1 .TH fio 1 "August 2017" "User Manual"
3 fio \- flexible I/O tester
6 [\fIoptions\fR] [\fIjobfile\fR]...
9 is a tool that will spawn a number of threads or processes doing a
10 particular type of I/O action as specified by the user.
11 The typical use of fio is to write a job file matching the I/O load
12 one wants to simulate.
15 .BI \-\-debug \fR=\fPtype
16 Enable verbose tracing \fItype\fR of various fio actions. May be `all' for all \fItype\fRs
17 or individual types separated by a comma (e.g. `\-\-debug=file,mem' will enable
18 file and memory debugging). `help' will list all available tracing options.
21 Parse options only, don't start any I/O.
23 .BI \-\-merge\-blktrace\-only
24 Merge blktraces only, don't start any I/O.
26 .BI \-\-output \fR=\fPfilename
27 Write output to \fIfilename\fR.
29 .BI \-\-output\-format \fR=\fPformat
30 Set the reporting \fIformat\fR to `normal', `terse', `json', or
31 `json+'. Multiple formats can be selected, separate by a comma. `terse'
32 is a CSV based format. `json+' is like `json', except it adds a full
33 dump of the latency buckets.
35 .BI \-\-bandwidth\-log
36 Generate aggregate bandwidth logs.
39 Print statistics in a terse, semicolon\-delimited format.
42 Print statistics in selected mode AND terse, semicolon\-delimited format.
43 \fBDeprecated\fR, use \fB\-\-output\-format\fR instead to select multiple formats.
45 .BI \-\-terse\-version \fR=\fPversion
46 Set terse \fIversion\fR output format (default `3', or `2', `4', `5').
49 Print version information and exit.
52 Print a summary of the command line options and exit.
54 .BI \-\-cpuclock\-test
55 Perform test and validation of internal CPU clock.
57 .BI \-\-crctest \fR=\fP[test]
58 Test the speed of the built\-in checksumming functions. If no argument is given,
59 all of them are tested. Alternatively, a comma separated list can be passed, in which
60 case the given ones are tested.
62 .BI \-\-cmdhelp \fR=\fPcommand
63 Print help information for \fIcommand\fR. May be `all' for all commands.
65 .BI \-\-enghelp \fR=\fP[ioengine[,command]]
66 List all commands defined by \fIioengine\fR, or print help for \fIcommand\fR
67 defined by \fIioengine\fR. If no \fIioengine\fR is given, list all
71 Convert given \fIjobfile\fRs to a set of command\-line options.
74 Turn on safety read\-only checks, preventing writes and trims. The \fB\-\-readonly\fR
75 option is an extra safety guard to prevent users from accidentally starting
76 a write or trim workload when that is not desired. Fio will only modify the
77 device under test if `rw=write/randwrite/rw/randrw/trim/randtrim/trimwrite'
78 is given. This safety net can be used as an extra precaution.
80 .BI \-\-eta \fR=\fPwhen
81 Specifies when real\-time ETA estimate should be printed. \fIwhen\fR may
82 be `always', `never' or `auto'. `auto' is the default, it prints ETA when
83 requested if the output is a TTY. `always' disregards the output type, and
84 prints ETA when requested. `never' never prints ETA.
86 .BI \-\-eta\-interval \fR=\fPtime
87 By default, fio requests client ETA status roughly every second. With this
88 option, the interval is configurable. Fio imposes a minimum allowed time to
89 avoid flooding the console, less than 250 msec is not supported.
91 .BI \-\-eta\-newline \fR=\fPtime
92 Force a new line for every \fItime\fR period passed. When the unit is omitted,
93 the value is interpreted in seconds.
95 .BI \-\-status\-interval \fR=\fPtime
96 Force a full status dump of cumulative (from job start) values at \fItime\fR
97 intervals. This option does *not* provide per-period measurements. So
98 values such as bandwidth are running averages. When the time unit is omitted,
99 \fItime\fR is interpreted in seconds. Note that using this option with
100 `\-\-output-format=json' will yield output that technically isn't valid json,
101 since the output will be collated sets of valid json. It will need to be split
102 into valid sets of json after the run.
104 .BI \-\-section \fR=\fPname
105 Only run specified section \fIname\fR in job file. Multiple sections can be specified.
106 The \fB\-\-section\fR option allows one to combine related jobs into one file.
107 E.g. one job file could define light, moderate, and heavy sections. Tell
108 fio to run only the "heavy" section by giving `\-\-section=heavy'
109 command line option. One can also specify the "write" operations in one
110 section and "verify" operation in another section. The \fB\-\-section\fR option
111 only applies to job sections. The reserved *global* section is always
114 .BI \-\-alloc\-size \fR=\fPkb
115 Allocate additional internal smalloc pools of size \fIkb\fR in KiB. The
116 \fB\-\-alloc\-size\fR option increases shared memory set aside for use by fio.
117 If running large jobs with randommap enabled, fio can run out of memory.
118 Smalloc is an internal allocator for shared structures from a fixed size
119 memory pool and can grow to 16 pools. The pool size defaults to 16MiB.
120 NOTE: While running `.fio_smalloc.*' backing store files are visible
123 .BI \-\-warnings\-fatal
124 All fio parser warnings are fatal, causing fio to exit with an error.
126 .BI \-\-max\-jobs \fR=\fPnr
127 Set the maximum number of threads/processes to support to \fInr\fR.
128 NOTE: On Linux, it may be necessary to increase the shared-memory limit
129 (`/proc/sys/kernel/shmmax') if fio runs into errors while creating jobs.
131 .BI \-\-server \fR=\fPargs
132 Start a backend server, with \fIargs\fR specifying what to listen to.
133 See \fBCLIENT/SERVER\fR section.
135 .BI \-\-daemonize \fR=\fPpidfile
136 Background a fio server, writing the pid to the given \fIpidfile\fR file.
138 .BI \-\-client \fR=\fPhostname
139 Instead of running the jobs locally, send and run them on the given \fIhostname\fR
140 or set of \fIhostname\fRs. See \fBCLIENT/SERVER\fR section.
142 .BI \-\-remote\-config \fR=\fPfile
143 Tell fio server to load this local \fIfile\fR.
145 .BI \-\-idle\-prof \fR=\fPoption
146 Report CPU idleness. \fIoption\fR is one of the following:
151 Run unit work calibration only and exit.
154 Show aggregate system idleness and unit work.
157 As \fBsystem\fR but also show per CPU idleness.
161 .BI \-\-inflate\-log \fR=\fPlog
162 Inflate and output compressed \fIlog\fR.
164 .BI \-\-trigger\-file \fR=\fPfile
165 Execute trigger command when \fIfile\fR exists.
167 .BI \-\-trigger\-timeout \fR=\fPtime
168 Execute trigger at this \fItime\fR.
170 .BI \-\-trigger \fR=\fPcommand
171 Set this \fIcommand\fR as local trigger.
173 .BI \-\-trigger\-remote \fR=\fPcommand
174 Set this \fIcommand\fR as remote trigger.
176 .BI \-\-aux\-path \fR=\fPpath
177 Use the directory specified by \fIpath\fP for generated state files instead
178 of the current working directory.
179 .SH "JOB FILE FORMAT"
180 Any parameters following the options will be assumed to be job files, unless
181 they match a job file parameter. Multiple job files can be listed and each job
182 file will be regarded as a separate group. Fio will \fBstonewall\fR execution
185 Fio accepts one or more job files describing what it is
186 supposed to do. The job file format is the classic ini file, where the names
187 enclosed in [] brackets define the job name. You are free to use any ASCII name
188 you want, except *global* which has special meaning. Following the job name is
189 a sequence of zero or more parameters, one per line, that define the behavior of
190 the job. If the first character in a line is a ';' or a '#', the entire line is
191 discarded as a comment.
193 A *global* section sets defaults for the jobs described in that file. A job may
194 override a *global* section parameter, and a job file may even have several
195 *global* sections if so desired. A job is only affected by a *global* section
198 The \fB\-\-cmdhelp\fR option also lists all options. If used with an \fIcommand\fR
199 argument, \fB\-\-cmdhelp\fR will detail the given \fIcommand\fR.
201 See the `examples/' directory for inspiration on how to write job files. Note
202 the copyright and license requirements currently apply to
205 Note that the maximum length of a line in the job file is 8192 bytes.
206 .SH "JOB FILE PARAMETERS"
207 Some parameters take an option of a given type, such as an integer or a
208 string. Anywhere a numeric value is required, an arithmetic expression may be
209 used, provided it is surrounded by parentheses. Supported operators are:
216 .B multiplication (*)
222 .B exponentiation (^)
225 For time values in expressions, units are microseconds by default. This is
226 different than for time values not in expressions (not enclosed in
228 .SH "PARAMETER TYPES"
229 The following parameter types are used.
232 String. A sequence of alphanumeric characters.
235 Integer with possible time suffix. Without a unit value is interpreted as
236 seconds unless otherwise specified. Accepts a suffix of 'd' for days, 'h' for
237 hours, 'm' for minutes, 's' for seconds, 'ms' (or 'msec') for milliseconds and 'us'
238 (or 'usec') for microseconds. For example, use 10m for 10 minutes.
241 Integer. A whole number value, which may contain an integer prefix
242 and an integer suffix.
246 [*integer prefix*] **number** [*integer suffix*]
249 The optional *integer prefix* specifies the number's base. The default
250 is decimal. *0x* specifies hexadecimal.
252 The optional *integer suffix* specifies the number's units, and includes an
253 optional unit prefix and an optional unit. For quantities of data, the
254 default unit is bytes. For quantities of time, the default unit is seconds
255 unless otherwise specified.
257 With `kb_base=1000', fio follows international standards for unit
258 prefixes. To specify power-of-10 decimal values defined in the
259 International System of Units (SI):
263 K means kilo (K) or 1000
265 M means mega (M) or 1000**2
267 G means giga (G) or 1000**3
269 T means tera (T) or 1000**4
271 P means peta (P) or 1000**5
275 To specify power-of-2 binary values defined in IEC 80000-13:
279 Ki means kibi (Ki) or 1024
281 Mi means mebi (Mi) or 1024**2
283 Gi means gibi (Gi) or 1024**3
285 Ti means tebi (Ti) or 1024**4
287 Pi means pebi (Pi) or 1024**5
291 For Zone Block Device Mode:
300 With `kb_base=1024' (the default), the unit prefixes are opposite
301 from those specified in the SI and IEC 80000-13 standards to provide
302 compatibility with old scripts. For example, 4k means 4096.
304 For quantities of data, an optional unit of 'B' may be included
305 (e.g., 'kB' is the same as 'k').
307 The *integer suffix* is not case sensitive (e.g., m/mi mean mebi/mega,
308 not milli). 'b' and 'B' both mean byte, not bit.
310 Examples with `kb_base=1000':
314 4 KiB: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB
316 1 MiB: 1048576, 1m, 1024k
318 1 MB: 1000000, 1mi, 1000ki
320 1 TiB: 1073741824, 1t, 1024m, 1048576k
322 1 TB: 1000000000, 1ti, 1000mi, 1000000ki
326 Examples with `kb_base=1024' (default):
330 4 KiB: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB
332 1 MiB: 1048576, 1m, 1024k
334 1 MB: 1000000, 1mi, 1000ki
336 1 TiB: 1073741824, 1t, 1024m, 1048576k
338 1 TB: 1000000000, 1ti, 1000mi, 1000000ki
342 To specify times (units are not case sensitive):
352 s or sec means seconds (default)
354 ms or msec means milliseconds
356 us or usec means microseconds
360 `z' suffix specifies that the value is measured in zones.
361 Value is recalculated once block device's zone size becomes known.
363 If the option accepts an upper and lower range, use a colon ':' or
364 minus '\-' to separate such values. See \fIirange\fR parameter type.
365 If the lower value specified happens to be larger than the upper value
366 the two values are swapped.
370 Boolean. Usually parsed as an integer, however only defined for
371 true and false (1 and 0).
374 Integer range with suffix. Allows value range to be given, such as
375 1024\-4096. A colon may also be used as the separator, e.g. 1k:4k. If the
376 option allows two sets of ranges, they can be specified with a ',' or '/'
377 delimiter: 1k\-4k/8k\-32k. Also see \fIint\fR parameter type.
380 A list of floating point numbers, separated by a ':' character.
382 With the above in mind, here follows the complete list of fio job parameters.
385 .BI kb_base \fR=\fPint
386 Select the interpretation of unit prefixes in input parameters.
391 Inputs comply with IEC 80000-13 and the International
392 System of Units (SI). Use:
396 \- power-of-2 values with IEC prefixes (e.g., KiB)
398 \- power-of-10 values with SI prefixes (e.g., kB)
403 Compatibility mode (default). To avoid breaking old scripts:
407 \- power-of-2 values with SI prefixes
409 \- power-of-10 values with IEC prefixes
414 See \fBbs\fR for more details on input parameters.
416 Outputs always use correct prefixes. Most outputs include both
420 bw=2383.3kB/s (2327.4KiB/s)
423 If only one value is reported, then kb_base selects the one to use:
427 1000 \-\- SI prefixes
429 1024 \-\- IEC prefixes
434 .BI unit_base \fR=\fPint
435 Base unit for reporting. Allowed values are:
440 Use auto-detection (default).
449 .SS "Job description"
452 ASCII name of the job. This may be used to override the name printed by fio
453 for this job. Otherwise the job name is used. On the command line this
454 parameter has the special purpose of also signaling the start of a new job.
456 .BI description \fR=\fPstr
457 Text description of the job. Doesn't do anything except dump this text
458 description when this job is run. It's not parsed.
461 Run the specified number of iterations of this job. Used to repeat the same
462 workload a given number of times. Defaults to 1.
464 .BI numjobs \fR=\fPint
465 Create the specified number of clones of this job. Each clone of job
466 is spawned as an independent thread or process. May be used to setup a
467 larger number of threads/processes doing the same thing. Each thread is
468 reported separately; to see statistics for all clones as a whole, use
469 \fBgroup_reporting\fR in conjunction with \fBnew_group\fR.
470 See \fB\-\-max\-jobs\fR. Default: 1.
471 .SS "Time related parameters"
473 .BI runtime \fR=\fPtime
474 Limit runtime. The test will run until it completes the configured I/O
475 workload or until it has run for this specified amount of time, whichever
476 occurs first. It can be quite hard to determine for how long a specified
477 job will run, so this parameter is handy to cap the total runtime to a
478 given time. When the unit is omitted, the value is interpreted in
482 If set, fio will run for the duration of the \fBruntime\fR specified
483 even if the file(s) are completely read or written. It will simply loop over
484 the same workload as many times as the \fBruntime\fR allows.
486 .BI startdelay \fR=\fPirange(int)
487 Delay the start of job for the specified amount of time. Can be a single
488 value or a range. When given as a range, each thread will choose a value
489 randomly from within the range. Value is in seconds if a unit is omitted.
491 .BI ramp_time \fR=\fPtime
492 If set, fio will run the specified workload for this amount of time before
493 logging any performance numbers. Useful for letting performance settle
494 before logging results, thus minimizing the runtime required for stable
495 results. Note that the \fBramp_time\fR is considered lead in time for a job,
496 thus it will increase the total runtime if a special timeout or
497 \fBruntime\fR is specified. When the unit is omitted, the value is
500 .BI clocksource \fR=\fPstr
501 Use the given clocksource as the base of timing. The supported options are:
506 \fBgettimeofday\fR\|(2)
509 \fBclock_gettime\fR\|(2)
512 Internal CPU clock source
515 \fBcpu\fR is the preferred clocksource if it is reliable, as it is very fast (and
516 fio is heavy on time calls). Fio will automatically use this clocksource if
517 it's supported and considered reliable on the system it is running on,
518 unless another clocksource is specifically set. For x86/x86\-64 CPUs, this
519 means supporting TSC Invariant.
522 .BI gtod_reduce \fR=\fPbool
523 Enable all of the \fBgettimeofday\fR\|(2) reducing options
524 (\fBdisable_clat\fR, \fBdisable_slat\fR, \fBdisable_bw_measurement\fR) plus
525 reduce precision of the timeout somewhat to really shrink the
526 \fBgettimeofday\fR\|(2) call count. With this option enabled, we only do
527 about 0.4% of the \fBgettimeofday\fR\|(2) calls we would have done if all
528 time keeping was enabled.
530 .BI gtod_cpu \fR=\fPint
531 Sometimes it's cheaper to dedicate a single thread of execution to just
532 getting the current time. Fio (and databases, for instance) are very
533 intensive on \fBgettimeofday\fR\|(2) calls. With this option, you can set
534 one CPU aside for doing nothing but logging current time to a shared memory
535 location. Then the other threads/processes that run I/O workloads need only
536 copy that segment, instead of entering the kernel with a
537 \fBgettimeofday\fR\|(2) call. The CPU set aside for doing these time
538 calls will be excluded from other uses. Fio will manually clear it from the
539 CPU mask of other jobs.
540 .SS "Target file/device"
542 .BI directory \fR=\fPstr
543 Prefix \fBfilename\fRs with this directory. Used to place files in a different
544 location than `./'. You can specify a number of directories by
545 separating the names with a ':' character. These directories will be
546 assigned equally distributed to job clones created by \fBnumjobs\fR as
547 long as they are using generated filenames. If specific \fBfilename\fR(s) are
548 set fio will use the first listed directory, and thereby matching the
549 \fBfilename\fR semantic (which generates a file for each clone if not
550 specified, but lets all clones use the same file if set).
553 See the \fBfilename\fR option for information on how to escape ':'
554 characters within the directory path itself.
556 Note: To control the directory fio will use for internal state files
557 use \fB\-\-aux\-path\fR.
560 .BI filename \fR=\fPstr
561 Fio normally makes up a \fBfilename\fR based on the job name, thread number, and
562 file number (see \fBfilename_format\fR). If you want to share files
563 between threads in a job or several
564 jobs with fixed file paths, specify a \fBfilename\fR for each of them to override
565 the default. If the ioengine is file based, you can specify a number of files
566 by separating the names with a ':' colon. So if you wanted a job to open
567 `/dev/sda' and `/dev/sdb' as the two working files, you would use
568 `filename=/dev/sda:/dev/sdb'. This also means that whenever this option is
569 specified, \fBnrfiles\fR is ignored. The size of regular files specified
570 by this option will be \fBsize\fR divided by number of files unless an
571 explicit size is specified by \fBfilesize\fR.
574 Each colon in the wanted path must be escaped with a '\e'
575 character. For instance, if the path is `/dev/dsk/foo@3,0:c' then you
576 would use `filename=/dev/dsk/foo@3,0\\:c' and if the path is
577 `F:\\filename' then you would use `filename=F\\:\\filename'.
579 On Windows, disk devices are accessed as `\\\\.\\PhysicalDrive0' for
580 the first device, `\\\\.\\PhysicalDrive1' for the second etc.
581 Note: Windows and FreeBSD prevent write access to areas
582 of the disk containing in-use data (e.g. filesystems).
584 The filename `\-' is a reserved name, meaning *stdin* or *stdout*. Which
585 of the two depends on the read/write direction set.
588 .BI filename_format \fR=\fPstr
589 If sharing multiple files between jobs, it is usually necessary to have fio
590 generate the exact names that you want. By default, fio will name a file
591 based on the default file format specification of
592 `jobname.jobnumber.filenumber'. With this option, that can be
593 customized. Fio will recognize and replace the following keywords in this
599 The name of the worker thread or process.
602 IP of the fio process when using client/server mode.
605 The incremental number of the worker thread or process.
608 The incremental number of the file for that worker thread or process.
611 To have dependent jobs share a set of files, this option can be set to have
612 fio generate filenames that are shared between the two. For instance, if
613 `testfiles.$filenum' is specified, file number 4 for any job will be
614 named `testfiles.4'. The default of `$jobname.$jobnum.$filenum'
615 will be used if no other format specifier is given.
617 If you specify a path then the directories will be created up to the main
618 directory for the file. So for example if you specify `a/b/c/$jobnum` then the
619 directories a/b/c will be created before the file setup part of the job. If you
620 specify \fBdirectory\fR then the path will be relative that directory, otherwise
621 it is treated as the absolute path.
624 .BI unique_filename \fR=\fPbool
625 To avoid collisions between networked clients, fio defaults to prefixing any
626 generated filenames (with a directory specified) with the source of the
627 client connecting. To disable this behavior, set this option to 0.
629 .BI opendir \fR=\fPstr
630 Recursively open any files below directory \fIstr\fR.
632 .BI lockfile \fR=\fPstr
633 Fio defaults to not locking any files before it does I/O to them. If a file
634 or file descriptor is shared, fio can serialize I/O to that file to make the
635 end result consistent. This is usual for emulating real workloads that share
636 files. The lock modes are:
641 No locking. The default.
644 Only one thread or process may do I/O at a time, excluding all others.
647 Read\-write locking on the file. Many readers may
648 access the file at the same time, but writes get exclusive access.
652 .BI nrfiles \fR=\fPint
653 Number of files to use for this job. Defaults to 1. The size of files
654 will be \fBsize\fR divided by this unless explicit size is specified by
655 \fBfilesize\fR. Files are created for each thread separately, and each
656 file will have a file number within its name by default, as explained in
657 \fBfilename\fR section.
659 .BI openfiles \fR=\fPint
660 Number of files to keep open at the same time. Defaults to the same as
661 \fBnrfiles\fR, can be set smaller to limit the number simultaneous
664 .BI file_service_type \fR=\fPstr
665 Defines how fio decides which file from a job to service next. The following
671 Choose a file at random.
674 Round robin over opened files. This is the default.
677 Finish one file before moving on to the next. Multiple files can
678 still be open depending on \fBopenfiles\fR.
681 Use a Zipf distribution to decide what file to access.
684 Use a Pareto distribution to decide what file to access.
687 Use a Gaussian (normal) distribution to decide what file to access.
693 For \fBrandom\fR, \fBroundrobin\fR, and \fBsequential\fR, a postfix can be appended to
694 tell fio how many I/Os to issue before switching to a new file. For example,
695 specifying `file_service_type=random:8' would cause fio to issue
696 8 I/Os before selecting a new file at random. For the non-uniform
697 distributions, a floating point postfix can be given to influence how the
698 distribution is skewed. See \fBrandom_distribution\fR for a description
699 of how that would work.
