_clat_min;write_clat_max;write_clat_mean;write_clat_dev;write_clat_pct01;write_clat_pct02;write_clat_pct03;write_clat_pct04;write_clat_pct05;write_clat_pct06;write_clat_pct07;write_clat_pct08;write_clat_pct09;write_clat_pct10;write_clat_pct11;write_clat_pct12;write_clat_pct13;write_clat_pct14;write_clat_pct15;write_clat_pct16;write_clat_pct17;write_clat_pct18;write_clat_pct19;write_clat_pct20;write_tlat_min;write_lat_max;write_lat_mean;write_lat_dev;write_bw_min;write_bw_max;write_bw_agg_pct;write_bw_mean;write_bw_dev;cpu_user;cpu_sys;cpu_csw;cpu_mjf;cpu_minf;iodepth_1;iodepth_2;iodepth_4;iodepth_8;iodepth_16;iodepth_32;iodepth_64;lat_2us;lat_4us;lat_10us;lat_20us;lat_50us;lat_100us;lat_250us;lat_500us;lat_750us;lat_1000us;lat_2ms;lat_4ms;lat_10ms;lat_20ms;lat_50ms;lat_100ms;lat_250ms;lat_500ms;lat_750ms;lat_1000ms;lat_2000ms;lat_over_2000ms;disk_name;disk_read_iops;disk_write_iops;disk_read_merges;disk_write_merges;disk_read_ticks;write_ticks;disk_queue_time;disk_util
3334 There are two trace file format that you can encounter. The older (v1) format is
3335 unsupported since version 1.20-rc3 (March 2008). It will still be described
3336 below in case that you get an old trace and want to understand it.
3338 In any case the trace is a simple text file with a single action per line.
3341 Trace file format v1
3342 ~~~~~~~~~~~~~~~~~~~~
3344 Each line represents a single I/O action in the following format::
3348 where `rw=0/1` for read/write, and the offset and length entries being in bytes.
3350 This format is not supported in fio versions >= 1.20-rc3.
3353 Trace file format v2
3354 ~~~~~~~~~~~~~~~~~~~~
3356 The second version of the trace file format was added in fio version 1.17. It
3357 allows to access more then one file per trace and has a bigger set of possible
3360 The first line of the trace file has to be::
3364 Following this can be lines in two different formats, which are described below.
3366 The file management format::
3370 The filename is given as an absolute path. The action can be one of these:
3373 Add the given filename to the trace.
3375 Open the file with the given filename. The filename has to have
3376 been added with the **add** action before.
3378 Close the file with the given filename. The file has to have been
3382 The file I/O action format::
3384 filename action offset length
3386 The `filename` is given as an absolute path, and has to have been added and
3387 opened before it can be used with this format. The `offset` and `length` are
3388 given in bytes. The `action` can be one of these:
3391 Wait for `offset` microseconds. Everything below 100 is discarded.
3392 The time is relative to the previous `wait` statement.
3394 Read `length` bytes beginning from `offset`.
3396 Write `length` bytes beginning from `offset`.
3398 :manpage:`fsync(2)` the file.
3400 :manpage:`fdatasync(2)` the file.
3402 Trim the given file from the given `offset` for `length` bytes.
3404 CPU idleness profiling
3405 ----------------------
3407 In some cases, we want to understand CPU overhead in a test. For example, we
3408 test patches for the specific goodness of whether they reduce CPU usage.
3409 Fio implements a balloon approach to create a thread per CPU that runs at idle
3410 priority, meaning that it only runs when nobody else needs the cpu.
3411 By measuring the amount of work completed by the thread, idleness of each CPU
3412 can be derived accordingly.
3414 An unit work is defined as touching a full page of unsigned characters. Mean and
3415 standard deviation of time to complete an unit work is reported in "unit work"
3416 section. Options can be chosen to report detailed percpu idleness or overall
3417 system idleness by aggregating percpu stats.
3420 Verification and triggers
3421 -------------------------
3423 Fio is usually run in one of two ways, when data verification is done. The first
3424 is a normal write job of some sort with verify enabled. When the write phase has
3425 completed, fio switches to reads and verifies everything it wrote. The second
3426 model is running just the write phase, and then later on running the same job
3427 (but with reads instead of writes) to repeat the same I/O patterns and verify
3428 the contents. Both of these methods depend on the write phase being completed,
3429 as fio otherwise has no idea how much data was written.
3431 With verification triggers, fio supports dumping the current write state to
3432 local files. Then a subsequent read verify workload can load this state and know
3433 exactly where to stop. This is useful for testing cases where power is cut to a
3434 server in a managed fashion, for instance.
3436 A verification trigger consists of two things:
3438 1) Storing the write state of each job.
3439 2) Executing a trigger command.
3441 The write state is relatively small, on the order of hundreds of bytes to single
3442 kilobytes. It contains information on the number of completions done, the last X
3445 A trigger is invoked either through creation ('touch') of a specified file in
3446 the system, or through a timeout setting. If fio is run with
3447 :option:`--trigger-file`\= :file:`/tmp/trigger-file`, then it will continually
3448 check for the existence of :file:`/tmp/trigger-file`. When it sees this file, it
3449 will fire off the trigger (thus saving state, and executing the trigger
3452 For client/server runs, there's both a local and remote trigger. If fio is
3453 running as a server backend, it will send the job states back to the client for
3454 safe storage, then execute the remote trigger, if specified. If a local trigger
3455 is specified, the server will still send back the write state, but the client
3456 will then execute the trigger.