702 .BI ioscheduler \fR=\fPstr
703 Attempt to switch the device hosting the file to the specified I/O scheduler
704 before running. If the file is a pipe, a character device file or if device
705 hosting the file could not be determined, this option is ignored.
707 .BI create_serialize \fR=\fPbool
708 If true, serialize the file creation for the jobs. This may be handy to
709 avoid interleaving of data files, which may greatly depend on the filesystem
710 used and even the number of processors in the system. Default: true.
712 .BI create_fsync \fR=\fPbool
713 \fBfsync\fR\|(2) the data file after creation. This is the default.
715 .BI create_on_open \fR=\fPbool
716 If true, don't pre-create files but allow the job's open() to create a file
717 when it's time to do I/O. Default: false \-\- pre-create all necessary files
720 .BI create_only \fR=\fPbool
721 If true, fio will only run the setup phase of the job. If files need to be
722 laid out or updated on disk, only that will be done \-\- the actual job contents
723 are not executed. Default: false.
725 .BI allow_file_create \fR=\fPbool
726 If true, fio is permitted to create files as part of its workload. If this
727 option is false, then fio will error out if
728 the files it needs to use don't already exist. Default: true.
730 .BI allow_mounted_write \fR=\fPbool
731 If this isn't set, fio will abort jobs that are destructive (e.g. that write)
732 to what appears to be a mounted device or partition. This should help catch
733 creating inadvertently destructive tests, not realizing that the test will
734 destroy data on the mounted file system. Note that some platforms don't allow
735 writing against a mounted device regardless of this option. Default: false.
737 .BI pre_read \fR=\fPbool
738 If this is given, files will be pre-read into memory before starting the
739 given I/O operation. This will also clear the \fBinvalidate\fR flag,
740 since it is pointless to pre-read and then drop the cache. This will only
741 work for I/O engines that are seek-able, since they allow you to read the
742 same data multiple times. Thus it will not work on non-seekable I/O engines
743 (e.g. network, splice). Default: false.
745 .BI unlink \fR=\fPbool
746 Unlink the job files when done. Not the default, as repeated runs of that
747 job would then waste time recreating the file set again and again. Default:
750 .BI unlink_each_loop \fR=\fPbool
751 Unlink job files after each iteration or loop. Default: false.
753 .BI zonemode \fR=\fPstr
759 The \fBzonerange\fR, \fBzonesize\fR \fBzonecapacity\fR and \fBzoneskip\fR
760 parameters are ignored.
763 I/O happens in a single zone until \fBzonesize\fR bytes have been transferred.
764 After that number of bytes has been transferred processing of the next zone
765 starts. The \fBzonecapacity\fR parameter is ignored.
768 Zoned block device mode. I/O happens sequentially in each zone, even if random
769 I/O has been selected. Random I/O happens across all zones instead of being
770 restricted to a single zone.
771 Trim is handled using a zone reset operation. Trim only considers non-empty
772 sequential write required and sequential write preferred zones.
776 .BI zonerange \fR=\fPint
777 For \fBzonemode\fR=strided, this is the size of a single zone. See also
778 \fBzonesize\fR and \fBzoneskip\fR.
780 For \fBzonemode\fR=zbd, this parameter is ignored.
782 .BI zonesize \fR=\fPint
783 For \fBzonemode\fR=strided, this is the number of bytes to transfer before
784 skipping \fBzoneskip\fR bytes. If this parameter is smaller than
785 \fBzonerange\fR then only a fraction of each zone with \fBzonerange\fR bytes
786 will be accessed. If this parameter is larger than \fBzonerange\fR then each
787 zone will be accessed multiple times before skipping to the next zone.
789 For \fBzonemode\fR=zbd, this is the size of a single zone. The
790 \fBzonerange\fR parameter is ignored in this mode. For a job accessing a
791 zoned block device, the specified \fBzonesize\fR must be 0 or equal to the
792 device zone size. For a regular block device or file, the specified
793 \fBzonesize\fR must be at least 512B.
795 .BI zonecapacity \fR=\fPint
796 For \fBzonemode\fR=zbd, this defines the capacity of a single zone, which is
797 the accessible area starting from the zone start address. This parameter only
798 applies when using \fBzonemode\fR=zbd in combination with regular block devices.
799 If not specified it defaults to the zone size. If the target device is a zoned
800 block device, the zone capacity is obtained from the device information and this
803 .BI zoneskip \fR=\fPint[z]
804 For \fBzonemode\fR=strided, the number of bytes to skip after \fBzonesize\fR
805 bytes of data have been transferred.
807 For \fBzonemode\fR=zbd, the \fBzonesize\fR aligned number of bytes to skip
808 once a zone is fully written (write workloads) or all written data in the
809 zone have been read (read workloads). This parameter is valid only for
810 sequential workloads and ignored for random workloads. For read workloads,
811 see also \fBread_beyond_wp\fR.
814 .BI read_beyond_wp \fR=\fPbool
815 This parameter applies to \fBzonemode=zbd\fR only.
817 Zoned block devices are block devices that consist of multiple zones. Each
818 zone has a type, e.g. conventional or sequential. A conventional zone can be
819 written at any offset that is a multiple of the block size. Sequential zones
820 must be written sequentially. The position at which a write must occur is
821 called the write pointer. A zoned block device can be either host managed or
822 host aware. For host managed devices the host must ensure that writes happen
823 sequentially. Fio recognizes host managed devices and serializes writes to
824 sequential zones for these devices.
826 If a read occurs in a sequential zone beyond the write pointer then the zoned
827 block device will complete the read without reading any data from the storage
828 medium. Since such reads lead to unrealistically high bandwidth and IOPS
829 numbers fio only reads beyond the write pointer if explicitly told to do
832 .BI max_open_zones \fR=\fPint
833 A zone of a zoned block device is in the open state when it is partially written
834 (i.e. not all sectors of the zone have been written). Zoned block devices may
835 have limit a on the total number of zones that can be simultaneously in the
836 open state, that is, the number of zones that can be written to simultaneously.
837 The \fBmax_open_zones\fR parameter limits the number of zones to which write
838 commands are issued by all fio jobs, that is, limits the number of zones that
839 will be in the open state. This parameter is relevant only if the
840 \fBzonemode=zbd\fR is used. The default value is always equal to maximum number
841 of open zones of the target zoned block device and a value higher than this
842 limit cannot be specified by users unless the option \fBignore_zone_limits\fR is
843 specified. When \fBignore_zone_limits\fR is specified or the target device has
844 no limit on the number of zones that can be in an open state,
845 \fBmax_open_zones\fR can specify 0 to disable any limit on the number of zones
846 that can be simultaneously written to by all jobs.
848 .BI job_max_open_zones \fR=\fPint
849 In the same manner as \fBmax_open_zones\fR, limit the number of open zones per
850 fio job, that is, the number of zones that a single job can simultaneously write
851 to. A value of zero indicates no limit. Default: zero.
853 .BI ignore_zone_limits \fR=\fPbool
854 If this option is used, fio will ignore the maximum number of open zones limit
855 of the zoned block device in use, thus allowing the option \fBmax_open_zones\fR
856 value to be larger than the device reported limit. Default: false.
858 .BI zone_reset_threshold \fR=\fPfloat
859 A number between zero and one that indicates the ratio of written bytes in the
860 zones with write pointers in the IO range to the size of the IO range. When
861 current ratio is above this ratio, zones are reset periodically as
862 \fBzone_reset_frequency\fR specifies. If there are multiple jobs when using this
863 option, the IO range for all write jobs has to be the same.
865 .BI zone_reset_frequency \fR=\fPfloat
866 A number between zero and one that indicates how often a zone reset should be
867 issued if the zone reset threshold has been exceeded. A zone reset is
868 submitted after each (1 / zone_reset_frequency) write requests. This and the
869 previous parameter can be used to simulate garbage collection activity.
873 .BI direct \fR=\fPbool
874 If value is true, use non-buffered I/O. This is usually O_DIRECT. Note that
875 OpenBSD and ZFS on Solaris don't support direct I/O. On Windows the synchronous
876 ioengines don't support direct I/O. Default: false.
878 .BI buffered \fR=\fPbool
879 If value is true, use buffered I/O. This is the opposite of the
880 \fBdirect\fR option. Defaults to true.
882 .BI readwrite \fR=\fPstr "\fR,\fP rw" \fR=\fPstr
883 Type of I/O pattern. Accepted values are:
894 Sequential trims (Linux block devices and SCSI character devices only).
903 Random trims (Linux block devices and SCSI character devices only).
906 Sequential mixed reads and writes.
909 Random mixed reads and writes.
912 Sequential trim+write sequences. Blocks will be trimmed first,
913 then the same blocks will be written to. So if `io_size=64K' is specified,
914 Fio will trim a total of 64K bytes and also write 64K bytes on the same
915 trimmed blocks. This behaviour will be consistent with `number_ios' or
916 other Fio options limiting the total bytes or number of I/O's.
921 but uses random offsets rather than sequential writes.
924 Fio defaults to read if the option is not specified. For the mixed I/O
925 types, the default is to split them 50/50. For certain types of I/O the
926 result may still be skewed a bit, since the speed may be different.
928 It is possible to specify the number of I/Os to do before getting a new
929 offset by appending `:<nr>' to the end of the string given. For a
930 random read, it would look like `rw=randread:8' for passing in an offset
931 modifier with a value of 8. If the suffix is used with a sequential I/O
932 pattern, then the `<nr>' value specified will be added to the generated
933 offset for each I/O turning sequential I/O into sequential I/O with holes.
934 For instance, using `rw=write:4k' will skip 4k for every write. Also see
935 the \fBrw_sequencer\fR option.
938 .BI rw_sequencer \fR=\fPstr
939 If an offset modifier is given by appending a number to the `rw=\fIstr\fR'
940 line, then this option controls how that number modifies the I/O offset
941 being generated. Accepted values are:
946 Generate sequential offset.
949 Generate the same offset.
952 \fBsequential\fR is only useful for random I/O, where fio would normally
953 generate a new random offset for every I/O. If you append e.g. 8 to randread,
954 i.e. `rw=randread:8' you would get a new random offset for every 8 I/Os. The
955 result would be a sequence of 8 sequential offsets with a random starting
956 point. However this behavior may change if a sequential I/O reaches end of the
957 file. As sequential I/O is already sequential, setting \fBsequential\fR for
958 that would not result in any difference. \fBidentical\fR behaves in a similar
959 fashion, except it sends the same offset 8 number of times before generating a
969 rw_sequencer=sequential
975 The generated sequence of offsets will look like this:
976 4k, 8k, 12k, 16k, 20k, 24k, 28k, 32k, 92k, 96k, 100k, 104k, 108k, 112k, 116k,
986 rw_sequencer=identical
992 The generated sequence of offsets will look like this:
993 4k, 4k, 4k, 4k, 4k, 4k, 4k, 4k, 92k, 92k, 92k, 92k, 92k, 92k, 92k, 92k, 48k,
997 .BI unified_rw_reporting \fR=\fPstr
998 Fio normally reports statistics on a per data direction basis, meaning that
999 reads, writes, and trims are accounted and reported separately. This option
1000 determines whether fio reports the results normally, summed together, or as
1002 Accepted values are:
1006 Normal statistics reporting.
1009 Statistics are summed per data direction and reported together.
1012 Statistics are reported normally, followed by the mixed statistics.
1015 Backward-compatible alias for \fBnone\fR.
1018 Backward-compatible alias for \fBmixed\fR.
1021 Alias for \fBboth\fR.
1024 .BI randrepeat \fR=\fPbool
1025 Seed the random number generator used for random I/O patterns in a
1026 predictable way so the pattern is repeatable across runs. Default: true.
1028 .BI allrandrepeat \fR=\fPbool
1029 Seed all random number generators in a predictable way so results are
1030 repeatable across runs. Default: false.
1032 .BI randseed \fR=\fPint
1033 Seed the random number generators based on this seed value, to be able to
1034 control what sequence of output is being generated. If not set, the random
1035 sequence depends on the \fBrandrepeat\fR setting.
1037 .BI fallocate \fR=\fPstr
1038 Whether pre-allocation is performed when laying down files.
1039 Accepted values are:
1044 Do not pre-allocate space.
1047 Use a platform's native pre-allocation call but fall back to
1048 \fBnone\fR behavior if it fails/is not implemented.
1051 Pre-allocate via \fBposix_fallocate\fR\|(3).
1054 Pre-allocate via \fBfallocate\fR\|(2) with
1055 FALLOC_FL_KEEP_SIZE set.
1058 Extend file to final size using \fBftruncate\fR|(2)
1059 instead of allocating.
1062 Backward-compatible alias for \fBnone\fR.
1065 Backward-compatible alias for \fBposix\fR.
1068 May not be available on all supported platforms. \fBkeep\fR is only available
1069 on Linux. If using ZFS on Solaris this cannot be set to \fBposix\fR
1070 because ZFS doesn't support pre-allocation. Default: \fBnative\fR if any
1071 pre-allocation methods except \fBtruncate\fR are available, \fBnone\fR if not.
1073 Note that using \fBtruncate\fR on Windows will interact surprisingly
1074 with non-sequential write patterns. When writing to a file that has
1075 been extended by setting the end-of-file information, Windows will
1076 backfill the unwritten portion of the file up to that offset with
1077 zeroes before issuing the new write. This means that a single small
1078 write to the end of an extended file will stall until the entire
1079 file has been filled with zeroes.
1082 .BI fadvise_hint \fR=\fPstr
1083 Use \fBposix_fadvise\fR\|(2) or \fBposix_madvise\fR\|(2) to advise the kernel
1084 what I/O patterns are likely to be issued. Accepted values are:
1089 Backwards compatible hint for "no hint".
1092 Backwards compatible hint for "advise with fio workload type". This
1093 uses FADV_RANDOM for a random workload, and FADV_SEQUENTIAL
1094 for a sequential workload.
1097 Advise using FADV_SEQUENTIAL.
1100 Advise using FADV_RANDOM.
1103 Advise using FADV_NOREUSE. This may be a no-op on older Linux
1104 kernels. Since Linux 6.3, it provides a hint to the LRU algorithm.
1105 See the \fBposix_fadvise\fR\|(2) man page.
1109 .BI write_hint \fR=\fPstr
1110 Use \fBfcntl\fR\|(2) to advise the kernel what life time to expect
1111 from a write. Only supported on Linux, as of version 4.13. Accepted
1117 No particular life time associated with this file.
1120 Data written to this file has a short life time.
1123 Data written to this file has a medium life time.
1126 Data written to this file has a long life time.
1129 Data written to this file has a very long life time.
1132 The values are all relative to each other, and no absolute meaning
1133 should be associated with them.
1136 .BI offset \fR=\fPint[%|z]
1137 Start I/O at the provided offset in the file, given as either a fixed size in
1138 bytes, zones or a percentage. If a percentage is given, the generated offset will be
1139 aligned to the minimum \fBblocksize\fR or to the value of \fBoffset_align\fR if
1140 provided. Data before the given offset will not be touched. This
1141 effectively caps the file size at `real_size \- offset'. Can be combined with
1142 \fBsize\fR to constrain the start and end range of the I/O workload.
1143 A percentage can be specified by a number between 1 and 100 followed by '%',
1144 for example, `offset=20%' to specify 20%. In ZBD mode, value can be set as
1145 number of zones using 'z'.
1147 .BI offset_align \fR=\fPint
1148 If set to non-zero value, the byte offset generated by a percentage \fBoffset\fR
1149 is aligned upwards to this value. Defaults to 0 meaning that a percentage
1150 offset is aligned to the minimum block size.
1152 .BI offset_increment \fR=\fPint[%|z]
1153 If this is provided, then the real offset becomes `\fBoffset\fR + \fBoffset_increment\fR
1154 * thread_number', where the thread number is a counter that starts at 0 and
1155 is incremented for each sub-job (i.e. when \fBnumjobs\fR option is
1156 specified). This option is useful if there are several jobs which are
1157 intended to operate on a file in parallel disjoint segments, with even
1158 spacing between the starting points. Percentages can be used for this option.
1159 If a percentage is given, the generated offset will be aligned to the minimum
1160 \fBblocksize\fR or to the value of \fBoffset_align\fR if provided.In ZBD mode, value
1161 can be set as number of zones using 'z'.
1163 .BI number_ios \fR=\fPint
1164 Fio will normally perform I/Os until it has exhausted the size of the region
1165 set by \fBsize\fR, or if it exhaust the allocated time (or hits an error
1166 condition). With this setting, the range/size can be set independently of
1167 the number of I/Os to perform. When fio reaches this number, it will exit
1168 normally and report status. Note that this does not extend the amount of I/O
1169 that will be done, it will only stop fio if this condition is met before
1170 other end-of-job criteria.
1172 .BI fsync \fR=\fPint
1173 If writing to a file, issue an \fBfsync\fR\|(2) (or its equivalent) of
1174 the dirty data for every number of blocks given. For example, if you give 32
1175 as a parameter, fio will sync the file after every 32 writes issued. If fio is
1176 using non-buffered I/O, we may not sync the file. The exception is the sg
1177 I/O engine, which synchronizes the disk cache anyway. Defaults to 0, which
1178 means fio does not periodically issue and wait for a sync to complete. Also
1179 see \fBend_fsync\fR and \fBfsync_on_close\fR.
1181 .BI fdatasync \fR=\fPint
1182 Like \fBfsync\fR but uses \fBfdatasync\fR\|(2) to only sync data and
1183 not metadata blocks. In Windows, DragonFlyBSD or OSX there is no
1184 \fBfdatasync\fR\|(2) so this falls back to using \fBfsync\fR\|(2).
1185 Defaults to 0, which means fio does not periodically issue and wait for a
1186 data-only sync to complete.
1188 .BI write_barrier \fR=\fPint
1189 Make every N\-th write a barrier write.
1191 .BI sync_file_range \fR=\fPstr:int
1192 Use \fBsync_file_range\fR\|(2) for every \fIint\fR number of write
1193 operations. Fio will track range of writes that have happened since the last
1194 \fBsync_file_range\fR\|(2) call. \fIstr\fR can currently be one or more of:
1199 SYNC_FILE_RANGE_WAIT_BEFORE
1202 SYNC_FILE_RANGE_WRITE
1205 SYNC_FILE_RANGE_WRITE_AFTER
1208 So if you do `sync_file_range=wait_before,write:8', fio would use
1209 `SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE' for every 8
1210 writes. Also see the \fBsync_file_range\fR\|(2) man page. This option is
1214 .BI overwrite \fR=\fPbool
1215 If true, writes to a file will always overwrite existing data. If the file
1216 doesn't already exist, it will be created before the write phase begins. If
1217 the file exists and is large enough for the specified write phase, nothing
1218 will be done. Default: false.
1220 .BI end_fsync \fR=\fPbool
1221 If true, \fBfsync\fR\|(2) file contents when a write stage has completed.
1224 .BI fsync_on_close \fR=\fPbool
1225 If true, fio will \fBfsync\fR\|(2) a dirty file on close. This differs
1226 from \fBend_fsync\fR in that it will happen on every file close, not
1227 just at the end of the job. Default: false.
1229 .BI rwmixread \fR=\fPint
1230 Percentage of a mixed workload that should be reads. Default: 50.
1232 .BI rwmixwrite \fR=\fPint
1233 Percentage of a mixed workload that should be writes. If both
1234 \fBrwmixread\fR and \fBrwmixwrite\fR is given and the values do not
1235 add up to 100%, the latter of the two will be used to override the
1236 first. This may interfere with a given rate setting, if fio is asked to
1237 limit reads or writes to a certain rate. If that is the case, then the
1238 distribution may be skewed. Default: 50.
1240 .BI random_distribution \fR=\fPstr:float[:float][,str:float][,str:float]
1241 By default, fio will use a completely uniform random distribution when asked
1242 to perform random I/O. Sometimes it is useful to skew the distribution in
1243 specific ways, ensuring that some parts of the data is more hot than others.
1244 fio includes the following distribution models:
1249 Uniform random distribution
1258 Normal (Gaussian) distribution
1261 Zoned random distribution
1263 Zoned absolute random distribution
1266 When using a \fBzipf\fR or \fBpareto\fR distribution, an input value is also
1267 needed to define the access pattern. For \fBzipf\fR, this is the `Zipf theta'.
1268 For \fBpareto\fR, it's the `Pareto power'. Fio includes a test
1269 program, \fBfio\-genzipf\fR, that can be used visualize what the given input
1270 values will yield in terms of hit rates. If you wanted to use \fBzipf\fR with
1271 a `theta' of 1.2, you would use `random_distribution=zipf:1.2' as the
1272 option. If a non\-uniform model is used, fio will disable use of the random
1273 map. For the \fBnormal\fR distribution, a normal (Gaussian) deviation is
1274 supplied as a value between 0 and 100.
1276 The second, optional float is allowed for \fBpareto\fR, \fBzipf\fR and \fBnormal\fR
1277 distributions. It allows one to set base of distribution in non-default place, giving
1278 more control over most probable outcome. This value is in range [0-1] which maps linearly to
1279 range of possible random values.
1280 Defaults are: random for \fBpareto\fR and \fBzipf\fR, and 0.5 for \fBnormal\fR.
1281 If you wanted to use \fBzipf\fR with a `theta` of 1.2 centered on 1/4 of allowed value range,
1282 you would use `random_distribution=zipf:1.2:0.25`.
1284 For a \fBzoned\fR distribution, fio supports specifying percentages of I/O
1285 access that should fall within what range of the file or device. For
1286 example, given a criteria of:
1290 60% of accesses should be to the first 10%
1292 30% of accesses should be to the next 20%
1294 8% of accesses should be to the next 30%
1296 2% of accesses should be to the next 40%
1300 we can define that through zoning of the random accesses. For the above
1301 example, the user would do:
1304 random_distribution=zoned:60/10:30/20:8/30:2/40
1307 A \fBzoned_abs\fR distribution works exactly like the\fBzoned\fR, except that
1308 it takes absolute sizes. For example, let's say you wanted to define access
1309 according to the following criteria:
1313 60% of accesses should be to the first 20G
1315 30% of accesses should be to the next 100G
1317 10% of accesses should be to the next 500G
1321 we can define an absolute zoning distribution with:
1324 random_distribution=zoned:60/10:30/20:8/30:2/40
1327 For both \fBzoned\fR and \fBzoned_abs\fR, fio supports defining up to 256
1330 Similarly to how \fBbssplit\fR works for setting ranges and percentages
1331 of block sizes. Like \fBbssplit\fR, it's possible to specify separate
1332 zones for reads, writes, and trims. If just one set is given, it'll apply to
1336 .BI percentage_random \fR=\fPint[,int][,int]
1337 For a random workload, set how big a percentage should be random. This
1338 defaults to 100%, in which case the workload is fully random. It can be set
1339 from anywhere from 0 to 100. Setting it to 0 would make the workload fully
1340 sequential. Any setting in between will result in a random mix of sequential
1341 and random I/O, at the given percentages. Comma-separated values may be
1342 specified for reads, writes, and trims as described in \fBblocksize\fR.
1345 Normally fio will cover every block of the file when doing random I/O. If
1346 this option is given, fio will just get a new random offset without looking
1347 at past I/O history. This means that some blocks may not be read or written,
1348 and that some blocks may be read/written more than once. If this option is
1349 used with \fBverify\fR and multiple blocksizes (via \fBbsrange\fR),
1350 only intact blocks are verified, i.e., partially-overwritten blocks are
1351 ignored. With an async I/O engine and an I/O depth > 1, it is possible for
1352 the same block to be overwritten, which can cause verification errors. Either
1353 do not use norandommap in this case, or also use the lfsr random generator.
1355 .BI softrandommap \fR=\fPbool
1356 See \fBnorandommap\fR. If fio runs with the random block map enabled and
1357 it fails to allocate the map, if this option is set it will continue without
1358 a random block map. As coverage will not be as complete as with random maps,
1359 this option is disabled by default.
1361 .BI random_generator \fR=\fPstr
1362 Fio supports the following engines for generating I/O offsets for random I/O:
1367 Strong 2^88 cycle random number generator.
1370 Linear feedback shift register generator.
1373 Strong 64\-bit 2^258 cycle random number generator.
1376 \fBtausworthe\fR is a strong random number generator, but it requires tracking
1377 on the side if we want to ensure that blocks are only read or written
1378 once. \fBlfsr\fR guarantees that we never generate the same offset twice, and
1379 it's also less computationally expensive. It's not a true random generator,
1380 however, though for I/O purposes it's typically good enough. \fBlfsr\fR only
1381 works with single block sizes, not with workloads that use multiple block
1382 sizes. If used with such a workload, fio may read or write some blocks
1383 multiple times. The default value is \fBtausworthe\fR, unless the required
1384 space exceeds 2^32 blocks. If it does, then \fBtausworthe64\fR is
1385 selected automatically.
1389 .BI blocksize \fR=\fPint[,int][,int] "\fR,\fB bs" \fR=\fPint[,int][,int]
1390 The block size in bytes used for I/O units. Default: 4096. A single value
1391 applies to reads, writes, and trims. Comma-separated values may be
1392 specified for reads, writes, and trims. A value not terminated in a comma
1393 applies to subsequent types. Examples:
1398 bs=256k means 256k for reads, writes and trims.
1400 bs=8k,32k means 8k for reads, 32k for writes and trims.
1402 bs=8k,32k, means 8k for reads, 32k for writes, and default for trims.
1404 bs=,8k means default for reads, 8k for writes and trims.
1406 bs=,8k, means default for reads, 8k for writes, and default for trims.
1411 .BI blocksize_range \fR=\fPirange[,irange][,irange] "\fR,\fB bsrange" \fR=\fPirange[,irange][,irange]
1412 A range of block sizes in bytes for I/O units. The issued I/O unit will
1413 always be a multiple of the minimum size, unless
1414 \fBblocksize_unaligned\fR is set.
1415 Comma-separated ranges may be specified for reads, writes, and trims as
1416 described in \fBblocksize\fR. Example:
1420 bsrange=1k\-4k,2k\-8k
1424 .BI bssplit \fR=\fPstr[,str][,str]
1425 Sometimes you want even finer grained control of the block sizes issued, not
1426 just an even split between them. This option allows you to weight various
1427 block sizes, so that you are able to define a specific amount of block sizes
1428 issued. The format for this option is:
1432 bssplit=blocksize/percentage:blocksize/percentage
1435 for as many block sizes as needed. So if you want to define a workload that
1436 has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would write:
1439 bssplit=4k/10:64k/50:32k/40
1442 Ordering does not matter. If the percentage is left blank, fio will fill in
1443 the remaining values evenly. So a bssplit option like this one:
1446 bssplit=4k/50:1k/:32k/
1449 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always add up
1450 to 100, if bssplit is given a range that adds up to more, it will error out.
1452 Comma-separated values may be specified for reads, writes, and trims as
1453 described in \fBblocksize\fR.
1455 If you want a workload that has 50% 2k reads and 50% 4k reads, while having
1456 90% 4k writes and 10% 8k writes, you would specify:
1459 bssplit=2k/50:4k/50,4k/90:8k/10
1462 Fio supports defining up to 64 different weights for each data direction.
1465 .BI blocksize_unaligned "\fR,\fB bs_unaligned"
1466 If set, fio will issue I/O units with any size within
1467 \fBblocksize_range\fR, not just multiples of the minimum size. This
1468 typically won't work with direct I/O, as that normally requires sector
1471 .BI bs_is_seq_rand \fR=\fPbool
1472 If this option is set, fio will use the normal read,write blocksize settings
1473 as sequential,random blocksize settings instead. Any random read or write
1474 will use the WRITE blocksize settings, and any sequential read or write will
1475 use the READ blocksize settings.
1477 .BI blockalign \fR=\fPint[,int][,int] "\fR,\fB ba" \fR=\fPint[,int][,int]
1478 Boundary to which fio will align random I/O units. Default:
1479 \fBblocksize\fR. Minimum alignment is typically 512b for using direct
1480 I/O, though it usually depends on the hardware block size. This option is
1481 mutually exclusive with using a random map for files, so it will turn off
1482 that option. Comma-separated values may be specified for reads, writes, and
1483 trims as described in \fBblocksize\fR.
1484 .SS "Buffers and memory"
1487 Initialize buffers with all zeros. Default: fill buffers with random data.
1490 If this option is given, fio will refill the I/O buffers on every
1491 submit. The default is to only fill it at init time and reuse that
1492 data. Only makes sense if zero_buffers isn't specified, naturally. If data
1493 verification is enabled, \fBrefill_buffers\fR is also automatically enabled.
1495 .BI scramble_buffers \fR=\fPbool
1496 If \fBrefill_buffers\fR is too costly and the target is using data
1497 deduplication, then setting this option will slightly modify the I/O buffer
1498 contents to defeat normal de-dupe attempts. This is not enough to defeat
1499 more clever block compression attempts, but it will stop naive dedupe of
1500 blocks. Default: true.
1502 .BI buffer_compress_percentage \fR=\fPint
1503 If this is set, then fio will attempt to provide I/O buffer content
1504 (on WRITEs) that compresses to the specified level. Fio does this by
1505 providing a mix of random data followed by fixed pattern data. The
1506 fixed pattern is either zeros, or the pattern specified by
1507 \fBbuffer_pattern\fR. If the \fBbuffer_pattern\fR option is used, it
1508 might skew the compression ratio slightly. Setting
1509 \fBbuffer_compress_percentage\fR to a value other than 100 will also
1510 enable \fBrefill_buffers\fR in order to reduce the likelihood that
1511 adjacent blocks are so similar that they over compress when seen
1512 together. See \fBbuffer_compress_chunk\fR for how to set a finer or
1513 coarser granularity of the random/fixed data regions. Defaults to unset
1514 i.e., buffer data will not adhere to any compression level.
1516 .BI buffer_compress_chunk \fR=\fPint
1517 This setting allows fio to manage how big the random/fixed data region
1518 is when using \fBbuffer_compress_percentage\fR. When
1519 \fBbuffer_compress_chunk\fR is set to some non-zero value smaller than the
1520 block size, fio can repeat the random/fixed region throughout the I/O
1521 buffer at the specified interval (which particularly useful when
1522 bigger block sizes are used for a job). When set to 0, fio will use a
1523 chunk size that matches the block size resulting in a single
1524 random/fixed region within the I/O buffer. Defaults to 512. When the
1525 unit is omitted, the value is interpreted in bytes.
1527 .BI buffer_pattern \fR=\fPstr
1528 If set, fio will fill the I/O buffers with this pattern or with the contents
1529 of a file. If not set, the contents of I/O buffers are defined by the other
1530 options related to buffer contents. The setting can be any pattern of bytes,
1531 and can be prefixed with 0x for hex values. It may also be a string, where
1532 the string must then be wrapped with "". Or it may also be a filename,
1533 where the filename must be wrapped with '' in which case the file is
1534 opened and read. Note that not all the file contents will be read if that
1535 would cause the buffers to overflow. So, for example:
1540 buffer_pattern='filename'
1544 buffer_pattern="abcd"
1552 buffer_pattern=0xdeadface
1556 Also you can combine everything together in any order:
1559 buffer_pattern=0xdeadface"abcd"\-12'filename'
1563 .BI dedupe_percentage \fR=\fPint
1564 If set, fio will generate this percentage of identical buffers when
1565 writing. These buffers will be naturally dedupable. The contents of the
1566 buffers depend on what other buffer compression settings have been set. It's
1567 possible to have the individual buffers either fully compressible, or not at
1568 all \-\- this option only controls the distribution of unique buffers. Setting
1569 this option will also enable \fBrefill_buffers\fR to prevent every buffer
1572 .BI dedupe_mode \fR=\fPstr
1573 If \fBdedupe_percentage\fR is given, then this option controls how fio
1574 generates the dedupe buffers.
1581 Generate dedupe buffers by repeating previous writes
1587 Generate dedupe buffers from working set
1591 \fBrepeat\fR is the default option for fio. Dedupe buffers are generated
1592 by repeating previous unique write.
1594 \fBworking_set\fR is a more realistic workload.
1595 With \fBworking_set\fR, \fBdedupe_working_set_percentage\fR should be provided.
1596 Given that, fio will use the initial unique write buffers as its working set.
1597 Upon deciding to dedupe, fio will randomly choose a buffer from the working set.
1598 Note that by using \fBworking_set\fR the dedupe percentage will converge
1599 to the desired over time while \fBrepeat\fR maintains the desired percentage
1604 .BI dedupe_working_set_percentage \fR=\fPint
1605 If \fBdedupe_mode\fR is set to \fBworking_set\fR, then this controls
1606 the percentage of size of the file or device used as the buffers
1607 fio will choose to generate the dedupe buffers from
1610 Note that \fBsize\fR needs to be explicitly provided and only 1 file
1611 per job is supported
1614 .BI dedupe_global \fR=\fPbool
1615 This controls whether the deduplication buffers will be shared amongst
1616 all jobs that have this option set. The buffers are spread evenly between
1620 Note that \fBdedupe_mode\fR must be set to \fBworking_set\fR for this to work.
1621 Can be used in combination with compression
1623 .BI invalidate \fR=\fPbool
1624 Invalidate the buffer/page cache parts of the files to be used prior to
1625 starting I/O if the platform and file type support it. Defaults to true.
1626 This will be ignored if \fBpre_read\fR is also specified for the
1630 Whether, and what type, of synchronous I/O to use for writes. The allowed
1636 Do not use synchronous IO, the default.
1642 Use synchronous file IO. For the majority of I/O engines,
1643 this means using O_SYNC.
1649 Use synchronous data IO. For the majority of I/O engines,
1650 this means using O_DSYNC.
1655 .BI iomem \fR=\fPstr "\fR,\fP mem" \fR=\fPstr
1656 Fio can use various types of memory as the I/O unit buffer. The allowed
1662 Use memory from \fBmalloc\fR\|(3) as the buffers. Default memory type.
1665 Use shared memory as the buffers. Allocated through \fBshmget\fR\|(2).
1668 Same as \fBshm\fR, but use huge pages as backing.
1671 Use \fBmmap\fR\|(2) to allocate buffers. May either be anonymous memory, or can
1672 be file backed if a filename is given after the option. The format
1673 is `mem=mmap:/path/to/file'.
1676 Use a memory mapped huge file as the buffer backing. Append filename
1677 after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file'.
1680 Same as \fBmmap\fR, but use a MMAP_SHARED mapping.
1683 Use GPU memory as the buffers for GPUDirect RDMA benchmark.
1684 The \fBioengine\fR must be \fBrdma\fR.
1687 The area allocated is a function of the maximum allowed bs size for the job,
1688 multiplied by the I/O depth given. Note that for \fBshmhuge\fR and
1689 \fBmmaphuge\fR to work, the system must have free huge pages allocated. This
1690 can normally be checked and set by reading/writing
1691 `/proc/sys/vm/nr_hugepages' on a Linux system. Fio assumes a huge page
1692 is 2 or 4MiB in size depending on the platform. So to calculate the number of
1693 huge pages you need for a given job file, add up the I/O depth of all jobs
1694 (normally one unless \fBiodepth\fR is used) and multiply by the maximum bs set.
1695 Then divide that number by the huge page size. You can see the size of the huge
1696 pages in `/proc/meminfo'. If no huge pages are allocated by having a non-zero
1697 number in `nr_hugepages', using \fBmmaphuge\fR or \fBshmhuge\fR will fail. Also
1698 see \fBhugepage\-size\fR.
1700 \fBmmaphuge\fR also needs to have hugetlbfs mounted and the file location
1701 should point there. So if it's mounted in `/huge', you would use
1702 `mem=mmaphuge:/huge/somefile'.
1705 .BI iomem_align \fR=\fPint "\fR,\fP mem_align" \fR=\fPint
1706 This indicates the memory alignment of the I/O memory buffers. Note that
1707 the given alignment is applied to the first I/O unit buffer, if using
1708 \fBiodepth\fR the alignment of the following buffers are given by the
1709 \fBbs\fR used. In other words, if using a \fBbs\fR that is a
1710 multiple of the page sized in the system, all buffers will be aligned to
1711 this value. If using a \fBbs\fR that is not page aligned, the alignment
1712 of subsequent I/O memory buffers is the sum of the \fBiomem_align\fR and
1715 .BI hugepage\-size \fR=\fPint
1716 Defines the size of a huge page. Must at least be equal to the system setting,
1717 see `/proc/meminfo' and `/sys/kernel/mm/hugepages/'. Defaults to 2 or 4MiB
1718 depending on the platform. Should probably always be a multiple of megabytes,
1719 so using `hugepage\-size=Xm' is the preferred way to set this to avoid setting
1720 a non-pow-2 bad value.
1722 .BI lockmem \fR=\fPint
1723 Pin the specified amount of memory with \fBmlock\fR\|(2). Can be used to
1724 simulate a smaller amount of memory. The amount specified is per worker.
1727 .BI size \fR=\fPint[%|z]
1728 The total size of file I/O for each thread of this job. Fio will run until
1729 this many bytes has been transferred, unless runtime is altered by other means
1730 such as (1) \fBruntime\fR, (2) \fBio_size\fR, (3) \fBnumber_ios\fR, (4)
1731 gaps/holes while doing I/O's such as `rw=read:16K', or (5) sequential I/O
1732 reaching end of the file which is possible when \fBpercentage_random\fR is
1734 Fio will divide this size between the available files determined by options
1735 such as \fBnrfiles\fR, \fBfilename\fR, unless \fBfilesize\fR is
1736 specified by the job. If the result of division happens to be 0, the size is
1737 set to the physical size of the given files or devices if they exist.
1738 If this option is not specified, fio will use the full size of the given
1739 files or devices. If the files do not exist, size must be given. It is also
1740 possible to give size as a percentage between 1 and 100. If `size=20%' is
1741 given, fio will use 20% of the full size of the given files or devices. In ZBD mode,
1742 size can be given in units of number of zones using 'z'. Can be combined with \fBoffset\fR to
1743 constrain the start and end range that I/O will be done within.
1745 .BI io_size \fR=\fPint[%|z] "\fR,\fB io_limit" \fR=\fPint[%|z]
1746 Normally fio operates within the region set by \fBsize\fR, which means
1747 that the \fBsize\fR option sets both the region and size of I/O to be
1748 performed. Sometimes that is not what you want. With this option, it is
1749 possible to define just the amount of I/O that fio should do. For instance,
1750 if \fBsize\fR is set to 20GiB and \fBio_size\fR is set to 5GiB, fio
1751 will perform I/O within the first 20GiB but exit when 5GiB have been
1752 done. The opposite is also possible \-\- if \fBsize\fR is set to 20GiB,
1753 and \fBio_size\fR is set to 40GiB, then fio will do 40GiB of I/O within
1754 the 0..20GiB region. Value can be set as percentage: \fBio_size\fR=N%.
1755 In this case \fBio_size\fR multiplies \fBsize\fR= value. In ZBD mode, value can
1756 also be set as number of zones using 'z'.
1758 .BI filesize \fR=\fPirange(int)
1759 Individual file sizes. May be a range, in which case fio will select sizes
1760 for files at random within the given range. If not given, each created file
1761 is the same size. This option overrides \fBsize\fR in terms of file size,
1762 i.e. \fBsize\fR becomes merely the default for \fBio_size\fR (and
1763 has no effect it all if \fBio_size\fR is set explicitly).
1765 .BI file_append \fR=\fPbool
1766 Perform I/O after the end of the file. Normally fio will operate within the
1767 size of a file. If this option is set, then fio will append to the file
1768 instead. This has identical behavior to setting \fBoffset\fR to the size
1769 of a file. This option is ignored on non-regular files.
1771 .BI fill_device \fR=\fPbool "\fR,\fB fill_fs" \fR=\fPbool
1772 Sets size to something really large and waits for ENOSPC (no space left on
1773 device) or EDQUOT (disk quota exceeded)
1774 as the terminating condition. Only makes sense with sequential
1775 write. For a read workload, the mount point will be filled first then I/O
1776 started on the result.
1779 .BI ioengine \fR=\fPstr
1780 Defines how the job issues I/O to the file. The following types are defined:
1785 Basic \fBread\fR\|(2) or \fBwrite\fR\|(2)
1786 I/O. \fBlseek\fR\|(2) is used to position the I/O location.
1787 See \fBfsync\fR and \fBfdatasync\fR for syncing write I/Os.
1790 Basic \fBpread\fR\|(2) or \fBpwrite\fR\|(2) I/O. Default on
1791 all supported operating systems except for Windows.
1794 Basic \fBreadv\fR\|(2) or \fBwritev\fR\|(2) I/O. Will emulate
1795 queuing by coalescing adjacent I/Os into a single submission.
1798 Basic \fBpreadv\fR\|(2) or \fBpwritev\fR\|(2) I/O.
1801 Basic \fBpreadv2\fR\|(2) or \fBpwritev2\fR\|(2) I/O.
1804 Fast Linux native asynchronous I/O. Supports async IO
1805 for both direct and buffered IO.
1806 This engine defines engine specific options.
1809 Fast Linux native asynchronous I/O for passthrough commands.
1810 This engine defines engine specific options.
1813 Linux native asynchronous I/O. Note that Linux may only support
1814 queued behavior with non-buffered I/O (set `direct=1' or
1816 This engine defines engine specific options.
1819 POSIX asynchronous I/O using \fBaio_read\fR\|(3) and
1820 \fBaio_write\fR\|(3).
1823 Solaris native asynchronous I/O.
1826 Windows native asynchronous I/O. Default on Windows.
1829 File is memory mapped with \fBmmap\fR\|(2) and data copied
1830 to/from using \fBmemcpy\fR\|(3).
1833 \fBsplice\fR\|(2) is used to transfer the data and
1834 \fBvmsplice\fR\|(2) to transfer data from user space to the
1838 SCSI generic sg v3 I/O. May either be synchronous using the SG_IO
1839 ioctl, or if the target is an sg character device we use
1840 \fBread\fR\|(2) and \fBwrite\fR\|(2) for asynchronous
1841 I/O. Requires \fBfilename\fR option to specify either block or
1842 character devices. This engine supports trim operations. The
1843 sg engine includes engine specific options.
1846 Read, write, trim and ZBC/ZAC operations to a zoned block device using
1847 \fBlibzbc\fR library. The target can be either an SG character device or
1848 a block device file.
1851 Doesn't transfer any data, just pretends to. This is mainly used to
1852 exercise fio itself and for debugging/testing purposes.
1855 Transfer over the network to given `host:port'. Depending on the
1856 \fBprotocol\fR used, the \fBhostname\fR, \fBport\fR,
1857 \fBlisten\fR and \fBfilename\fR options are used to specify
1858 what sort of connection to make, while the \fBprotocol\fR option
1859 determines which protocol will be used. This engine defines engine
1863 Like \fBnet\fR, but uses \fBsplice\fR\|(2) and
1864 \fBvmsplice\fR\|(2) to map data and send/receive.
1865 This engine defines engine specific options.
1868 Doesn't transfer any data, but burns CPU cycles according to the
1869 \fBcpuload\fR, \fBcpuchunks\fR and \fBcpumode\fR options.
1870 A job never finishes unless there is at least one non-cpuio job.
1874 \fBcpuload\fR\=85 will cause that job to do nothing but burn 85% of the CPU.
1875 In case of SMP machines, use \fBnumjobs=<nr_of_cpu>\fR\ to get desired CPU usage,
1876 as the cpuload only loads a single CPU at the desired rate.
1879 \fBcpumode\fR\=qsort replace the default noop instructions loop
1880 by a qsort algorithm to consume more energy.
1886 The RDMA I/O engine supports both RDMA memory semantics
1887 (RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the
1888 InfiniBand, RoCE and iWARP protocols. This engine defines engine
1892 I/O engine that does regular fallocate to simulate data transfer as
1897 DDIR_READ does fallocate(,mode = FALLOC_FL_KEEP_SIZE,).
1899 DIR_WRITE does fallocate(,mode = 0).
1901 DDIR_TRIM does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE).
1906 I/O engine that sends \fBftruncate\fR\|(2) operations in response
1907 to write (DDIR_WRITE) events. Each ftruncate issued sets the file's
1908 size to the current block offset. \fBblocksize\fR is ignored.
1911 I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate
1912 defragment activity in request to DDIR_WRITE event.