3458 Verification trigger example
3459 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3461 Let's say we want to run a powercut test on the remote machine 'server'. Our
3462 write workload is in :file:`write-test.fio`. We want to cut power to 'server' at
3463 some point during the run, and we'll run this test from the safety or our local
3464 machine, 'localbox'. On the server, we'll start the fio backend normally::
3466 server# fio --server
3468 and on the client, we'll fire off the workload::
3470 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger-remote="bash -c \"echo b > /proc/sysrq-triger\""
3472 We set :file:`/tmp/my-trigger` as the trigger file, and we tell fio to execute::
3474 echo b > /proc/sysrq-trigger
3476 on the server once it has received the trigger and sent us the write state. This
3477 will work, but it's not **really** cutting power to the server, it's merely
3478 abruptly rebooting it. If we have a remote way of cutting power to the server
3479 through IPMI or similar, we could do that through a local trigger command
3480 instead. Let's assume we have a script that does IPMI reboot of a given hostname,
3481 ipmi-reboot. On localbox, we could then have run fio with a local trigger
3484 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger="ipmi-reboot server"
3486 For this case, fio would wait for the server to send us the write state, then
3487 execute ``ipmi-reboot server`` when that happened.
3489 Loading verify state
3490 ~~~~~~~~~~~~~~~~~~~~
3492 To load stored write state, a read verification job file must contain the
3493 :option:`verify_state_load` option. If that is set, fio will load the previously
3494 stored state. For a local fio run this is done by loading the files directly,
3495 and on a client/server run, the server backend will ask the client to send the
3496 files over and load them from there.
3502 Fio supports a variety of log file formats, for logging latencies, bandwidth,
3503 and IOPS. The logs share a common format, which looks like this:
3505 *time* (`msec`), *value*, *data direction*, *offset*
3507 Time for the log entry is always in milliseconds. The *value* logged depends
3508 on the type of log, it will be one of the following:
3511 Value is latency in usecs
3517 *Data direction* is one of the following:
3526 The *offset* is the offset, in bytes, from the start of the file, for that
3527 particular I/O. The logging of the offset can be toggled with
3528 :option:`log_offset`.
3530 Fio defaults to logging every individual I/O. When IOPS are logged for individual
3531 I/Os the value entry will always be 1. If windowed logging is enabled through
3532 :option:`log_avg_msec`, fio logs the average values over the specified period of time.
3533 If windowed logging is enabled and :option:`log_max_value` is set, then fio logs
3534 maximum values in that window instead of averages. Since 'data direction' and
3535 'offset' are per-I/O values, they aren't applicable if windowed logging is enabled.
3540 Normally fio is invoked as a stand-alone application on the machine where the
3541 I/O workload should be generated. However, the backend and frontend of fio can
3542 be run separately i.e., the fio server can generate an I/O workload on the "Device
3543 Under Test" while being controlled by a client on another machine.
3545 Start the server on the machine which has access to the storage DUT::
3549 where `args` defines what fio listens to. The arguments are of the form
3550 ``type,hostname`` or ``IP,port``. *type* is either ``ip`` (or ip4) for TCP/IP
3551 v4, ``ip6`` for TCP/IP v6, or ``sock`` for a local unix domain socket.
3552 *hostname* is either a hostname or IP address, and *port* is the port to listen
3553 to (only valid for TCP/IP, not a local socket). Some examples:
3557 Start a fio server, listening on all interfaces on the default port (8765).
3559 2) ``fio --server=ip:hostname,4444``
3561 Start a fio server, listening on IP belonging to hostname and on port 4444.
3563 3) ``fio --server=ip6:::1,4444``
3565 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
3567 4) ``fio --server=,4444``
3569 Start a fio server, listening on all interfaces on port 4444.
3571 5) ``fio --server=1.2.3.4``
3573 Start a fio server, listening on IP 1.2.3.4 on the default port.
3575 6) ``fio --server=sock:/tmp/fio.sock``
3577 Start a fio server, listening on the local socket :file:`/tmp/fio.sock`.
3579 Once a server is running, a "client" can connect to the fio server with::
3581 fio <local-args> --client=<server> <remote-args> <job file(s)>
3583 where `local-args` are arguments for the client where it is running, `server`
3584 is the connect string, and `remote-args` and `job file(s)` are sent to the
3585 server. The `server` string follows the same format as it does on the server
3586 side, to allow IP/hostname/socket and port strings.
3588 Fio can connect to multiple servers this way::
3590 fio --client=<server1> <job file(s)> --client=<server2> <job file(s)>
3592 If the job file is located on the fio server, then you can tell the server to
3593 load a local file as well. This is done by using :option:`--remote-config` ::
3595 fio --client=server --remote-config /path/to/file.fio
3597 Then fio will open this local (to the server) job file instead of being passed
3598 one from the client.
3600 If you have many servers (example: 100 VMs/containers), you can input a pathname
3601 of a file containing host IPs/names as the parameter value for the
3602 :option:`--client` option. For example, here is an example :file:`host.list`
3603 file containing 2 hostnames::
3605 host1.your.dns.domain
3606 host2.your.dns.domain
3608 The fio command would then be::
3610 fio --client=host.list <job file(s)>
3612 In this mode, you cannot input server-specific parameters or job files -- all
3613 servers receive the same job file.
3615 In order to let ``fio --client`` runs use a shared filesystem from multiple
3616 hosts, ``fio --client`` now prepends the IP address of the server to the
3617 filename. For example, if fio is using the directory :file:`/mnt/nfs/fio` and is
3618 writing filename :file:`fileio.tmp`, with a :option:`--client` `hostfile`
3619 containing two hostnames ``h1`` and ``h2`` with IP addresses 192.168.10.120 and
3620 192.168.10.121, then fio will create two files::
3622 /mnt/nfs/fio/192.168.10.120.fileio.tmp
3623 /mnt/nfs/fio/192.168.10.121.fileio.tmp