1915 I/O engine supporting direct access to Ceph Reliable Autonomic Distributed
1916 Object Store (RADOS) via librados. This ioengine defines engine specific
1920 I/O engine supporting direct access to Ceph Rados Block Devices
1921 (RBD) via librbd without the need to use the kernel rbd driver. This
1922 ioengine defines engine specific options.
1925 I/O engine supporting GET/PUT requests over HTTP(S) with libcurl to
1926 a WebDAV or S3 endpoint. This ioengine defines engine specific options.
1928 This engine only supports direct IO of iodepth=1; you need to scale this
1929 via numjobs. blocksize defines the size of the objects to be created.
1931 TRIM is translated to object deletion.
1934 Using GlusterFS libgfapi sync interface to direct access to
1935 GlusterFS volumes without having to go through FUSE. This ioengine
1936 defines engine specific options.
1939 Using GlusterFS libgfapi async interface to direct access to
1940 GlusterFS volumes without having to go through FUSE. This ioengine
1941 defines engine specific options.
1944 Read and write through Hadoop (HDFS). The \fBfilename\fR option
1945 is used to specify host,port of the hdfs name\-node to connect. This
1946 engine interprets offsets a little differently. In HDFS, files once
1947 created cannot be modified so random writes are not possible. To
1948 imitate this the libhdfs engine expects a bunch of small files to be
1949 created over HDFS and will randomly pick a file from them
1950 based on the offset generated by fio backend (see the example
1951 job file to create such files, use `rw=write' option). Please
1952 note, it may be necessary to set environment variables to work
1953 with HDFS/libhdfs properly. Each job uses its own connection to
1957 Read, write and erase an MTD character device (e.g.,
1958 `/dev/mtd0'). Discards are treated as erases. Depending on the
1959 underlying device type, the I/O may have to go in a certain pattern,
1960 e.g., on NAND, writing sequentially to erase blocks and discarding
1961 before overwriting. The \fBtrimwrite\fR mode works well for this
1965 Read and write using device DAX to a persistent memory device (e.g.,
1966 /dev/dax0.0) through the PMDK libpmem library.
1969 Prefix to specify loading an external I/O engine object file. Append
1970 the engine filename, e.g. `ioengine=external:/tmp/foo.o' to load
1971 ioengine `foo.o' in `/tmp'. The path can be either
1972 absolute or relative. See `engines/skeleton_external.c' in the fio source for
1973 details of writing an external I/O engine.
1976 Simply create the files and do no I/O to them. You still need to set
1977 \fBfilesize\fR so that all the accounting still occurs, but no actual I/O will be
1978 done other than creating the file.
1981 Simply do stat() and do no I/O to the file. You need to set 'filesize'
1982 and 'nrfiles', so that files will be created.
1983 This engine is to measure file lookup and meta data access.
1986 Simply delete files by unlink() and do no I/O to the file. You need to set 'filesize'
1987 and 'nrfiles', so that files will be created.
1988 This engine is to measure file delete.
1991 Read and write using mmap I/O to a file on a filesystem
1992 mounted with DAX on a persistent memory device through the PMDK
1996 Synchronous read and write using DDN's Infinite Memory Engine (IME). This
1997 engine is very basic and issues calls to IME whenever an IO is queued.
2000 Synchronous read and write using DDN's Infinite Memory Engine (IME). This
2001 engine uses iovecs and will try to stack as much IOs as possible (if the IOs
2002 are "contiguous" and the IO depth is not exceeded) before issuing a call to IME.
2005 Asynchronous read and write using DDN's Infinite Memory Engine (IME). This
2006 engine will try to stack as much IOs as possible by creating requests for IME.
2007 FIO will then decide when to commit these requests.
2010 Read and write iscsi lun with libiscsi.
2013 Synchronous read and write a Network Block Device (NBD).
2016 I/O engine supporting libcufile synchronous access to nvidia-fs and a
2017 GPUDirect Storage-supported filesystem. This engine performs
2018 I/O without transferring buffers between user-space and the kernel,
2019 unless \fBverify\fR is set or \fBcuda_io\fR is \fBposix\fR. \fBiomem\fR must
2020 not be \fBcudamalloc\fR. This ioengine defines engine specific options.
2023 I/O engine supporting asynchronous read and write operations to the DAOS File
2024 System (DFS) via libdfs.
2027 I/O engine supporting asynchronous read and write operations to
2028 NFS filesystems from userspace via libnfs. This is useful for
2029 achieving higher concurrency and thus throughput than is possible
2033 Execute 3rd party tools. Could be used to perform monitoring during jobs runtime.
2036 I/O engine using the xNVMe C API, for NVMe devices. The xnvme engine provides
2037 flexibility to access GNU/Linux Kernel NVMe driver via libaio, IOCTLs, io_uring,
2038 the SPDK NVMe driver, or your own custom NVMe driver. The xnvme engine includes
2039 engine specific options. (See \fIhttps://xnvme.io/\fR).
2042 Use the libblkio library (\fIhttps://gitlab.com/libblkio/libblkio\fR). The
2043 specific driver to use must be set using \fBlibblkio_driver\fR. If
2044 \fBmem\fR/\fBiomem\fR is not specified, memory allocation is delegated to
2045 libblkio (and so is guaranteed to work with the selected driver). One libblkio
2046 instance is used per process, so all jobs setting option \fBthread\fR will share
2047 a single instance (with one queue per thread) and must specify compatible
2048 options. Note that some drivers don't allow several instances to access the same
2049 device or file simultaneously, but allow it for threads.
2050 .SS "I/O engine specific parameters"
2051 In addition, there are some parameters which are only valid when a specific
2052 \fBioengine\fR is in use. These are used identically to normal parameters,
2053 with the caveat that when used on the command line, they must come after the
2054 \fBioengine\fR that defines them is selected.
2056 .BI (io_uring,libaio)cmdprio_percentage \fR=\fPint[,int]
2057 Set the percentage of I/O that will be issued with the highest priority.
2058 Default: 0. A single value applies to reads and writes. Comma-separated
2059 values may be specified for reads and writes. For this option to be effective,
2060 NCQ priority must be supported and enabled, and `direct=1' option must be
2061 used. fio must also be run as the root user. Unlike slat/clat/lat stats, which
2062 can be tracked and reported independently, per priority stats only track and
2063 report a single type of latency. By default, completion latency (clat) will be
2064 reported, if \fBlat_percentiles\fR is set, total latency (lat) will be reported.
2066 .BI (io_uring,libaio)cmdprio_class \fR=\fPint[,int]
2067 Set the I/O priority class to use for I/Os that must be issued with a
2068 priority when \fBcmdprio_percentage\fR or \fBcmdprio_bssplit\fR is set.
2069 If not specified when \fBcmdprio_percentage\fR or \fBcmdprio_bssplit\fR
2070 is set, this defaults to the highest priority class. A single value applies
2071 to reads and writes. Comma-separated values may be specified for reads and
2072 writes. See man \fBionice\fR\|(1). See also the \fBprioclass\fR option.
2074 .BI (io_uring,libaio)cmdprio \fR=\fPint[,int]
2075 Set the I/O priority value to use for I/Os that must be issued with a
2076 priority when \fBcmdprio_percentage\fR or \fBcmdprio_bssplit\fR is set.
2077 If not specified when \fBcmdprio_percentage\fR or \fBcmdprio_bssplit\fR
2078 is set, this defaults to 0. Linux limits us to a positive value between
2079 0 and 7, with 0 being the highest. A single value applies to reads and writes.
2080 Comma-separated values may be specified for reads and writes. See man
2081 \fBionice\fR\|(1). Refer to an appropriate manpage for other operating systems
2082 since the meaning of priority may differ. See also the \fBprio\fR option.
2084 .BI (io_uring,libaio)cmdprio_bssplit \fR=\fPstr[,str]
2085 To get a finer control over I/O priority, this option allows specifying
2086 the percentage of IOs that must have a priority set depending on the block
2087 size of the IO. This option is useful only when used together with the option
2088 \fBbssplit\fR, that is, multiple different block sizes are used for reads and
2092 The first accepted format for this option is the same as the format of the
2093 \fBbssplit\fR option:
2096 cmdprio_bssplit=blocksize/percentage:blocksize/percentage
2099 In this case, each entry will use the priority class and priority level defined
2100 by the options \fBcmdprio_class\fR and \fBcmdprio\fR respectively.
2102 The second accepted format for this option is:
2105 cmdprio_bssplit=blocksize/percentage/class/level:blocksize/percentage/class/level
2108 In this case, the priority class and priority level is defined inside each
2109 entry. In comparison with the first accepted format, the second accepted format
2110 does not restrict all entries to have the same priority class and priority
2113 For both formats, only the read and write data directions are supported, values
2114 for trim IOs are ignored. This option is mutually exclusive with the
2115 \fBcmdprio_percentage\fR option.
2118 .BI (io_uring,io_uring_cmd)fixedbufs
2119 If fio is asked to do direct IO, then Linux will map pages for each IO call, and
2120 release them when IO is done. If this option is set, the pages are pre-mapped
2121 before IO is started. This eliminates the need to map and release for each IO.
2122 This is more efficient, and reduces the IO latency as well.
2124 .BI (io_uring,io_uring_cmd)nonvectored \fR=\fPint
2125 With this option, fio will use non-vectored read/write commands, where address
2126 must contain the address directly. Default is -1.
2128 .BI (io_uring,io_uring_cmd)force_async
2129 Normal operation for io_uring is to try and issue an sqe as non-blocking first,
2130 and if that fails, execute it in an async manner. With this option set to N,
2131 then every N request fio will ask sqe to be issued in an async manner. Default
2134 .BI (io_uring,io_uring_cmd,xnvme)hipri
2135 If this option is set, fio will attempt to use polled IO completions. Normal IO
2136 completions generate interrupts to signal the completion of IO, polled
2137 completions do not. Hence they are require active reaping by the application.
2138 The benefits are more efficient IO for high IOPS scenarios, and lower latencies
2139 for low queue depth IO.
2141 .BI (io_uring,io_uring_cmd)registerfiles
2142 With this option, fio registers the set of files being used with the kernel.
2143 This avoids the overhead of managing file counts in the kernel, making the
2144 submission and completion part more lightweight. Required for the below
2145 sqthread_poll option.
2147 .BI (io_uring,io_uring_cmd,xnvme)sqthread_poll
2148 Normally fio will submit IO by issuing a system call to notify the kernel of
2149 available items in the SQ ring. If this option is set, the act of submitting IO
2150 will be done by a polling thread in the kernel. This frees up cycles for fio, at
2151 the cost of using more CPU in the system. As submission is just the time it
2152 takes to fill in the sqe entries and any syscall required to wake up the idle
2153 kernel thread, fio will not report submission latencies.
2155 .BI (io_uring,io_uring_cmd)sqthread_poll_cpu \fR=\fPint
2156 When `sqthread_poll` is set, this option provides a way to define which CPU
2157 should be used for the polling thread.
2159 .BI (io_uring_cmd)cmd_type \fR=\fPstr
2160 Specifies the type of uring passthrough command to be used. Supported
2161 value is nvme. Default is nvme.
2163 .BI (libaio)userspace_reap
2164 Normally, with the libaio engine in use, fio will use the
2165 \fBio_getevents\fR\|(3) system call to reap newly returned events. With
2166 this flag turned on, the AIO ring will be read directly from user-space to
2167 reap events. The reaping mode is only enabled when polling for a minimum of
2168 0 events (e.g. when `iodepth_batch_complete=0').
2171 Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority
2174 .BI (pvsync2)hipri_percentage
2175 When hipri is set this determines the probability of a pvsync2 I/O being high
2176 priority. The default is 100%.
2178 .BI (pvsync2,libaio,io_uring,io_uring_cmd)nowait \fR=\fPbool
2179 By default if a request cannot be executed immediately (e.g. resource starvation,
2180 waiting on locks) it is queued and the initiating process will be blocked until
2181 the required resource becomes free.
2182 This option sets the RWF_NOWAIT flag (supported from the 4.14 Linux kernel) and
2183 the call will return instantly with EAGAIN or a partial result rather than waiting.
2185 It is useful to also use \fBignore_error\fR=EAGAIN when using this option.
2186 Note: glibc 2.27, 2.28 have a bug in syscall wrappers preadv2, pwritev2.
2187 They return EOPNOTSUP instead of EAGAIN.
2189 For cached I/O, using this option usually means a request operates only with
2190 cached data. Currently the RWF_NOWAIT flag does not supported for cached write.
2191 For direct I/O, requests will only succeed if cache invalidation isn't required,
2192 file blocks are fully allocated and the disk request could be issued immediately.
2194 .BI (io_uring_cmd)fdp \fR=\fPbool
2195 Enable Flexible Data Placement mode for write commands.
2197 .BI (io_uring_cmd)fdp_pli \fR=\fPstr
2198 Select which Placement ID Index/Indicies this job is allowed to use for writes.
2199 By default, the job will cycle through all available Placement IDs, so use this
2200 to isolate these identifiers to specific jobs. If you want fio to use placement
2201 identifier only at indices 0, 2 and 5 specify, you would set `fdp_pli=0,2,5`.
2203 .BI (cpuio)cpuload \fR=\fPint
2204 Attempt to use the specified percentage of CPU cycles. This is a mandatory
2205 option when using cpuio I/O engine.
2207 .BI (cpuio)cpuchunks \fR=\fPint
2208 Split the load into cycles of the given time. In microseconds.
2210 .BI (cpuio)cpumode \fR=\fPstr
2211 Specify how to stress the CPU. It can take these two values:
2216 This is the default and directs the CPU to execute noop instructions.
2219 Replace the default noop instructions with a qsort algorithm to consume more energy.
2223 .BI (cpuio)exit_on_io_done \fR=\fPbool
2224 Detect when I/O threads are done, then exit.
2226 .BI (libhdfs)namenode \fR=\fPstr
2227 The hostname or IP address of a HDFS cluster namenode to contact.
2229 .BI (libhdfs)port \fR=\fPint
2230 The listening port of the HFDS cluster namenode.
2232 .BI (netsplice,net)port \fR=\fPint
2233 The TCP or UDP port to bind to or connect to. If this is used with
2234 \fBnumjobs\fR to spawn multiple instances of the same job type, then
2235 this will be the starting port number since fio will use a range of
2238 .BI (rdma,librpma_*)port \fR=\fPint
2239 The port to use for RDMA-CM communication. This should be the same
2240 value on the client and the server side.
2242 .BI (netsplice,net,rdma)hostname \fR=\fPstr
2243 The hostname or IP address to use for TCP, UDP or RDMA-CM based I/O.
2244 If the job is a TCP listener or UDP reader, the hostname is not used
2245 and must be omitted unless it is a valid UDP multicast address.
2247 .BI (librpma_*)serverip \fR=\fPstr
2248 The IP address to be used for RDMA-CM based I/O.
2250 .BI (librpma_*_server)direct_write_to_pmem \fR=\fPbool
2251 Set to 1 only when Direct Write to PMem from the remote host is possible. Otherwise, set to 0.
2253 .BI (librpma_*_server)busy_wait_polling \fR=\fPbool
2254 Set to 0 to wait for completion instead of busy-wait polling completion.
2257 .BI (netsplice,net)interface \fR=\fPstr
2258 The IP address of the network interface used to send or receive UDP
2261 .BI (netsplice,net)ttl \fR=\fPint
2262 Time\-to\-live value for outgoing UDP multicast packets. Default: 1.
2264 .BI (netsplice,net)nodelay \fR=\fPbool
2265 Set TCP_NODELAY on TCP connections.
2267 .BI (netsplice,net)protocol \fR=\fPstr "\fR,\fP proto" \fR=\fPstr
2268 The network protocol to use. Accepted values are:
2273 Transmission control protocol.
2276 Transmission control protocol V6.
2279 User datagram protocol.
2282 User datagram protocol V6.
2288 When the protocol is TCP or UDP, the port must also be given, as well as the
2289 hostname if the job is a TCP listener or UDP reader. For unix sockets, the
2290 normal \fBfilename\fR option should be used and the port is invalid.
2293 .BI (netsplice,net)listen
2294 For TCP network connections, tell fio to listen for incoming connections
2295 rather than initiating an outgoing connection. The \fBhostname\fR must
2296 be omitted if this option is used.
2298 .BI (netsplice,net)pingpong
2299 Normally a network writer will just continue writing data, and a network
2300 reader will just consume packages. If `pingpong=1' is set, a writer will
2301 send its normal payload to the reader, then wait for the reader to send the
2302 same payload back. This allows fio to measure network latencies. The
2303 submission and completion latencies then measure local time spent sending or
2304 receiving, and the completion latency measures how long it took for the
2305 other end to receive and send back. For UDP multicast traffic
2306 `pingpong=1' should only be set for a single reader when multiple readers
2307 are listening to the same address.
2309 .BI (netsplice,net)window_size \fR=\fPint
2310 Set the desired socket buffer size for the connection.
2312 .BI (netsplice,net)mss \fR=\fPint
2313 Set the TCP maximum segment size (TCP_MAXSEG).
2315 .BI (e4defrag)donorname \fR=\fPstr
2316 File will be used as a block donor (swap extents between files).
2318 .BI (e4defrag)inplace \fR=\fPint
2319 Configure donor file blocks allocation strategy:
2324 Default. Preallocate donor's file on init.
2327 Allocate space immediately inside defragment event, and free right
2332 .BI (rbd,rados)clustername \fR=\fPstr
2333 Specifies the name of the Ceph cluster.
2335 .BI (rbd)rbdname \fR=\fPstr
2336 Specifies the name of the RBD.
2338 .BI (rbd,rados)pool \fR=\fPstr
2339 Specifies the name of the Ceph pool containing RBD or RADOS data.
2341 .BI (rbd,rados)clientname \fR=\fPstr
2342 Specifies the username (without the 'client.' prefix) used to access the
2343 Ceph cluster. If the \fBclustername\fR is specified, the \fBclientname\fR shall be
2344 the full *type.id* string. If no type. prefix is given, fio will add 'client.'
2347 .BI (rados)conf \fR=\fPstr
2348 Specifies the configuration path of ceph cluster, so conf file does not
2349 have to be /etc/ceph/ceph.conf.
2351 .BI (rbd,rados)busy_poll \fR=\fPbool
2352 Poll store instead of waiting for completion. Usually this provides better
2353 throughput at cost of higher(up to 100%) CPU utilization.
2355 .BI (rados)touch_objects \fR=\fPbool
2356 During initialization, touch (create if do not exist) all objects (files).
2357 Touching all objects affects ceph caches and likely impacts test results.
2360 .BI (http)http_host \fR=\fPstr
2361 Hostname to connect to. For S3, this could be the bucket name. Default
2364 .BI (http)http_user \fR=\fPstr
2365 Username for HTTP authentication.
2367 .BI (http)http_pass \fR=\fPstr
2368 Password for HTTP authentication.
2370 .BI (http)https \fR=\fPstr
2371 Whether to use HTTPS instead of plain HTTP. \fRon\fP enables HTTPS;
2372 \fRinsecure\fP will enable HTTPS, but disable SSL peer verification (use
2373 with caution!). Default is \fBoff\fR.
2375 .BI (http)http_mode \fR=\fPstr
2376 Which HTTP access mode to use: webdav, swift, or s3. Default is
2379 .BI (http)http_s3_region \fR=\fPstr
2380 The S3 region/zone to include in the request. Default is \fBus-east-1\fR.
2382 .BI (http)http_s3_key \fR=\fPstr
2385 .BI (http)http_s3_keyid \fR=\fPstr
2386 The S3 key/access id.
2388 .BI (http)http_s3_sse_customer_key \fR=\fPstr
2389 The encryption customer key in SSE server side.
2391 .BI (http)http_s3_sse_customer_algorithm \fR=\fPstr
2392 The encryption customer algorithm in SSE server side. Default is \fBAES256\fR
2394 .BI (http)http_s3_storage_class \fR=\fPstr
2395 Which storage class to access. User-customizable settings. Default is \fBSTANDARD\fR
2397 .BI (http)http_swift_auth_token \fR=\fPstr
2398 The Swift auth token. See the example configuration file on how to
2401 .BI (http)http_verbose \fR=\fPint
2402 Enable verbose requests from libcurl. Useful for debugging. 1 turns on
2403 verbose logging from libcurl, 2 additionally enables HTTP IO tracing.
2406 .BI (mtd)skip_bad \fR=\fPbool
2407 Skip operations against known bad blocks.
2409 .BI (libhdfs)hdfsdirectory
2410 libhdfs will create chunk in this HDFS directory.
2412 .BI (libhdfs)chunk_size
2413 The size of the chunk to use for each file.
2415 .BI (rdma)verb \fR=\fPstr
2416 The RDMA verb to use on this side of the RDMA ioengine
2417 connection. Valid values are write, read, send and recv. These
2418 correspond to the equivalent RDMA verbs (e.g. write = rdma_write
2419 etc.). Note that this only needs to be specified on the client side of
2420 the connection. See the examples folder.
2422 .BI (rdma)bindname \fR=\fPstr
2423 The name to use to bind the local RDMA-CM connection to a local RDMA
2424 device. This could be a hostname or an IPv4 or IPv6 address. On the
2425 server side this will be passed into the rdma_bind_addr() function and
2426 on the client site it will be used in the rdma_resolve_add()
2427 function. This can be useful when multiple paths exist between the
2428 client and the server or in certain loopback configurations.
2430 .BI (filestat)stat_type \fR=\fPstr
2431 Specify stat system call type to measure lookup/getattr performance.
2432 Default is \fBstat\fR for \fBstat\fR\|(2).
2435 If this option is set, fio will attempt to use polled IO completions. This
2436 will have a similar effect as (io_uring)hipri. Only SCSI READ and WRITE
2437 commands will have the SGV4_FLAG_HIPRI set (not UNMAP (trim) nor VERIFY).
2438 Older versions of the Linux sg driver that do not support hipri will simply
2439 ignore this flag and do normal IO. The Linux SCSI Low Level Driver (LLD)
2440 that "owns" the device also needs to support hipri (also known as iopoll
2441 and mq_poll). The MegaRAID driver is an example of a SCSI LLD.
2442 Default: clear (0) which does normal (interrupted based) IO.
2444 .BI (sg)readfua \fR=\fPbool
2445 With readfua option set to 1, read operations include the force
2446 unit access (fua) flag. Default: 0.
2448 .BI (sg)writefua \fR=\fPbool
2449 With writefua option set to 1, write operations include the force
2450 unit access (fua) flag. Default: 0.
2452 .BI (sg)sg_write_mode \fR=\fPstr
2453 Specify the type of write commands to issue. This option can take multiple
2459 Write opcodes are issued as usual
2462 Issue WRITE AND VERIFY commands. The BYTCHK bit is set to 00b. This directs the
2463 device to carry out a medium verification with no data comparison for the data
2464 that was written. The writefua option is ignored with this selection.
2467 This option is deprecated. Use write_and_verify instead.
2470 Issue WRITE SAME commands. This transfers a single block to the device
2471 and writes this same block of data to a contiguous sequence of LBAs
2472 beginning at the specified offset. fio's block size parameter
2473 specifies the amount of data written with each command. However, the
2474 amount of data actually transferred to the device is equal to the
2475 device's block (sector) size. For a device with 512 byte sectors,
2476 blocksize=8k will write 16 sectors with each command. fio will still
2477 generate 8k of data for each command butonly the first 512 bytes will
2478 be used and transferred to the device. The writefua option is ignored
2479 with this selection.
2482 This option is deprecated. Use write_same instead.
2485 Issue WRITE SAME(16) commands as above but with the No Data Output
2486 Buffer (NDOB) bit set. No data will be transferred to the device with
2487 this bit set. Data written will be a pre-determined pattern such as
2491 Issue WRITE STREAM(16) commands. Use the stream_id option to specify
2492 the stream identifier.
2495 Issue VERIFY commands with BYTCHK set to 00. This directs the device to carry
2496 out a medium verification with no data comparison.
2499 Issue VERIFY commands with BYTCHK set to 01. This directs the device to
2500 compare the data on the device with the data transferred to the device.
2503 Issue VERIFY commands with BYTCHK set to 11. This transfers a single block to
2504 the device and compares the contents of this block with the data on the device
2505 beginning at the specified offset. fio's block size parameter specifies the
2506 total amount of data compared with this command. However, only one block
2507 (sector) worth of data is transferred to the device. This is similar to the
2508 WRITE SAME command except that data is compared instead of written.
2512 .BI (sg)stream_id \fR=\fPint
2513 Set the stream identifier for WRITE STREAM commands. If this is set to 0 (which is not
2514 a valid stream identifier) fio will open a stream and then close it when done. Default
2517 .BI (nbd)uri \fR=\fPstr
2518 Specify the NBD URI of the server to test.
2519 The string is a standard NBD URI (see
2520 \fIhttps://github.com/NetworkBlockDevice/nbd/tree/master/doc\fR).
2525 \fInbd://localhost:10809\fR
2527 \fInbd+unix:///?socket=/tmp/socket\fR
2529 \fInbds://tlshost/exportname\fR
2533 .BI (libcufile)gpu_dev_ids\fR=\fPstr
2534 Specify the GPU IDs to use with CUDA. This is a colon-separated list of int.
2535 GPUs are assigned to workers roundrobin. Default is 0.
2537 .BI (libcufile)cuda_io\fR=\fPstr
2538 Specify the type of I/O to use with CUDA. This option
2539 takes the following values:
2544 Use libcufile and nvidia-fs. This option performs I/O directly
2545 between a GPUDirect Storage filesystem and GPU buffers,
2546 avoiding use of a bounce buffer. If \fBverify\fR is set,
2547 cudaMemcpy is used to copy verification data between RAM and GPU(s).
2548 Verification data is copied from RAM to GPU before a write
2549 and from GPU to RAM after a read.
2550 \fBdirect\fR must be 1.
2553 Use POSIX to perform I/O with a RAM buffer, and use
2554 cudaMemcpy to transfer data between RAM and the GPU(s).
2555 Data is copied from GPU to RAM before a write and copied
2556 from RAM to GPU after a read. \fBverify\fR does not affect
2557 the use of cudaMemcpy.
2562 Specify the label or UUID of the DAOS pool to connect to.
2565 Specify the label or UUID of the DAOS container to open.
2568 Specify a different chunk size (in bytes) for the dfs file.
2569 Use DAOS container's chunk size by default.
2571 .BI (dfs)object_class
2572 Specify a different object class for the dfs file.
2573 Use DAOS container's object class by default.
2576 URL in libnfs format, eg nfs://<server|ipv4|ipv6>/path[?arg=val[&arg=val]*]
2577 Refer to the libnfs README for more details.
2579 .BI (exec)program\fR=\fPstr
2580 Specify the program to execute.
2581 Note the program will receive a SIGTERM when the job is reaching the time limit.
2582 A SIGKILL is sent once the job is over. The delay between the two signals is defined by \fBgrace_time\fR option.
2584 .BI (exec)arguments\fR=\fPstr
2585 Specify arguments to pass to program.
2586 Some special variables can be expanded to pass fio's job details to the program :
2591 replaced by the duration of the job in seconds
2594 replaced by the name of the job
2598 .BI (exec)grace_time\fR=\fPint
2599 Defines the time between the SIGTERM and SIGKILL signals. Default is 1 second.
2601 .BI (exec)std_redirect\fR=\fPbool
2602 If set, stdout and stderr streams are redirected to files named from the job name. Default is true.
2604 .BI (xnvme)xnvme_async\fR=\fPstr
2605 Select the xnvme async command interface. This can take these values.
2610 This is default and use to emulate asynchronous I/O by using a single thread to
2611 create a queue pair on top of a synchronous I/O interface using the NVMe driver
2615 Emulate an asynchronous I/O interface with a pool of userspace threads on top
2616 of a synchronous I/O interface using the NVMe driver IOCTL. By default four
2620 Linux native asynchronous I/O interface which supports both direct and buffered
2624 Use Linux aio for Asynchronous I/O
2627 Use the posix asynchronous I/O interface to perform one or more I/O operations
2631 Use the user-space VFIO-based backend, implemented using libvfn instead of
2635 Do not transfer any data; just pretend to. This is mainly used for
2636 introspective performance evaluation.
2640 .BI (xnvme)xnvme_sync\fR=\fPstr
2641 Select the xnvme synchronous command interface. This can take these values.
2646 This is default and uses Linux NVMe Driver ioctl() for synchronous I/O.
2649 This supports regular as well as vectored pread() and pwrite() commands.
2652 This is the same as psync except that it also supports zone management
2653 commands using Linux block layer IOCTLs.
2657 .BI (xnvme)xnvme_admin\fR=\fPstr
2658 Select the xnvme admin command interface. This can take these values.
2663 This is default and uses Linux NVMe Driver ioctl() for admin commands.
2666 Use Linux Block Layer ioctl() and sysfs for admin commands.
2670 .BI (xnvme)xnvme_dev_nsid\fR=\fPint
2671 xnvme namespace identifier for userspace NVMe driver SPDK or vfio.
2673 .BI (xnvme)xnvme_dev_subnqn\fR=\fPstr
2674 Sets the subsystem NQN for fabrics. This is for xNVMe to utilize a fabrics
2675 target with multiple systems.
2677 .BI (xnvme)xnvme_mem\fR=\fPstr
2678 Select the xnvme memory backend. This can take these values.
2683 This is the default posix memory backend for linux NVMe driver.
2686 Use hugepages, instead of existing posix memory backend. The memory backend
2687 uses hugetlbfs. This require users to allocate hugepages, mount hugetlbfs and
2688 set an enviornment variable for XNVME_HUGETLB_PATH.
2691 Uses SPDK's memory allocator.
2694 Uses libvfn's memory allocator. This also specifies the use of libvfn backend
2699 .BI (xnvme)xnvme_iovec
2700 If this option is set, xnvme will use vectored read/write commands.
2702 .BI (libblkio)libblkio_driver \fR=\fPstr
2703 The libblkio driver to use. Different drivers access devices through different
2704 underlying interfaces. Available drivers depend on the libblkio version in use
2705 and are listed at \fIhttps://libblkio.gitlab.io/libblkio/blkio.html#drivers\fR
2707 .BI (libblkio)libblkio_path \fR=\fPstr
2708 Sets the value of the driver-specific "path" property before connecting the
2709 libblkio instance, which identifies the target device or file on which to
2710 perform I/O. Its exact semantics are driver-dependent and not all drivers may
2711 support it; see \fIhttps://libblkio.gitlab.io/libblkio/blkio.html#drivers\fR
2713 .BI (libblkio)libblkio_pre_connect_props \fR=\fPstr
2714 A colon-separated list of additional libblkio properties to be set after
2715 creating but before connecting the libblkio instance. Each property must have
2716 the format \fB<name>=<value>\fR. Colons can be escaped as \fB\\:\fR. These are
2717 set after the engine sets any other properties, so those can be overriden.
2718 Available properties depend on the libblkio version in use and are listed at
2719 \fIhttps://libblkio.gitlab.io/libblkio/blkio.html#properties\fR
2721 .BI (libblkio)libblkio_num_entries \fR=\fPint
2722 Sets the value of the driver-specific "num-entries" property before starting the
2723 libblkio instance. Its exact semantics are driver-dependent and not all drivers
2724 may support it; see \fIhttps://libblkio.gitlab.io/libblkio/blkio.html#drivers\fR
2726 .BI (libblkio)libblkio_queue_size \fR=\fPint
2727 Sets the value of the driver-specific "queue-size" property before starting the
2728 libblkio instance. Its exact semantics are driver-dependent and not all drivers
2729 may support it; see \fIhttps://libblkio.gitlab.io/libblkio/blkio.html#drivers\fR
2731 .BI (libblkio)libblkio_pre_start_props \fR=\fPstr
2732 A colon-separated list of additional libblkio properties to be set after
2733 connecting but before starting the libblkio instance. Each property must have
2734 the format \fB<name>=<value>\fR. Colons can be escaped as \fB\\:\fR. These are
2735 set after the engine sets any other properties, so those can be overriden.
2736 Available properties depend on the libblkio version in use and are listed at
2737 \fIhttps://libblkio.gitlab.io/libblkio/blkio.html#properties\fR
2740 Use poll queues. This is incompatible with \fBlibblkio_wait_mode=eventfd\fR and
2741 \fBlibblkio_force_enable_completion_eventfd\fR.
2743 .BI (libblkio)libblkio_vectored
2744 Submit vectored read and write requests.
2746 .BI (libblkio)libblkio_write_zeroes_on_trim
2747 Submit trims as "write zeroes" requests instead of discard requests.
2749 .BI (libblkio)libblkio_wait_mode \fR=\fPstr
2750 How to wait for completions:
2754 .B block \fR(default)
2755 Use a blocking call to \fBblkioq_do_io()\fR.
2758 Use a blocking call to \fBread()\fR on the completion eventfd.
2761 Use a busy loop with a non-blocking call to \fBblkioq_do_io()\fR.
2765 .BI (libblkio)libblkio_force_enable_completion_eventfd
2766 Enable the queue's completion eventfd even when unused. This may impact
2767 performance. The default is to enable it only if
2768 \fBlibblkio_wait_mode=eventfd\fR.
2771 .BI iodepth \fR=\fPint
2772 Number of I/O units to keep in flight against the file. Note that
2773 increasing \fBiodepth\fR beyond 1 will not affect synchronous ioengines (except
2774 for small degrees when \fBverify_async\fR is in use). Even async
2775 engines may impose OS restrictions causing the desired depth not to be
2776 achieved. This may happen on Linux when using libaio and not setting
2777 `direct=1', since buffered I/O is not async on that OS. Keep an
2778 eye on the I/O depth distribution in the fio output to verify that the
2779 achieved depth is as expected. Default: 1.
2781 .BI iodepth_batch_submit \fR=\fPint "\fR,\fP iodepth_batch" \fR=\fPint
2782 This defines how many pieces of I/O to submit at once. It defaults to 1
2783 which means that we submit each I/O as soon as it is available, but can be
2784 raised to submit bigger batches of I/O at the time. If it is set to 0 the
2785 \fBiodepth\fR value will be used.
2787 .BI iodepth_batch_complete_min \fR=\fPint "\fR,\fP iodepth_batch_complete" \fR=\fPint
2788 This defines how many pieces of I/O to retrieve at once. It defaults to 1
2789 which means that we'll ask for a minimum of 1 I/O in the retrieval process
2790 from the kernel. The I/O retrieval will go on until we hit the limit set by
2791 \fBiodepth_low\fR. If this variable is set to 0, then fio will always
2792 check for completed events before queuing more I/O. This helps reduce I/O
2793 latency, at the cost of more retrieval system calls.
2795 .BI iodepth_batch_complete_max \fR=\fPint
2796 This defines maximum pieces of I/O to retrieve at once. This variable should
2797 be used along with \fBiodepth_batch_complete_min\fR=\fIint\fR variable,
2798 specifying the range of min and max amount of I/O which should be
2799 retrieved. By default it is equal to \fBiodepth_batch_complete_min\fR
2805 iodepth_batch_complete_min=1
2807 iodepth_batch_complete_max=<iodepth>
2811 which means that we will retrieve at least 1 I/O and up to the whole
2812 submitted queue depth. If none of I/O has been completed yet, we will wait.
2817 iodepth_batch_complete_min=0
2819 iodepth_batch_complete_max=<iodepth>
2823 which means that we can retrieve up to the whole submitted queue depth, but
2824 if none of I/O has been completed yet, we will NOT wait and immediately exit
2825 the system call. In this example we simply do polling.
2828 .BI iodepth_low \fR=\fPint
2829 The low water mark indicating when to start filling the queue
2830 again. Defaults to the same as \fBiodepth\fR, meaning that fio will
2831 attempt to keep the queue full at all times. If \fBiodepth\fR is set to
2832 e.g. 16 and \fBiodepth_low\fR is set to 4, then after fio has filled the queue of
2833 16 requests, it will let the depth drain down to 4 before starting to fill
2836 .BI serialize_overlap \fR=\fPbool
2837 Serialize in-flight I/Os that might otherwise cause or suffer from data races.
2838 When two or more I/Os are submitted simultaneously, there is no guarantee that
2839 the I/Os will be processed or completed in the submitted order. Further, if
2840 two or more of those I/Os are writes, any overlapping region between them can
2841 become indeterminate/undefined on certain storage. These issues can cause
2842 verification to fail erratically when at least one of the racing I/Os is
2843 changing data and the overlapping region has a non-zero size. Setting
2844 \fBserialize_overlap\fR tells fio to avoid provoking this behavior by explicitly
2845 serializing in-flight I/Os that have a non-zero overlap. Note that setting
2846 this option can reduce both performance and the \fBiodepth\fR achieved.
2849 This option only applies to I/Os issued for a single job except when it is
2850 enabled along with \fBio_submit_mode\fR=offload. In offload mode, fio
2851 will check for overlap among all I/Os submitted by offload jobs with \fBserialize_overlap\fR
2857 .BI io_submit_mode \fR=\fPstr
2858 This option controls how fio submits the I/O to the I/O engine. The default
2859 is `inline', which means that the fio job threads submit and reap I/O
2860 directly. If set to `offload', the job threads will offload I/O submission
2861 to a dedicated pool of I/O threads. This requires some coordination and thus
2862 has a bit of extra overhead, especially for lower queue depth I/O where it
2863 can increase latencies. The benefit is that fio can manage submission rates
2864 independently of the device completion rates. This avoids skewed latency
2865 reporting if I/O gets backed up on the device side (the coordinated omission
2866 problem). Note that this option cannot reliably be used with async IO engines.
2869 .BI thinktime \fR=\fPtime
2870 Stall the job for the specified period of time after an I/O has completed before issuing the
2871 next. May be used to simulate processing being done by an application.
2872 When the unit is omitted, the value is interpreted in microseconds. See
2873 \fBthinktime_blocks\fR, \fBthinktime_iotime\fR and \fBthinktime_spin\fR.
2875 .BI thinktime_spin \fR=\fPtime
2876 Only valid if \fBthinktime\fR is set - pretend to spend CPU time doing
2877 something with the data received, before falling back to sleeping for the
2878 rest of the period specified by \fBthinktime\fR. When the unit is
2879 omitted, the value is interpreted in microseconds.
2881 .BI thinktime_blocks \fR=\fPint
2882 Only valid if \fBthinktime\fR is set - control how many blocks to issue,
2883 before waiting \fBthinktime\fR usecs. If not set, defaults to 1 which will make
2884 fio wait \fBthinktime\fR usecs after every block. This effectively makes any
2885 queue depth setting redundant, since no more than 1 I/O will be queued
2886 before we have to complete it and do our \fBthinktime\fR. In other words, this
2887 setting effectively caps the queue depth if the latter is larger.
2889 .BI thinktime_blocks_type \fR=\fPstr
2890 Only valid if \fBthinktime\fR is set - control how \fBthinktime_blocks\fR triggers.
2891 The default is `complete', which triggers \fBthinktime\fR when fio completes
2892 \fBthinktime_blocks\fR blocks. If this is set to `issue', then the trigger happens
2895 .BI thinktime_iotime \fR=\fPtime
2896 Only valid if \fBthinktime\fR is set - control \fBthinktime\fR interval by time.
2897 The \fBthinktime\fR stall is repeated after IOs are executed for
2898 \fBthinktime_iotime\fR. For example, `\-\-thinktime_iotime=9s \-\-thinktime=1s'
2899 repeat 10-second cycle with IOs for 9 seconds and stall for 1 second. When the
2900 unit is omitted, \fBthinktime_iotime\fR is interpreted as a number of seconds.
2901 If this option is used together with \fBthinktime_blocks\fR, the \fBthinktime\fR
2902 stall is repeated after \fBthinktime_iotime\fR or after \fBthinktime_blocks\fR
2903 IOs, whichever happens first.
2906 .BI rate \fR=\fPint[,int][,int]
2907 Cap the bandwidth used by this job. The number is in bytes/sec, the normal
2908 suffix rules apply. Comma-separated values may be specified for reads,
2909 writes, and trims as described in \fBblocksize\fR.
2912 For example, using `rate=1m,500k' would limit reads to 1MiB/sec and writes to
2913 500KiB/sec. Capping only reads or writes can be done with `rate=,500k' or
2914 `rate=500k,' where the former will only limit writes (to 500KiB/sec) and the
2915 latter will only limit reads.
2918 .BI rate_min \fR=\fPint[,int][,int]
2919 Tell fio to do whatever it can to maintain at least this bandwidth. Failing
2920 to meet this requirement will cause the job to exit. Comma-separated values
2921 may be specified for reads, writes, and trims as described in
2924 .BI rate_iops \fR=\fPint[,int][,int]
2925 Cap the bandwidth to this number of IOPS. Basically the same as
2926 \fBrate\fR, just specified independently of bandwidth. If the job is
2927 given a block size range instead of a fixed value, the smallest block size
2928 is used as the metric. Comma-separated values may be specified for reads,
2929 writes, and trims as described in \fBblocksize\fR.
2931 .BI rate_iops_min \fR=\fPint[,int][,int]
2932 If fio doesn't meet this rate of I/O, it will cause the job to exit.
2933 Comma-separated values may be specified for reads, writes, and trims as
2934 described in \fBblocksize\fR.
2936 .BI rate_process \fR=\fPstr
2937 This option controls how fio manages rated I/O submissions. The default is
2938 `linear', which submits I/O in a linear fashion with fixed delays between
2939 I/Os that gets adjusted based on I/O completion rates. If this is set to
2940 `poisson', fio will submit I/O based on a more real world random request
2941 flow, known as the Poisson process
2942 (\fIhttps://en.wikipedia.org/wiki/Poisson_point_process\fR). The lambda will be
2943 10^6 / IOPS for the given workload.
2945 .BI rate_ignore_thinktime \fR=\fPbool
2946 By default, fio will attempt to catch up to the specified rate setting, if any
2947 kind of thinktime setting was used. If this option is set, then fio will
2948 ignore the thinktime and continue doing IO at the specified rate, instead of
2949 entering a catch-up mode after thinktime is done.
2952 .BI latency_target \fR=\fPtime
2953 If set, fio will attempt to find the max performance point that the given
2954 workload will run at while maintaining a latency below this target. When
2955 the unit is omitted, the value is interpreted in microseconds. See
2956 \fBlatency_window\fR and \fBlatency_percentile\fR.
2958 .BI latency_window \fR=\fPtime
2959 Used with \fBlatency_target\fR to specify the sample window that the job
2960 is run at varying queue depths to test the performance. When the unit is
2961 omitted, the value is interpreted in microseconds.
2963 .BI latency_percentile \fR=\fPfloat
2964 The percentage of I/Os that must fall within the criteria specified by
2965 \fBlatency_target\fR and \fBlatency_window\fR. If not set, this
2966 defaults to 100.0, meaning that all I/Os must be equal or below to the value
2967 set by \fBlatency_target\fR.
2969 .BI latency_run \fR=\fPbool
2970 Used with \fBlatency_target\fR. If false (default), fio will find the highest
2971 queue depth that meets \fBlatency_target\fR and exit. If true, fio will continue
2972 running and try to meet \fBlatency_target\fR by adjusting queue depth.
2974 .BI max_latency \fR=\fPtime[,time][,time]
2975 If set, fio will exit the job with an ETIMEDOUT error if it exceeds this
2976 maximum latency. When the unit is omitted, the value is interpreted in
2977 microseconds. Comma-separated values may be specified for reads, writes,
2978 and trims as described in \fBblocksize\fR.
2980 .BI rate_cycle \fR=\fPint
2981 Average bandwidth for \fBrate\fR and \fBrate_min\fR over this number
2982 of milliseconds. Defaults to 1000.
2985 .BI write_iolog \fR=\fPstr
2986 Write the issued I/O patterns to the specified file. See
2987 \fBread_iolog\fR. Specify a separate file for each job, otherwise the
2988 iologs will be interspersed and the file may be corrupt. This file will be
2989 opened in append mode.
2991 .BI read_iolog \fR=\fPstr
2992 Open an iolog with the specified filename and replay the I/O patterns it
2993 contains. This can be used to store a workload and replay it sometime
2994 later. The iolog given may also be a blktrace binary file, which allows fio
2995 to replay a workload captured by blktrace. See
2996 \fBblktrace\fR\|(8) for how to capture such logging data. For blktrace
2997 replay, the file needs to be turned into a blkparse binary data file first
2998 (`blkparse <device> \-o /dev/null \-d file_for_fio.bin').
2999 You can specify a number of files by separating the names with a ':' character.
3000 See the \fBfilename\fR option for information on how to escape ':'
3001 characters within the file names. These files will be sequentially assigned to
3002 job clones created by \fBnumjobs\fR. '-' is a reserved name, meaning read from
3003 stdin, notably if \fBfilename\fR is set to '-' which means stdin as well,
3004 then this flag can't be set to '-'.
3006 .BI read_iolog_chunked \fR=\fPbool
3007 Determines how iolog is read. If false (default) entire \fBread_iolog\fR will
3008 be read at once. If selected true, input from iolog will be read gradually.
3009 Useful when iolog is very large, or it is generated.
3011 .BI merge_blktrace_file \fR=\fPstr
3012 When specified, rather than replaying the logs passed to \fBread_iolog\fR,
3013 the logs go through a merge phase which aggregates them into a single blktrace.
3014 The resulting file is then passed on as the \fBread_iolog\fR parameter. The
3015 intention here is to make the order of events consistent. This limits the
3016 influence of the scheduler compared to replaying multiple blktraces via
3019 .BI merge_blktrace_scalars \fR=\fPfloat_list
3020 This is a percentage based option that is index paired with the list of files
3021 passed to \fBread_iolog\fR. When merging is performed, scale the time of each
3022 event by the corresponding amount. For example,
3023 `\-\-merge_blktrace_scalars="50:100"' runs the first trace in halftime and the
3024 second trace in realtime. This knob is separately tunable from
3025 \fBreplay_time_scale\fR which scales the trace during runtime and will not
3026 change the output of the merge unlike this option.
3028 .BI merge_blktrace_iters \fR=\fPfloat_list
3029 This is a whole number option that is index paired with the list of files
3030 passed to \fBread_iolog\fR. When merging is performed, run each trace for
3031 the specified number of iterations. For example,
3032 `\-\-merge_blktrace_iters="2:1"' runs the first trace for two iterations
3033 and the second trace for one iteration.
3035 .BI replay_no_stall \fR=\fPbool
3036 When replaying I/O with \fBread_iolog\fR the default behavior is to
3037 attempt to respect the timestamps within the log and replay them with the
3038 appropriate delay between IOPS. By setting this variable fio will not
3039 respect the timestamps and attempt to replay them as fast as possible while
3040 still respecting ordering. The result is the same I/O pattern to a given
3041 device, but different timings.
3043 .BI replay_time_scale \fR=\fPint
3044 When replaying I/O with \fBread_iolog\fR, fio will honor the original timing
3045 in the trace. With this option, it's possible to scale the time. It's a
3046 percentage option, if set to 50 it means run at 50% the original IO rate in
3047 the trace. If set to 200, run at twice the original IO rate. Defaults to 100.
3049 .BI replay_redirect \fR=\fPstr
3050 While replaying I/O patterns using \fBread_iolog\fR the default behavior
3051 is to replay the IOPS onto the major/minor device that each IOP was recorded
3052 from. This is sometimes undesirable because on a different machine those
3053 major/minor numbers can map to a different device. Changing hardware on the
3054 same system can also result in a different major/minor mapping.
3055 \fBreplay_redirect\fR causes all I/Os to be replayed onto the single specified
3056 device regardless of the device it was recorded
3057 from. i.e. `replay_redirect=/dev/sdc' would cause all I/O
3058 in the blktrace or iolog to be replayed onto `/dev/sdc'. This means
3059 multiple devices will be replayed onto a single device, if the trace
3060 contains multiple devices. If you want multiple devices to be replayed
3061 concurrently to multiple redirected devices you must blkparse your trace
3062 into separate traces and replay them with independent fio invocations.
3063 Unfortunately this also breaks the strict time ordering between multiple
3066 .BI replay_align \fR=\fPint
3067 Force alignment of the byte offsets in a trace to this value. The value
3068 must be a power of 2.
3070 .BI replay_scale \fR=\fPint
3071 Scale bye offsets down by this factor when replaying traces. Should most
3072 likely use \fBreplay_align\fR as well.
3073 .SS "Threads, processes and job synchronization"
3075 .BI replay_skip \fR=\fPstr
3076 Sometimes it's useful to skip certain IO types in a replay trace. This could
3077 be, for instance, eliminating the writes in the trace. Or not replaying the
3078 trims/discards, if you are redirecting to a device that doesn't support them.
3079 This option takes a comma separated list of read, write, trim, sync.
3082 Fio defaults to creating jobs by using fork, however if this option is
3083 given, fio will create jobs by using POSIX Threads' function
3084 \fBpthread_create\fR\|(3) to create threads instead.
3086 .BI wait_for \fR=\fPstr
3087 If set, the current job won't be started until all workers of the specified
3088 waitee job are done.
3089 .\" ignore blank line here from HOWTO as it looks normal without it
3090 \fBwait_for\fR operates on the job name basis, so there are a few
3091 limitations. First, the waitee must be defined prior to the waiter job
3092 (meaning no forward references). Second, if a job is being referenced as a
3093 waitee, it must have a unique name (no duplicate waitees).
3096 Run the job with the given nice value. See man \fBnice\fR\|(2).
3097 .\" ignore blank line here from HOWTO as it looks normal without it
3098 On Windows, values less than \-15 set the process class to "High"; \-1 through
3099 \-15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle"
3103 Set the I/O priority value of this job. Linux limits us to a positive value
3104 between 0 and 7, with 0 being the highest. See man
3105 \fBionice\fR\|(1). Refer to an appropriate manpage for other operating
3106 systems since meaning of priority may differ. For per-command priority
3107 setting, see the I/O engine specific `cmdprio_percentage` and
3110 .BI prioclass \fR=\fPint
3111 Set the I/O priority class. See man \fBionice\fR\|(1). For per-command
3112 priority setting, see the I/O engine specific `cmdprio_percentage` and
3113 `cmdprio_class` options.
3115 .BI cpus_allowed \fR=\fPstr
3116 Controls the same options as \fBcpumask\fR, but accepts a textual
3117 specification of the permitted CPUs instead and CPUs are indexed from 0. So
3118 to use CPUs 0 and 5 you would specify `cpus_allowed=0,5'. This option also
3119 allows a range of CPUs to be specified \-\- say you wanted a binding to CPUs
3120 0, 5, and 8 to 15, you would set `cpus_allowed=0,5,8\-15'.
3123 On Windows, when `cpus_allowed' is unset only CPUs from fio's current
3124 processor group will be used and affinity settings are inherited from the
3125 system. An fio build configured to target Windows 7 makes options that set
3126 CPUs processor group aware and values will set both the processor group
3127 and a CPU from within that group. For example, on a system where processor
3128 group 0 has 40 CPUs and processor group 1 has 32 CPUs, `cpus_allowed'
3129 values between 0 and 39 will bind CPUs from processor group 0 and
3130 `cpus_allowed' values between 40 and 71 will bind CPUs from processor
3131 group 1. When using `cpus_allowed_policy=shared' all CPUs specified by a
3132 single `cpus_allowed' option must be from the same processor group. For
3133 Windows fio builds not built for Windows 7, CPUs will only be selected from
3134 (and be relative to) whatever processor group fio happens to be running in
3135 and CPUs from other processor groups cannot be used.
3138 .BI cpus_allowed_policy \fR=\fPstr
3139 Set the policy of how fio distributes the CPUs specified by
3140 \fBcpus_allowed\fR or \fBcpumask\fR. Two policies are supported:
3145 All jobs will share the CPU set specified.
3148 Each job will get a unique CPU from the CPU set.
3151 \fBshared\fR is the default behavior, if the option isn't specified. If
3152 \fBsplit\fR is specified, then fio will assign one cpu per job. If not
3153 enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs
3157 .BI cpumask \fR=\fPint
3158 Set the CPU affinity of this job. The parameter given is a bit mask of
3159 allowed CPUs the job may run on. So if you want the allowed CPUs to be 1
3160 and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man
3161 \fBsched_setaffinity\fR\|(2). This may not work on all supported
3162 operating systems or kernel versions. This option doesn't work well for a
3163 higher CPU count than what you can store in an integer mask, so it can only
3164 control cpus 1\-32. For boxes with larger CPU counts, use
3167 .BI numa_cpu_nodes \fR=\fPstr
3168 Set this job running on specified NUMA nodes' CPUs. The arguments allow
3169 comma delimited list of cpu numbers, A\-B ranges, or `all'. Note, to enable
3170 NUMA options support, fio must be built on a system with libnuma\-dev(el)
3173 .BI numa_mem_policy \fR=\fPstr
3174 Set this job's memory policy and corresponding NUMA nodes. Format of the
3182 `mode' is one of the following memory policies: `default', `prefer',
3183 `bind', `interleave' or `local'. For `default' and `local' memory
3184 policies, no node needs to be specified. For `prefer', only one node is
3185 allowed. For `bind' and `interleave' the `nodelist' may be as
3186 follows: a comma delimited list of numbers, A\-B ranges, or `all'.
3189 .BI cgroup \fR=\fPstr
3190 Add job to this control group. If it doesn't exist, it will be created. The
3191 system must have a mounted cgroup blkio mount point for this to work. If
3192 your system doesn't have it mounted, you can do so with:
3196 # mount \-t cgroup \-o blkio none /cgroup
3200 .BI cgroup_weight \fR=\fPint
3201 Set the weight of the cgroup to this value. See the documentation that comes
3202 with the kernel, allowed values are in the range of 100..1000.
3204 .BI cgroup_nodelete \fR=\fPbool
3205 Normally fio will delete the cgroups it has created after the job
3206 completion. To override this behavior and to leave cgroups around after the
3207 job completion, set `cgroup_nodelete=1'. This can be useful if one wants
3208 to inspect various cgroup files after job completion. Default: false.
3210 .BI flow_id \fR=\fPint
3211 The ID of the flow. If not specified, it defaults to being a global
3212 flow. See \fBflow\fR.
3215 Weight in token-based flow control. If this value is used,
3216 then fio regulates the activity between two or more jobs
3217 sharing the same flow_id.
3218 Fio attempts to keep each job activity proportional to other jobs' activities
3219 in the same flow_id group, with respect to requested weight per job.
3220 That is, if one job has `flow=3', another job has `flow=2'
3221 and another with `flow=1`, then there will be a roughly 3:2:1 ratio
3222 in how much one runs vs the others.
3224 .BI flow_sleep \fR=\fPint
3225 The period of time, in microseconds, to wait after the flow counter
3226 has exceeded its proportion before retrying operations.
3228 .BI stonewall "\fR,\fB wait_for_previous"
3229 Wait for preceding jobs in the job file to exit, before starting this
3230 one. Can be used to insert serialization points in the job file. A stone
3231 wall also implies starting a new reporting group, see
3232 \fBgroup_reporting\fR. Optionally you can use `stonewall=0` to disable or
3233 `stonewall=1` to enable it.
3236 By default, fio will continue running all other jobs when one job finishes.
3237 Sometimes this is not the desired action. Setting \fBexitall\fR will instead
3238 make fio terminate all jobs in the same group, as soon as one job of that
3241 .BI exit_what \fR=\fPstr
3242 By default, fio will continue running all other jobs when one job finishes.
3243 Sometimes this is not the desired action. Setting \fBexitall\fR will instead
3244 make fio terminate all jobs in the same group. The option \fBexit_what\fR
3245 allows you to control which jobs get terminated when \fBexitall\fR is enabled.
3246 The default value is \fBgroup\fR.
3247 The allowed values are:
3252 terminates all jobs.
3255 is the default and does not change the behaviour of \fBexitall\fR.
3258 terminates all currently running jobs across all groups and continues
3259 execution with the next stonewalled group.
3263 .BI exec_prerun \fR=\fPstr
3264 Before running this job, issue the command specified through
3265 \fBsystem\fR\|(3). Output is redirected in a file called `jobname.prerun.txt'.
3267 .BI exec_postrun \fR=\fPstr
3268 After the job completes, issue the command specified though
3269 \fBsystem\fR\|(3). Output is redirected in a file called `jobname.postrun.txt'.
3272 Instead of running as the invoking user, set the user ID to this value
3273 before the thread/process does any work.
3276 Set group ID, see \fBuid\fR.
3280 Do not perform specified workload, only verify data still matches previous
3281 invocation of this workload. This option allows one to check data multiple
3282 times at a later date without overwriting it. This option makes sense only
3283 for workloads that write data, and does not support workloads with the
3284 \fBtime_based\fR option set.
3286 .BI do_verify \fR=\fPbool
3287 Run the verify phase after a write phase. Only valid if \fBverify\fR is
3290 .BI verify \fR=\fPstr
3291 If writing to a file, fio can verify the file contents after each iteration
3292 of the job. Each verification method also implies verification of special
3293 header, which is written to the beginning of each block. This header also
3294 includes meta information, like offset of the block, block number, timestamp
3295 when block was written, etc. \fBverify\fR can be combined with
3296 \fBverify_pattern\fR option. The allowed values are:
3301 Use an md5 sum of the data area and store it in the header of
3305 Use an experimental crc64 sum of the data area and store it in the
3306 header of each block.
3309 Use a crc32c sum of the data area and store it in the header of
3310 each block. This will automatically use hardware acceleration
3311 (e.g. SSE4.2 on an x86 or CRC crypto extensions on ARM64) but will
3312 fall back to software crc32c if none is found. Generally the
3313 fastest checksum fio supports when hardware accelerated.
3319 Use a crc32 sum of the data area and store it in the header of each
3323 Use a crc16 sum of the data area and store it in the header of each
3327 Use a crc7 sum of the data area and store it in the header of each
3331 Use xxhash as the checksum function. Generally the fastest software
3332 checksum that fio supports.
3335 Use sha512 as the checksum function.
3338 Use sha256 as the checksum function.
3341 Use optimized sha1 as the checksum function.
3344 Use optimized sha3\-224 as the checksum function.
3347 Use optimized sha3\-256 as the checksum function.
3350 Use optimized sha3\-384 as the checksum function.
3353 Use optimized sha3\-512 as the checksum function.
3356 This option is deprecated, since now meta information is included in
3357 generic verification header and meta verification happens by
3358 default. For detailed information see the description of the
3359 \fBverify\fR setting. This option is kept because of
3360 compatibility's sake with old configurations. Do not use it.
3363 Verify a strict pattern. Normally fio includes a header with some
3364 basic information and checksumming, but if this option is set, only
3365 the specific pattern set with \fBverify_pattern\fR is verified.
3368 Only pretend to verify. Useful for testing internals with
3369 `ioengine=null', not for much else.
3372 This option can be used for repeated burn\-in tests of a system to make sure
3373 that the written data is also correctly read back. If the data direction
3374 given is a read or random read, fio will assume that it should verify a
3375 previously written file. If the data direction includes any form of write,
3376 the verify will be of the newly written data.
3378 To avoid false verification errors, do not use the norandommap option when
3379 verifying data with async I/O engines and I/O depths > 1. Or use the
3380 norandommap and the lfsr random generator together to avoid writing to the
3381 same offset with multiple outstanding I/Os.
3384 .BI verify_offset \fR=\fPint
3385 Swap the verification header with data somewhere else in the block before
3386 writing. It is swapped back before verifying.
3388 .BI verify_interval \fR=\fPint
3389 Write the verification header at a finer granularity than the
3390 \fBblocksize\fR. It will be written for chunks the size of
3391 \fBverify_interval\fR. \fBblocksize\fR should divide this evenly.
3393 .BI verify_pattern \fR=\fPstr
3394 If set, fio will fill the I/O buffers with this pattern. Fio defaults to
3395 filling with totally random bytes, but sometimes it's interesting to fill
3396 with a known pattern for I/O verification purposes. Depending on the width
3397 of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time (it can
3398 be either a decimal or a hex number). The \fBverify_pattern\fR if larger than
3399 a 32\-bit quantity has to be a hex number that starts with either "0x" or
3400 "0X". Use with \fBverify\fR. Also, \fBverify_pattern\fR supports %o
3401 format, which means that for each block offset will be written and then
3402 verified back, e.g.:
3409 Or use combination of everything:
3412 verify_pattern=0xff%o"abcd"\-12
3416 .BI verify_fatal \fR=\fPbool
3417 Normally fio will keep checking the entire contents before quitting on a
3418 block verification failure. If this option is set, fio will exit the job on
3419 the first observed failure. Default: false.
3421 .BI verify_dump \fR=\fPbool
3422 If set, dump the contents of both the original data block and the data block
3423 we read off disk to files. This allows later analysis to inspect just what
3424 kind of data corruption occurred. Off by default.
3426 .BI verify_async \fR=\fPint
3427 Fio will normally verify I/O inline from the submitting thread. This option
3428 takes an integer describing how many async offload threads to create for I/O
3429 verification instead, causing fio to offload the duty of verifying I/O
3430 contents to one or more separate threads. If using this offload option, even
3431 sync I/O engines can benefit from using an \fBiodepth\fR setting higher
3432 than 1, as it allows them to have I/O in flight while verifies are running.
3433 Defaults to 0 async threads, i.e. verification is not asynchronous.
3435 .BI verify_async_cpus \fR=\fPstr
3436 Tell fio to set the given CPU affinity on the async I/O verification
3437 threads. See \fBcpus_allowed\fR for the format used.
3439 .BI verify_backlog \fR=\fPint
3440 Fio will normally verify the written contents of a job that utilizes verify
3441 once that job has completed. In other words, everything is written then
3442 everything is read back and verified. You may want to verify continually
3443 instead for a variety of reasons. Fio stores the meta data associated with
3444 an I/O block in memory, so for large verify workloads, quite a bit of memory
3445 would be used up holding this meta data. If this option is enabled, fio will
3446 write only N blocks before verifying these blocks.
3448 .BI verify_backlog_batch \fR=\fPint
3449 Control how many blocks fio will verify if \fBverify_backlog\fR is
3450 set. If not set, will default to the value of \fBverify_backlog\fR
3451 (meaning the entire queue is read back and verified). If
3452 \fBverify_backlog_batch\fR is less than \fBverify_backlog\fR then not all
3453 blocks will be verified, if \fBverify_backlog_batch\fR is larger than
3454 \fBverify_backlog\fR, some blocks will be verified more than once.
3456 .BI verify_state_save \fR=\fPbool
3457 When a job exits during the write phase of a verify workload, save its
3458 current state. This allows fio to replay up until that point, if the verify
3459 state is loaded for the verify read phase. The format of the filename is,
3464 <type>\-<jobname>\-<jobindex>\-verify.state.
3467 <type> is "local" for a local run, "sock" for a client/server socket
3468 connection, and "ip" (192.168.0.1, for instance) for a networked
3469 client/server connection. Defaults to true.
3472 .BI verify_state_load \fR=\fPbool
3473 If a verify termination trigger was used, fio stores the current write state
3474 of each thread. This can be used at verification time so that fio knows how
3475 far it should verify. Without this information, fio will run a full
3476 verification pass, according to the settings in the job file used. Default
3479 .BI trim_percentage \fR=\fPint
3480 Number of verify blocks to discard/trim.
3482 .BI trim_verify_zero \fR=\fPbool
3483 Verify that trim/discarded blocks are returned as zeros.
3485 .BI trim_backlog \fR=\fPint
3486 Verify that trim/discarded blocks are returned as zeros.
3488 .BI trim_backlog_batch \fR=\fPint
3489 Trim this number of I/O blocks.
3491 .BI experimental_verify \fR=\fPbool
3492 Enable experimental verification. Standard verify records I/O metadata for
3493 later use during the verification phase. Experimental verify instead resets the
3494 file after the write phase and then replays I/Os for the verification phase.
3497 .BI steadystate \fR=\fPstr:float "\fR,\fP ss" \fR=\fPstr:float
3498 Define the criterion and limit for assessing steady state performance. The
3499 first parameter designates the criterion whereas the second parameter sets
3500 the threshold. When the criterion falls below the threshold for the
3501 specified duration, the job will stop. For example, `iops_slope:0.1%' will
3502 direct fio to terminate the job when the least squares regression slope
3503 falls below 0.1% of the mean IOPS. If \fBgroup_reporting\fR is enabled
3504 this will apply to all jobs in the group. Below is the list of available
3505 steady state assessment criteria. All assessments are carried out using only
3506 data from the rolling collection window. Threshold limits can be expressed
3507 as a fixed value or as a percentage of the mean in the collection window.
3510 When using this feature, most jobs should include the \fBtime_based\fR
3511 and \fBruntime\fR options or the \fBloops\fR option so that fio does not
3512 stop running after it has covered the full size of the specified file(s)
3518 Collect IOPS data. Stop the job if all individual IOPS measurements
3519 are within the specified limit of the mean IOPS (e.g., `iops:2'
3520 means that all individual IOPS values must be within 2 of the mean,
3521 whereas `iops:0.2%' means that all individual IOPS values must be
3522 within 0.2% of the mean IOPS to terminate the job).
3525 Collect IOPS data and calculate the least squares regression
3526 slope. Stop the job if the slope falls below the specified limit.
3529 Collect bandwidth data. Stop the job if all individual bandwidth
3530 measurements are within the specified limit of the mean bandwidth.
3533 Collect bandwidth data and calculate the least squares regression
3534 slope. Stop the job if the slope falls below the specified limit.
3538 .BI steadystate_duration \fR=\fPtime "\fR,\fP ss_dur" \fR=\fPtime
3539 A rolling window of this duration will be used to judge whether steady state
3540 has been reached. Data will be collected every \fBss_interval\fR. The default
3541 is 0 which disables steady state detection. When the unit is omitted, the value
3542 is interpreted in seconds.
3544 .BI steadystate_ramp_time \fR=\fPtime "\fR,\fP ss_ramp" \fR=\fPtime
3545 Allow the job to run for the specified duration before beginning data
3546 collection for checking the steady state job termination criterion. The
3547 default is 0. When the unit is omitted, the value is interpreted in seconds.
3549 .BI steadystate_check_interval \fR=\fPtime "\fR,\fP ss_interval" \fR=\fPtime
3550 The values suring the rolling window will be collected with a period of this
3551 value. If \fBss_interval\fR is 30s and \fBss_dur\fR is 300s, 10 measurements
3552 will be taken. Default is 1s but that might not converge, especially for slower
3553 devices, so set this accordingly. When the unit is omitted, the value is
3554 interpreted in seconds.
3555 .SS "Measurements and reporting"
3557 .BI per_job_logs \fR=\fPbool
3558 If set, this generates bw/clat/iops log with per file private filenames. If
3559 not set, jobs with identical names will share the log filename. Default:
3563 It may sometimes be interesting to display statistics for groups of jobs as
3564 a whole instead of for each individual job. This is especially true if
3565 \fBnumjobs\fR is used; looking at individual thread/process output
3566 quickly becomes unwieldy. To see the final report per-group instead of
3567 per-job, use \fBgroup_reporting\fR. Jobs in a file will be part of the
3568 same reporting group, unless if separated by a \fBstonewall\fR, or by
3569 using \fBnew_group\fR.
3572 Start a new reporting group. See: \fBgroup_reporting\fR. If not given,
3573 all jobs in a file will be part of the same reporting group, unless
3574 separated by a \fBstonewall\fR.
3576 .BI stats \fR=\fPbool
3577 By default, fio collects and shows final output results for all jobs
3578 that run. If this option is set to 0, then fio will ignore it in
3579 the final stat output.
3581 .BI write_bw_log \fR=\fPstr
3582 If given, write a bandwidth log for this job. Can be used to store data of
3583 the bandwidth of the jobs in their lifetime.
3586 If no str argument is given, the default filename of
3587 `jobname_type.x.log' is used. Even when the argument is given, fio
3588 will still append the type of log. So if one specifies:
3594 The actual log name will be `foo_bw.x.log' where `x' is the index
3595 of the job (1..N, where N is the number of jobs). If
3596 \fBper_job_logs\fR is false, then the filename will not include the
3599 The included \fBfio_generate_plots\fR script uses gnuplot to turn these
3600 text files into nice graphs. See the \fBLOG FILE FORMATS\fR section for how data is
3601 structured within the file.
3604 .BI write_lat_log \fR=\fPstr
3605 Same as \fBwrite_bw_log\fR, except this option creates I/O
3606 submission (e.g., `name_slat.x.log'), completion (e.g.,
3607 `name_clat.x.log'), and total (e.g., `name_lat.x.log') latency
3608 files instead. See \fBwrite_bw_log\fR for details about the
3609 filename format and the \fBLOG FILE FORMATS\fR section for how data is structured
3612 .BI write_hist_log \fR=\fPstr
3613 Same as \fBwrite_bw_log\fR but writes an I/O completion latency
3614 histogram file (e.g., `name_hist.x.log') instead. Note that this
3615 file will be empty unless \fBlog_hist_msec\fR has also been set.
3616 See \fBwrite_bw_log\fR for details about the filename format and
3617 the \fBLOG FILE FORMATS\fR section for how data is structured
3620 .BI write_iops_log \fR=\fPstr
3621 Same as \fBwrite_bw_log\fR, but writes an IOPS file (e.g.
3622 `name_iops.x.log`) instead. Because fio defaults to individual
3623 I/O logging, the value entry in the IOPS log will be 1 unless windowed
3624 logging (see \fBlog_avg_msec\fR) has been enabled. See
3625 \fBwrite_bw_log\fR for details about the filename format and \fBLOG
3626 FILE FORMATS\fR for how data is structured within the file.
3628 .BI log_entries \fR=\fPint
3629 By default, fio will log an entry in the iops, latency, or bw log for
3630 every I/O that completes. The initial number of I/O log entries is 1024.
3631 When the log entries are all used, new log entries are dynamically
3632 allocated. This dynamic log entry allocation may negatively impact
3633 time-related statistics such as I/O tail latencies (e.g. 99.9th percentile
3634 completion latency). This option allows specifying a larger initial
3635 number of log entries to avoid run-time allocation of new log entries,
3636 resulting in more precise time-related I/O statistics.
3637 Also see \fBlog_avg_msec\fR as well. Defaults to 1024.
3639 .BI log_avg_msec \fR=\fPint
3640 By default, fio will log an entry in the iops, latency, or bw log for every
3641 I/O that completes. When writing to the disk log, that can quickly grow to a
3642 very large size. Setting this option makes fio average the each log entry
3643 over the specified period of time, reducing the resolution of the log. See
3644 \fBlog_max_value\fR as well. Defaults to 0, logging all entries.
3645 Also see \fBLOG FILE FORMATS\fR section.
3647 .BI log_hist_msec \fR=\fPint
3648 Same as \fBlog_avg_msec\fR, but logs entries for completion latency
3649 histograms. Computing latency percentiles from averages of intervals using
3650 \fBlog_avg_msec\fR is inaccurate. Setting this option makes fio log
3651 histogram entries over the specified period of time, reducing log sizes for
3652 high IOPS devices while retaining percentile accuracy. See
3653 \fBlog_hist_coarseness\fR and \fBwrite_hist_log\fR as well.
3654 Defaults to 0, meaning histogram logging is disabled.
3656 .BI log_hist_coarseness \fR=\fPint
3657 Integer ranging from 0 to 6, defining the coarseness of the resolution of
3658 the histogram logs enabled with \fBlog_hist_msec\fR. For each increment
3659 in coarseness, fio outputs half as many bins. Defaults to 0, for which
3660 histogram logs contain 1216 latency bins. See \fBLOG FILE FORMATS\fR section.
3662 .BI log_max_value \fR=\fPbool
3663 If \fBlog_avg_msec\fR is set, fio logs the average over that window. If
3664 you instead want to log the maximum value, set this option to 1. Defaults to
3665 0, meaning that averaged values are logged.
3667 .BI log_offset \fR=\fPbool
3668 If this is set, the iolog options will include the byte offset for the I/O
3669 entry as well as the other data values. Defaults to 0 meaning that
3670 offsets are not present in logs. Also see \fBLOG FILE FORMATS\fR section.
3672 .BI log_prio \fR=\fPbool
3673 If this is set, the iolog options will include the I/O priority for the I/O
3674 entry as well as the other data values. Defaults to 0 meaning that
3675 I/O priorities are not present in logs. Also see \fBLOG FILE FORMATS\fR section.
3677 .BI log_compression \fR=\fPint
3678 If this is set, fio will compress the I/O logs as it goes, to keep the
3679 memory footprint lower. When a log reaches the specified size, that chunk is
3680 removed and compressed in the background. Given that I/O logs are fairly
3681 highly compressible, this yields a nice memory savings for longer runs. The
3682 downside is that the compression will consume some background CPU cycles, so
3683 it may impact the run. This, however, is also true if the logging ends up
3684 consuming most of the system memory. So pick your poison. The I/O logs are
3685 saved normally at the end of a run, by decompressing the chunks and storing
3686 them in the specified log file. This feature depends on the availability of
3689 .BI log_compression_cpus \fR=\fPstr
3690 Define the set of CPUs that are allowed to handle online log compression for
3691 the I/O jobs. This can provide better isolation between performance
3692 sensitive jobs, and background compression work. See \fBcpus_allowed\fR for
3695 .BI log_store_compressed \fR=\fPbool
3696 If set, fio will store the log files in a compressed format. They can be
3697 decompressed with fio, using the \fB\-\-inflate\-log\fR command line
3698 parameter. The files will be stored with a `.fz' suffix.
3700 .BI log_unix_epoch \fR=\fPbool
3701 If set, fio will log Unix timestamps to the log files produced by enabling
3702 write_type_log for each log type, instead of the default zero-based
3705 .BI log_alternate_epoch \fR=\fPbool
3706 If set, fio will log timestamps based on the epoch used by the clock specified
3707 in the \fBlog_alternate_epoch_clock_id\fR option, to the log files produced by
3708 enabling write_type_log for each log type, instead of the default zero-based
3711 .BI log_alternate_epoch_clock_id \fR=\fPint
3712 Specifies the clock_id to be used by clock_gettime to obtain the alternate epoch
3713 if either \fBBlog_unix_epoch\fR or \fBlog_alternate_epoch\fR are true. Otherwise has no
3714 effect. Default value is 0, or CLOCK_REALTIME.
3716 .BI block_error_percentiles \fR=\fPbool
3717 If set, record errors in trim block-sized units from writes and trims and
3718 output a histogram of how many trims it took to get to errors, and what kind
3719 of error was encountered.
3721 .BI bwavgtime \fR=\fPint
3722 Average the calculated bandwidth over the given time. Value is specified in
3723 milliseconds. If the job also does bandwidth logging through
3724 \fBwrite_bw_log\fR, then the minimum of this option and
3725 \fBlog_avg_msec\fR will be used. Default: 500ms.
3727 .BI iopsavgtime \fR=\fPint
3728 Average the calculated IOPS over the given time. Value is specified in
3729 milliseconds. If the job also does IOPS logging through
3730 \fBwrite_iops_log\fR, then the minimum of this option and
3731 \fBlog_avg_msec\fR will be used. Default: 500ms.
3733 .BI disk_util \fR=\fPbool
3734 Generate disk utilization statistics, if the platform supports it.
3737 .BI disable_lat \fR=\fPbool
3738 Disable measurements of total latency numbers. Useful only for cutting back
3739 the number of calls to \fBgettimeofday\fR\|(2), as that does impact
3740 performance at really high IOPS rates. Note that to really get rid of a
3741 large amount of these calls, this option must be used with
3742 \fBdisable_slat\fR and \fBdisable_bw_measurement\fR as well.
3744 .BI disable_clat \fR=\fPbool
3745 Disable measurements of completion latency numbers. See
3748 .BI disable_slat \fR=\fPbool
3749 Disable measurements of submission latency numbers. See
3752 .BI disable_bw_measurement \fR=\fPbool "\fR,\fP disable_bw" \fR=\fPbool
3753 Disable measurements of throughput/bandwidth numbers. See
3756 .BI slat_percentiles \fR=\fPbool
3757 Report submission latency percentiles. Submission latency is not recorded
3758 for synchronous ioengines.
3760 .BI clat_percentiles \fR=\fPbool
3761 Report completion latency percentiles.
3763 .BI lat_percentiles \fR=\fPbool
3764 Report total latency percentiles. Total latency is the sum of submission
3765 latency and completion latency.
3767 .BI percentile_list \fR=\fPfloat_list
3768 Overwrite the default list of percentiles for latencies and the
3769 block error histogram. Each number is a floating point number in the range
3770 (0,100], and the maximum length of the list is 20. Use ':' to separate the
3771 numbers. For example, `\-\-percentile_list=99.5:99.9' will cause fio to
3772 report the latency durations below which 99.5% and 99.9% of the observed
3773 latencies fell, respectively.
3775 .BI significant_figures \fR=\fPint
3776 If using \fB\-\-output\-format\fR of `normal', set the significant figures
3777 to this value. Higher values will yield more precise IOPS and throughput
3778 units, while lower values will round. Requires a minimum value of 1 and a
3779 maximum value of 10. Defaults to 4.
3780 .SS "Error handling"
3782 .BI exitall_on_error
3783 When one job finishes in error, terminate the rest. The default is to wait
3784 for each job to finish.
3786 .BI continue_on_error \fR=\fPstr
3787 Normally fio will exit the job on the first observed failure. If this option
3788 is set, fio will continue the job when there is a 'non-fatal error' (EIO or
3789 EILSEQ) until the runtime is exceeded or the I/O size specified is
3790 completed. If this option is used, there are two more stats that are
3791 appended, the total error count and the first error. The error field given
3792 in the stats is the first error that was hit during the run.
3795 Note: a write error from the device may go unnoticed by fio when using buffered
3796 IO, as the write() (or similar) system call merely dirties the kernel pages,
3797 unless `sync' or `direct' is used. Device IO errors occur when the dirty data is
3798 actually written out to disk. If fully sync writes aren't desirable, `fsync' or
3799 `fdatasync' can be used as well. This is specific to writes, as reads are always
3803 The allowed values are:
3808 Exit on any I/O or verify errors.
3811 Continue on read errors, exit on all others.
3814 Continue on write errors, exit on all others.
3817 Continue on any I/O error, exit on all others.
3820 Continue on verify errors, exit on all others.
3823 Continue on all errors.
3826 Backward-compatible alias for 'none'.
3829 Backward-compatible alias for 'all'.
3833 .BI ignore_error \fR=\fPstr
3834 Sometimes you want to ignore some errors during test in that case you can
3835 specify error list for each error type, instead of only being able to
3836 ignore the default 'non-fatal error' using \fBcontinue_on_error\fR.
3837 `ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST' errors for
3838 given error type is separated with ':'. Error may be symbol ('ENOSPC', 'ENOMEM')
3839 or integer. Example:
3843 ignore_error=EAGAIN,ENOSPC:122
3846 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from
3847 WRITE. This option works by overriding \fBcontinue_on_error\fR with
3848 the list of errors for each error type if any.
3851 .BI error_dump \fR=\fPbool
3852 If set dump every error even if it is non fatal, true by default. If
3853 disabled only fatal error will be dumped.
3854 .SS "Running predefined workloads"
3855 Fio includes predefined profiles that mimic the I/O workloads generated by
3858 .BI profile \fR=\fPstr
3859 The predefined workload to run. Current profiles are:
3864 Threaded I/O bench (tiotest/tiobench) like workload.
3867 Aerospike Certification Tool (ACT) like workload.
3871 To view a profile's additional options use \fB\-\-cmdhelp\fR after specifying
3872 the profile. For example:
3875 $ fio \-\-profile=act \-\-cmdhelp
3877 .SS "Act profile options"
3879 .BI device\-names \fR=\fPstr
3883 ACT load multiplier. Default: 1.
3885 .BI test\-duration\fR=\fPtime
3886 How long the entire test takes to run. When the unit is omitted, the value
3887 is given in seconds. Default: 24h.
3889 .BI threads\-per\-queue\fR=\fPint
3890 Number of read I/O threads per device. Default: 8.
3892 .BI read\-req\-num\-512\-blocks\fR=\fPint
3893 Number of 512B blocks to read at the time. Default: 3.
3895 .BI large\-block\-op\-kbytes\fR=\fPint
3896 Size of large block ops in KiB (writes). Default: 131072.
3899 Set to run ACT prep phase.
3900 .SS "Tiobench profile options"
3906 Block size in bytes. Default: 4096.
3908 .BI numruns\fR=\fPint
3914 .BI threads\fR=\fPint
3917 Fio spits out a lot of output. While running, fio will display the status of the
3918 jobs created. An example of that would be:
3921 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]
3924 The characters inside the first set of square brackets denote the current status of
3925 each thread. The first character is the first job defined in the job file, and so
3926 forth. The possible values (in typical life cycle order) are:
3931 Thread setup, but not started.
3937 Thread initialized, waiting or generating necessary data.
3940 Thread running pre-reading file(s).
3943 Thread is in ramp period.
3946 Running, doing sequential reads.
3949 Running, doing random reads.
3952 Running, doing sequential writes.
3955 Running, doing random writes.
3958 Running, doing mixed sequential reads/writes.
3961 Running, doing mixed random reads/writes.
3964 Running, doing sequential trims.
3967 Running, doing random trims.
3970 Running, currently waiting for \fBfsync\fR\|(2).
3973 Running, doing verification of written data.
3979 Thread exited, not reaped by main thread yet.
3985 Thread reaped, exited with an error.
3988 Thread reaped, exited due to signal.
3992 Fio will condense the thread string as not to take up more space on the command
3993 line than needed. For instance, if you have 10 readers and 10 writers running,
3994 the output would look like this:
3997 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]
4000 Note that the status string is displayed in order, so it's possible to tell which of
4001 the jobs are currently doing what. In the example above this means that jobs 1\-\-10
4002 are readers and 11\-\-20 are writers.
4004 The other values are fairly self explanatory \-\- number of threads currently
4005 running and doing I/O, the number of currently open files (f=), the estimated
4006 completion percentage, the rate of I/O since last check (read speed listed first,
4007 then write speed and optionally trim speed) in terms of bandwidth and IOPS,
4008 and time to completion for the current running group. It's impossible to estimate
4009 runtime of the following groups (if any).
4011 When fio is done (or interrupted by Ctrl\-C), it will show the data for
4012 each thread, group of threads, and disks in that order. For each overall thread (or
4013 group) the output looks like:
4016 Client1: (groupid=0, jobs=1): err= 0: pid=16109: Sat Jun 24 12:07:54 2017
4017 write: IOPS=88, BW=623KiB/s (638kB/s)(30.4MiB/50032msec)
4018 slat (nsec): min=500, max=145500, avg=8318.00, stdev=4781.50
4019 clat (usec): min=170, max=78367, avg=4019.02, stdev=8293.31
4020 lat (usec): min=174, max=78375, avg=4027.34, stdev=8291.79
4021 clat percentiles (usec):
4022 | 1.00th=[ 302], 5.00th=[ 326], 10.00th=[ 343], 20.00th=[ 363],
4023 | 30.00th=[ 392], 40.00th=[ 404], 50.00th=[ 416], 60.00th=[ 445],
4024 | 70.00th=[ 816], 80.00th=[ 6718], 90.00th=[12911], 95.00th=[21627],
4025 | 99.00th=[43779], 99.50th=[51643], 99.90th=[68682], 99.95th=[72877],
4027 bw ( KiB/s): min= 532, max= 686, per=0.10%, avg=622.87, stdev=24.82, samples= 100
4028 iops : min= 76, max= 98, avg=88.98, stdev= 3.54, samples= 100
4029 lat (usec) : 250=0.04%, 500=64.11%, 750=4.81%, 1000=2.79%
4030 lat (msec) : 2=4.16%, 4=1.84%, 10=4.90%, 20=11.33%, 50=5.37%
4031 lat (msec) : 100=0.65%
4032 cpu : usr=0.27%, sys=0.18%, ctx=12072, majf=0, minf=21
4033 IO depths : 1=85.0%, 2=13.1%, 4=1.8%, 8=0.1%, 16=0.0%, 32=0.0%, >=64=0.0%
4034 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
4035 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
4036 issued rwt: total=0,4450,0, short=0,0,0, dropped=0,0,0
4037 latency : target=0, window=0, percentile=100.00%, depth=8
4040 The job name (or first job's name when using \fBgroup_reporting\fR) is printed,
4041 along with the group id, count of jobs being aggregated, last error id seen (which
4042 is 0 when there are no errors), pid/tid of that thread and the time the job/group
4043 completed. Below are the I/O statistics for each data direction performed (showing
4044 writes in the example above). In the order listed, they denote:
4048 The string before the colon shows the I/O direction the statistics
4049 are for. \fIIOPS\fR is the average I/Os performed per second. \fIBW\fR
4050 is the average bandwidth rate shown as: value in power of 2 format
4051 (value in power of 10 format). The last two values show: (total
4052 I/O performed in power of 2 format / \fIruntime\fR of that thread).
4055 Submission latency (\fImin\fR being the minimum, \fImax\fR being the
4056 maximum, \fIavg\fR being the average, \fIstdev\fR being the standard
4057 deviation). This is the time it took to submit the I/O. For
4058 sync I/O this row is not displayed as the slat is really the
4059 completion latency (since queue/complete is one operation there).
4060 This value can be in nanoseconds, microseconds or milliseconds \-\-\-
4061 fio will choose the most appropriate base and print that (in the
4062 example above nanoseconds was the best scale). Note: in \fB\-\-minimal\fR mode
4063 latencies are always expressed in microseconds.
4066 Completion latency. Same names as slat, this denotes the time from
4067 submission to completion of the I/O pieces. For sync I/O, clat will
4068 usually be equal (or very close) to 0, as the time from submit to
4069 complete is basically just CPU time (I/O has already been done, see slat
4073 Total latency. Same names as slat and clat, this denotes the time from
4074 when fio created the I/O unit to completion of the I/O operation.
4077 Bandwidth statistics based on samples. Same names as the xlat stats,
4078 but also includes the number of samples taken (\fIsamples\fR) and an
4079 approximate percentage of total aggregate bandwidth this thread
4080 received in its group (\fIper\fR). This last value is only really
4081 useful if the threads in this group are on the same disk, since they
4082 are then competing for disk access.
4085 IOPS statistics based on samples. Same names as \fBbw\fR.
4087 .B lat (nsec/usec/msec)
4088 The distribution of I/O completion latencies. This is the time from when
4089 I/O leaves fio and when it gets completed. Unlike the separate
4090 read/write/trim sections above, the data here and in the remaining
4091 sections apply to all I/Os for the reporting group. 250=0.04% means that
4092 0.04% of the I/Os completed in under 250us. 500=64.11% means that 64.11%
4093 of the I/Os required 250 to 499us for completion.
4096 CPU usage. User and system time, along with the number of context
4097 switches this thread went through, usage of system and user time, and
4098 finally the number of major and minor page faults. The CPU utilization
4099 numbers are averages for the jobs in that reporting group, while the
4100 context and fault counters are summed.
4103 The distribution of I/O depths over the job lifetime. The numbers are
4104 divided into powers of 2 and each entry covers depths from that value
4105 up to those that are lower than the next entry \-\- e.g., 16= covers
4106 depths from 16 to 31. Note that the range covered by a depth
4107 distribution entry can be different to the range covered by the
4108 equivalent \fBsubmit\fR/\fBcomplete\fR distribution entry.
4111 How many pieces of I/O were submitting in a single submit call. Each
4112 entry denotes that amount and below, until the previous entry \-\- e.g.,
4113 16=100% means that we submitted anywhere between 9 to 16 I/Os per submit
4114 call. Note that the range covered by a \fBsubmit\fR distribution entry can
4115 be different to the range covered by the equivalent depth distribution
4119 Like the above \fBsubmit\fR number, but for completions instead.
4122 The number of \fBread/write/trim\fR requests issued, and how many of them were
4126 These values are for \fBlatency_target\fR and related options. When
4127 these options are engaged, this section describes the I/O depth required
4128 to meet the specified latency target.
4131 After each client has been listed, the group statistics are printed. They
4132 will look like this:
4135 Run status group 0 (all jobs):
4136 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
4137 WRITE: bw=1231KiB/s (1261kB/s), 616KiB/s\-621KiB/s (630kB/s\-636kB/s), io=64.0MiB (67.1MB), run=52747\-53223msec
4140 For each data direction it prints:
4144 Aggregate bandwidth of threads in this group followed by the
4145 minimum and maximum bandwidth of all the threads in this group.
4146 Values outside of brackets are power-of-2 format and those
4147 within are the equivalent value in a power-of-10 format.
4150 Aggregate I/O performed of all threads in this group. The
4151 format is the same as \fBbw\fR.
4154 The smallest and longest runtimes of the threads in this group.
4157 And finally, the disk statistics are printed. This is Linux specific.
4158 They will look like this:
4161 Disk stats (read/write):
4162 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
4165 Each value is printed for both reads and writes, with reads first. The
4170 Number of I/Os performed by all groups.
4173 Number of merges performed by the I/O scheduler.
4176 Number of ticks we kept the disk busy.
4179 Total time spent in the disk queue.
4182 The disk utilization. A value of 100% means we kept the disk
4183 busy constantly, 50% would be a disk idling half of the time.
4186 It is also possible to get fio to dump the current output while it is running,
4187 without terminating the job. To do that, send fio the USR1 signal. You can
4188 also get regularly timed dumps by using the \fB\-\-status\-interval\fR
4189 parameter, or by creating a file in `/tmp' named
4190 `fio\-dump\-status'. If fio sees this file, it will unlink it and dump the
4191 current output status.
4193 For scripted usage where you typically want to generate tables or graphs of the
4194 results, fio can output the results in a semicolon separated format. The format
4195 is one long line of values, such as:
4198 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%
4199 A description of this job goes here.
4202 The job description (if provided) follows on a second line for terse v2.
4203 It appears on the same line for other terse versions.
4205 To enable terse output, use the \fB\-\-minimal\fR or
4206 `\-\-output\-format=terse' command line options. The
4207 first value is the version of the terse output format. If the output has to be
4208 changed for some reason, this number will be incremented by 1 to signify that
4211 Split up, the format is as follows (comments in brackets denote when a
4212 field was introduced or whether it's specific to some terse version):
4215 terse version, fio version [v3], jobname, groupid, error
4224 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
4225 Submission latency: min, max, mean, stdev (usec)
4226 Completion latency: min, max, mean, stdev (usec)
4227 Completion latency percentiles: 20 fields (see below)
4228 Total latency: min, max, mean, stdev (usec)
4229 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
4230 IOPS [v5]: min, max, mean, stdev, number of samples
4239 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
4240 Submission latency: min, max, mean, stdev (usec)
4241 Completion latency: min, max, mean, stdev (usec)
4242 Completion latency percentiles: 20 fields (see below)
4243 Total latency: min, max, mean, stdev (usec)
4244 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
4245 IOPS [v5]: min, max, mean, stdev, number of samples
4250 TRIM status [all but version 3]:
4254 Fields are similar to \fBREAD/WRITE\fR status.
4263 user, system, context switches, major faults, minor faults
4272 <=1, 2, 4, 8, 16, 32, >=64
4277 I/O latencies microseconds:
4281 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
4286 I/O latencies milliseconds:
4290 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
4295 Disk utilization [v3]:
4299 disk name, read ios, write ios, read merges, write merges, read ticks, write ticks, time spent in queue, disk utilization percentage
4304 Additional Info (dependent on continue_on_error, default off):
4308 total # errors, first error code
4313 Additional Info (dependent on description being set):
4320 Completion latency percentiles can be a grouping of up to 20 sets, so for the
4321 terse output fio writes all of them. Each field will look like this:
4327 which is the Xth percentile, and the `usec' latency associated with it.
4329 For \fBDisk utilization\fR, all disks used by fio are shown. So for each disk there
4330 will be a disk utilization section.
4332 Below is a single line containing short names for each of the fields in the
4333 minimal output v3, separated by semicolons:
4336 terse_version_3;fio_version;jobname;groupid;error;read_kb;read_bandwidth_kb;read_iops;read_runtime_ms;read_slat_min_us;read_slat_max_us;read_slat_mean_us;read_slat_dev_us;read_clat_min_us;read_clat_max_us;read_clat_mean_us;read_clat_dev_us;read_clat_pct01;read_clat_pct02;read_clat_pct03;read_clat_pct04;read_clat_pct05;read_clat_pct06;read_clat_pct07;read_clat_pct08;read_clat_pct09;read_clat_pct10;read_clat_pct11;read_clat_pct12;read_clat_pct13;read_clat_pct14;read_clat_pct15;read_clat_pct16;read_clat_pct17;read_clat_pct18;read_clat_pct19;read_clat_pct20;read_tlat_min_us;read_lat_max_us;read_lat_mean_us;read_lat_dev_us;read_bw_min_kb;read_bw_max_kb;read_bw_agg_pct;read_bw_mean_kb;read_bw_dev_kb;write_kb;write_bandwidth_kb;write_iops;write_runtime_ms;write_slat_min_us;write_slat_max_us;write_slat_mean_us;write_slat_dev_us;write_clat_min_us;write_clat_max_us;write_clat_mean_us;write_clat_dev_us;write_clat_pct01;write_clat_pct02;write_clat_pct03;write_clat_pct04;write_clat_pct05;write_clat_pct06;write_clat_pct07;write_clat_pct08;write_clat_pct09;write_clat_pct10;write_clat_pct11;write_clat_pct12;write_clat_pct13;write_clat_pct14;write_clat_pct15;write_clat_pct16;write_clat_pct17;write_clat_pct18;write_clat_pct19;write_clat_pct20;write_tlat_min_us;write_lat_max_us;write_lat_mean_us;write_lat_dev_us;write_bw_min_kb;write_bw_max_kb;write_bw_agg_pct;write_bw_mean_kb;write_bw_dev_kb;cpu_user;cpu_sys;cpu_csw;cpu_mjf;cpu_minf;iodepth_1;iodepth_2;iodepth_4;iodepth_8;iodepth_16;iodepth_32;iodepth_64;lat_2us;lat_4us;lat_10us;lat_20us;lat_50us;lat_100us;lat_250us;lat_500us;lat_750us;lat_1000us;lat_2ms;lat_4ms;lat_10ms;lat_20ms;lat_50ms;lat_100ms;lat_250ms;lat_500ms;lat_750ms;lat_1000ms;lat_2000ms;lat_over_2000ms;disk_name;disk_read_iops;disk_write_iops;disk_read_merges;disk_write_merges;disk_read_ticks;write_ticks;disk_queue_time;disk_util
4339 In client/server mode terse output differs from what appears when jobs are run
4340 locally. Disk utilization data is omitted from the standard terse output and
4341 for v3 and later appears on its own separate line at the end of each terse
4344 The \fBjson\fR output format is intended to be both human readable and convenient
4345 for automated parsing. For the most part its sections mirror those of the
4346 \fBnormal\fR output. The \fBruntime\fR value is reported in msec and the \fBbw\fR value is
4347 reported in 1024 bytes per second units.
4350 The \fBjson+\fR output format is identical to the \fBjson\fR output format except that it
4351 adds a full dump of the completion latency bins. Each \fBbins\fR object contains a
4352 set of (key, value) pairs where keys are latency durations and values count how
4353 many I/Os had completion latencies of the corresponding duration. For example,
4357 "bins" : { "87552" : 1, "89600" : 1, "94720" : 1, "96768" : 1, "97792" : 1, "99840" : 1, "100864" : 2, "103936" : 6, "104960" : 534, "105984" : 5995, "107008" : 7529, ... }
4360 This data indicates that one I/O required 87,552ns to complete, two I/Os required
4361 100,864ns to complete, and 7529 I/Os required 107,008ns to complete.
4363 Also included with fio is a Python script \fBfio_jsonplus_clat2csv\fR that takes
4364 json+ output and generates CSV-formatted latency data suitable for plotting.
4366 The latency durations actually represent the midpoints of latency intervals.
4367 For details refer to `stat.h' in the fio source.
4368 .SH TRACE FILE FORMAT
4369 There are two trace file format that you can encounter. The older (v1) format is
4370 unsupported since version 1.20\-rc3 (March 2008). It will still be described
4371 below in case that you get an old trace and want to understand it.
4373 In any case the trace is a simple text file with a single action per line.
4375 .B Trace file format v1
4376 Each line represents a single I/O action in the following format:
4383 where `rw=0/1' for read/write, and the `offset' and `length' entries being in bytes.
4385 This format is not supported in fio versions >= 1.20\-rc3.
4388 .B Trace file format v2
4389 The second version of the trace file format was added in fio version 1.17. It
4390 allows one to access more than one file per trace and has a bigger set of possible
4394 The first line of the trace file has to be:
4397 "fio version 2 iolog"
4400 Following this can be lines in two different formats, which are described below.
4403 The file management format:
4407 The `filename' is given as an absolute path. The `action' can be one of these:
4411 Add the given `filename' to the trace.
4414 Open the file with the given `filename'. The `filename' has to have
4415 been added with the \fBadd\fR action before.
4418 Close the file with the given `filename'. The file has to have been
4419 \fBopen\fRed before.
4424 The file I/O action format:
4426 filename action offset length
4428 The `filename' is given as an absolute path, and has to have been \fBadd\fRed and
4429 \fBopen\fRed before it can be used with this format. The `offset' and `length' are
4430 given in bytes. The `action' can be one of these:
4434 Wait for `offset' microseconds. Everything below 100 is discarded.
4435 The time is relative to the previous `wait' statement. Note that action `wait`
4436 is not allowed as of version 3, as the same behavior can be achieved using
4440 Read `length' bytes beginning from `offset'.
4443 Write `length' bytes beginning from `offset'.
4446 \fBfsync\fR\|(2) the file.
4449 \fBfdatasync\fR\|(2) the file.
4452 Trim the given file from the given `offset' for `length' bytes.
4457 .B Trace file format v3
4458 The third version of the trace file format was added in fio version 3.31. It
4459 forces each action to have a timestamp associated with it.
4462 The first line of the trace file has to be:
4465 "fio version 3 iolog"
4468 Following this can be lines in two different formats, which are described below.
4471 The file management format:
4473 timestamp filename action
4477 The file I/O action format:
4479 timestamp filename action offset length
4481 The `timestamp` is relative to the beginning of the run (ie starts at 0). The
4482 `filename`, `action`, `offset` and `length` are identical to version 2, except
4483 that version 3 does not allow the `wait` action.
4486 .SH I/O REPLAY \- MERGING TRACES
4487 Colocation is a common practice used to get the most out of a machine.
4488 Knowing which workloads play nicely with each other and which ones don't is
4489 a much harder task. While fio can replay workloads concurrently via multiple
4490 jobs, it leaves some variability up to the scheduler making results harder to
4491 reproduce. Merging is a way to make the order of events consistent.
4493 Merging is integrated into I/O replay and done when a \fBmerge_blktrace_file\fR
4494 is specified. The list of files passed to \fBread_iolog\fR go through the merge
4495 process and output a single file stored to the specified file. The output file is
4496 passed on as if it were the only file passed to \fBread_iolog\fR. An example would
4500 $ fio \-\-read_iolog="<file1>:<file2>" \-\-merge_blktrace_file="<output_file>"
4503 Creating only the merged file can be done by passing the command line argument
4504 \fBmerge-blktrace-only\fR.
4506 Scaling traces can be done to see the relative impact of any particular trace
4507 being slowed down or sped up. \fBmerge_blktrace_scalars\fR takes in a colon
4508 separated list of percentage scalars. It is index paired with the files passed
4509 to \fBread_iolog\fR.
4511 With scaling, it may be desirable to match the running time of all traces.
4512 This can be done with \fBmerge_blktrace_iters\fR. It is index paired with
4513 \fBread_iolog\fR just like \fBmerge_blktrace_scalars\fR.
4515 In an example, given two traces, A and B, each 60s long. If we want to see
4516 the impact of trace A issuing IOs twice as fast and repeat trace A over the
4517 runtime of trace B, the following can be done:
4520 $ fio \-\-read_iolog="<trace_a>:"<trace_b>" \-\-merge_blktrace_file"<output_file>" \-\-merge_blktrace_scalars="50:100" \-\-merge_blktrace_iters="2:1"
4523 This runs trace A at 2x the speed twice for approximately the same runtime as
4524 a single run of trace B.
4525 .SH CPU IDLENESS PROFILING
4526 In some cases, we want to understand CPU overhead in a test. For example, we
4527 test patches for the specific goodness of whether they reduce CPU usage.
4528 Fio implements a balloon approach to create a thread per CPU that runs at idle
4529 priority, meaning that it only runs when nobody else needs the cpu.
4530 By measuring the amount of work completed by the thread, idleness of each CPU
4531 can be derived accordingly.
4533 An unit work is defined as touching a full page of unsigned characters. Mean and
4534 standard deviation of time to complete an unit work is reported in "unit work"
4535 section. Options can be chosen to report detailed percpu idleness or overall
4536 system idleness by aggregating percpu stats.
4537 .SH VERIFICATION AND TRIGGERS
4538 Fio is usually run in one of two ways, when data verification is done. The first
4539 is a normal write job of some sort with verify enabled. When the write phase has
4540 completed, fio switches to reads and verifies everything it wrote. The second
4541 model is running just the write phase, and then later on running the same job
4542 (but with reads instead of writes) to repeat the same I/O patterns and verify
4543 the contents. Both of these methods depend on the write phase being completed,
4544 as fio otherwise has no idea how much data was written.
4546 With verification triggers, fio supports dumping the current write state to
4547 local files. Then a subsequent read verify workload can load this state and know
4548 exactly where to stop. This is useful for testing cases where power is cut to a
4549 server in a managed fashion, for instance.
4551 A verification trigger consists of two things:
4554 1) Storing the write state of each job.
4556 2) Executing a trigger command.
4559 The write state is relatively small, on the order of hundreds of bytes to single
4560 kilobytes. It contains information on the number of completions done, the last X
4563 A trigger is invoked either through creation ('touch') of a specified file in
4564 the system, or through a timeout setting. If fio is run with
4565 `\-\-trigger\-file=/tmp/trigger\-file', then it will continually
4566 check for the existence of `/tmp/trigger\-file'. When it sees this file, it
4567 will fire off the trigger (thus saving state, and executing the trigger
4570 For client/server runs, there's both a local and remote trigger. If fio is
4571 running as a server backend, it will send the job states back to the client for
4572 safe storage, then execute the remote trigger, if specified. If a local trigger
4573 is specified, the server will still send back the write state, but the client
4574 will then execute the trigger.
4577 .B Verification trigger example
4579 Let's say we want to run a powercut test on the remote Linux machine 'server'.
4580 Our write workload is in `write\-test.fio'. We want to cut power to 'server' at
4581 some point during the run, and we'll run this test from the safety or our local
4582 machine, 'localbox'. On the server, we'll start the fio backend normally:
4585 server# fio \-\-server
4588 and on the client, we'll fire off the workload:
4591 localbox$ fio \-\-client=server \-\-trigger\-file=/tmp/my\-trigger \-\-trigger\-remote="bash \-c "echo b > /proc/sysrq\-triger""
4594 We set `/tmp/my\-trigger' as the trigger file, and we tell fio to execute:
4597 echo b > /proc/sysrq\-trigger
4600 on the server once it has received the trigger and sent us the write state. This
4601 will work, but it's not really cutting power to the server, it's merely
4602 abruptly rebooting it. If we have a remote way of cutting power to the server
4603 through IPMI or similar, we could do that through a local trigger command
4604 instead. Let's assume we have a script that does IPMI reboot of a given hostname,
4605 ipmi\-reboot. On localbox, we could then have run fio with a local trigger
4609 localbox$ fio \-\-client=server \-\-trigger\-file=/tmp/my\-trigger \-\-trigger="ipmi\-reboot server"
4612 For this case, fio would wait for the server to send us the write state, then
4613 execute `ipmi\-reboot server' when that happened.
4616 .B Loading verify state
4618 To load stored write state, a read verification job file must contain the
4619 \fBverify_state_load\fR option. If that is set, fio will load the previously
4620 stored state. For a local fio run this is done by loading the files directly,
4621 and on a client/server run, the server backend will ask the client to send the
4622 files over and load them from there.
4624 .SH LOG FILE FORMATS
4625 Fio supports a variety of log file formats, for logging latencies, bandwidth,
4626 and IOPS. The logs share a common format, which looks like this:
4629 time (msec), value, data direction, block size (bytes), offset (bytes),
4633 `Time' for the log entry is always in milliseconds. The `value' logged depends
4634 on the type of log, it will be one of the following:
4638 Value is latency in nsecs
4647 `Data direction' is one of the following:
4660 The entry's `block size' is always in bytes. The `offset' is the position in bytes
4661 from the start of the file for that particular I/O. The logging of the offset can be
4662 toggled with \fBlog_offset\fR.
4664 If \fBlog_prio\fR is not set, the entry's `Command priority` is 1 for an IO executed
4665 with the highest RT priority class (\fBprioclass\fR=1 or \fBcmdprio_class\fR=1) and 0
4666 otherwise. This is controlled by the \fBprioclass\fR option and the ioengine specific
4667 \fBcmdprio_percentage\fR \fBcmdprio_class\fR options. If \fBlog_prio\fR is set, the
4668 entry's `Command priority` is the priority set for the IO, as a 16-bits hexadecimal
4669 number with the lowest 13 bits indicating the priority value (\fBprio\fR and
4670 \fBcmdprio\fR options) and the highest 3 bits indicating the IO priority class
4671 (\fBprioclass\fR and \fBcmdprio_class\fR options).
4673 Fio defaults to logging every individual I/O but when windowed logging is set
4674 through \fBlog_avg_msec\fR, either the average (by default) or the maximum
4675 (\fBlog_max_value\fR is set) `value' seen over the specified period of time
4676 is recorded. Each `data direction' seen within the window period will aggregate
4677 its values in a separate row. Further, when using windowed logging the `block
4678 size' and `offset' entries will always contain 0.
4680 Normally fio is invoked as a stand-alone application on the machine where the
4681 I/O workload should be generated. However, the backend and frontend of fio can
4682 be run separately i.e., the fio server can generate an I/O workload on the "Device
4683 Under Test" while being controlled by a client on another machine.
4685 Start the server on the machine which has access to the storage DUT:
4688 $ fio \-\-server=args
4691 where `args' defines what fio listens to. The arguments are of the form
4692 `type,hostname' or `IP,port'. `type' is either `ip' (or ip4) for TCP/IP
4693 v4, `ip6' for TCP/IP v6, or `sock' for a local unix domain socket.
4694 `hostname' is either a hostname or IP address, and `port' is the port to listen
4695 to (only valid for TCP/IP, not a local socket). Some examples:
4698 1) \fBfio \-\-server\fR
4699 Start a fio server, listening on all interfaces on the default port (8765).
4701 2) \fBfio \-\-server=ip:hostname,4444\fR
4702 Start a fio server, listening on IP belonging to hostname and on port 4444.
4704 3) \fBfio \-\-server=ip6:::1,4444\fR
4705 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
4707 4) \fBfio \-\-server=,4444\fR
4708 Start a fio server, listening on all interfaces on port 4444.
4710 5) \fBfio \-\-server=1.2.3.4\fR
4711 Start a fio server, listening on IP 1.2.3.4 on the default port.
4713 6) \fBfio \-\-server=sock:/tmp/fio.sock\fR
4714 Start a fio server, listening on the local socket `/tmp/fio.sock'.
4717 Once a server is running, a "client" can connect to the fio server with:
4720 $ fio <local\-args> \-\-client=<server> <remote\-args> <job file(s)>
4723 where `local\-args' are arguments for the client where it is running, `server'
4724 is the connect string, and `remote\-args' and `job file(s)' are sent to the
4725 server. The `server' string follows the same format as it does on the server
4726 side, to allow IP/hostname/socket and port strings.
4728 Fio can connect to multiple servers this way:
4731 $ fio \-\-client=<server1> <job file(s)> \-\-client=<server2> <job file(s)>
4734 If the job file is located on the fio server, then you can tell the server to
4735 load a local file as well. This is done by using \fB\-\-remote\-config\fR:
4738 $ fio \-\-client=server \-\-remote\-config /path/to/file.fio
4741 Then fio will open this local (to the server) job file instead of being passed
4742 one from the client.
4744 If you have many servers (example: 100 VMs/containers), you can input a pathname
4745 of a file containing host IPs/names as the parameter value for the
4746 \fB\-\-client\fR option. For example, here is an example `host.list'
4747 file containing 2 hostnames:
4751 host1.your.dns.domain
4753 host2.your.dns.domain
4757 The fio command would then be:
4760 $ fio \-\-client=host.list <job file(s)>
4763 In this mode, you cannot input server-specific parameters or job files \-\- all
4764 servers receive the same job file.
4766 In order to let `fio \-\-client' runs use a shared filesystem from multiple
4767 hosts, `fio \-\-client' now prepends the IP address of the server to the
4768 filename. For example, if fio is using the directory `/mnt/nfs/fio' and is
4769 writing filename `fileio.tmp', with a \fB\-\-client\fR `hostfile'
4770 containing two hostnames `h1' and `h2' with IP addresses 192.168.10.120 and
4771 192.168.10.121, then fio will create two files:
4775 /mnt/nfs/fio/192.168.10.120.fileio.tmp
4777 /mnt/nfs/fio/192.168.10.121.fileio.tmp
4781 Terse output in client/server mode will differ slightly from what is produced
4782 when fio is run in stand-alone mode. See the terse output section for details.
4785 was written by Jens Axboe <axboe@kernel.dk>.
4787 This man page was written by Aaron Carroll <aaronc@cse.unsw.edu.au> based
4788 on documentation by Jens Axboe.
4790 This man page was rewritten by Tomohiro Kusumi <tkusumi@tuxera.com> based
4791 on documentation by Jens Axboe.
4792 .SH "REPORTING BUGS"
4793 Report bugs to the \fBfio\fR mailing list <fio@vger.kernel.org>.
4795 See \fBREPORTING\-BUGS\fR.
4797 \fBREPORTING\-BUGS\fR: \fIhttp://git.kernel.dk/cgit/fio/plain/REPORTING\-BUGS\fR
4799 For further documentation see \fBHOWTO\fR and \fBREADME\fR.
4801 Sample jobfiles are available in the `examples/' directory.
4803 These are typically located under `/usr/share/doc/fio'.
4805 \fBHOWTO\fR: \fIhttp://git.kernel.dk/cgit/fio/plain/HOWTO\fR
4807 \fBREADME\fR: \fIhttp://git.kernel.dk/cgit/fio/plain/README\fR