blob: 138208f22e926b8582011f5cab58aa9208f43436 [file] [log] [blame]
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001.TH fio 1 "September 2007" "User Manual"
2.SH NAME
3fio \- flexible I/O tester
4.SH SYNOPSIS
5.B fio
6[\fIoptions\fR] [\fIjobfile\fR]...
7.SH DESCRIPTION
8.B fio
9is a tool that will spawn a number of threads or processes doing a
10particular type of I/O action as specified by the user.
11The typical use of fio is to write a job file matching the I/O load
12one wants to simulate.
13.SH OPTIONS
14.TP
Jens Axboe49da1242011-10-13 20:17:02 +020015.BI \-\-debug \fR=\fPtype
16Enable verbose tracing of various fio actions. May be `all' for all types
17or individual types separated by a comma (eg \-\-debug=io,file). `help' will
18list all available tracing options.
19.TP
Aaron Carrolld60e92d2007-09-17 10:32:59 +020020.BI \-\-output \fR=\fPfilename
21Write output to \fIfilename\fR.
22.TP
23.BI \-\-timeout \fR=\fPtimeout
24Limit run time to \fItimeout\fR seconds.
25.TP
26.B \-\-latency\-log
27Generate per-job latency logs.
28.TP
29.B \-\-bandwidth\-log
30Generate per-job bandwidth logs.
31.TP
32.B \-\-minimal
Aaron Carrolld1429b52007-09-18 08:10:57 +020033Print statistics in a terse, semicolon-delimited format.
Aaron Carrolld60e92d2007-09-17 10:32:59 +020034.TP
Jens Axboe49da1242011-10-13 20:17:02 +020035.B \-\-version
36Display version information and exit.
37.TP
Jens Axboe065248b2011-10-13 20:51:05 +020038.BI \-\-terse\-version \fR=\fPversion
Jens Axboe4d658652011-10-17 15:05:47 +020039Set terse version output format (Current version 3, or older version 2).
Jens Axboe49da1242011-10-13 20:17:02 +020040.TP
41.B \-\-help
42Display usage information and exit.
43.TP
44.BI \-\-cmdhelp \fR=\fPcommand
45Print help information for \fIcommand\fR. May be `all' for all commands.
46.TP
Steven Langde890a12011-11-09 14:03:34 +010047.BI \-\-enghelp \fR=\fPioengine[,command]
48List all commands defined by \fIioengine\fR, or print help for \fIcommand\fR defined by \fIioengine\fR.
49.TP
Aaron Carrolld60e92d2007-09-17 10:32:59 +020050.BI \-\-showcmd \fR=\fPjobfile
51Convert \fIjobfile\fR to a set of command-line options.
52.TP
53.B \-\-readonly
54Enable read-only safety checks.
55.TP
56.BI \-\-eta \fR=\fPwhen
57Specifies when real-time ETA estimate should be printed. \fIwhen\fR may
58be one of `always', `never' or `auto'.
59.TP
Jens Axboe49da1242011-10-13 20:17:02 +020060.BI \-\-readonly
61Turn on safety read-only checks, preventing any attempted write.
62.TP
Aaron Carrollc0a5d352008-02-26 23:10:39 +010063.BI \-\-section \fR=\fPsec
Jens Axboe49da1242011-10-13 20:17:02 +020064Only run section \fIsec\fR from job file. Multiple of these options can be given, adding more sections to run.
Aaron Carrollc0a5d352008-02-26 23:10:39 +010065.TP
Jens Axboe49da1242011-10-13 20:17:02 +020066.BI \-\-alloc\-size \fR=\fPkb
67Set the internal smalloc pool size to \fIkb\fP kilobytes.
Aaron Carrolld60e92d2007-09-17 10:32:59 +020068.TP
Jens Axboe49da1242011-10-13 20:17:02 +020069.BI \-\-warnings\-fatal
70All fio parser warnings are fatal, causing fio to exit with an error.
Jens Axboe91837882008-02-05 12:02:07 +010071.TP
Jens Axboe49da1242011-10-13 20:17:02 +020072.BI \-\-max\-jobs \fR=\fPnr
73Set the maximum allowed number of jobs (threads/processes) to suport.
Aaron Carrolld60e92d2007-09-17 10:32:59 +020074.TP
Jens Axboe49da1242011-10-13 20:17:02 +020075.BI \-\-server \fR=\fPargs
76Start a backend server, with \fIargs\fP specifying what to listen to. See client/server section.
Jens Axboef57a9c52011-09-09 21:01:37 +020077.TP
Jens Axboe49da1242011-10-13 20:17:02 +020078.BI \-\-daemonize \fR=\fPpidfile
79Background a fio server, writing the pid to the given pid file.
80.TP
81.BI \-\-client \fR=\fPhost
82Instead of running the jobs locally, send and run them on the given host.
Aaron Carrolld60e92d2007-09-17 10:32:59 +020083.SH "JOB FILE FORMAT"
84Job files are in `ini' format. They consist of one or more
85job definitions, which begin with a job name in square brackets and
86extend to the next job name. The job name can be any ASCII string
87except `global', which has a special meaning. Following the job name is
88a sequence of zero or more parameters, one per line, that define the
89behavior of the job. Any line starting with a `;' or `#' character is
Aaron Carrolld1429b52007-09-18 08:10:57 +020090considered a comment and ignored.
Aaron Carrolld9956b62007-11-16 12:12:45 +010091.P
92If \fIjobfile\fR is specified as `-', the job file will be read from
93standard input.
Aaron Carrolld60e92d2007-09-17 10:32:59 +020094.SS "Global Section"
95The global section contains default parameters for jobs specified in the
96job file. A job is only affected by global sections residing above it,
97and there may be any number of global sections. Specific job definitions
98may override any parameter set in global sections.
99.SH "JOB PARAMETERS"
100.SS Types
101Some parameters may take arguments of a specific type. The types used are:
102.TP
103.I str
104String: a sequence of alphanumeric characters.
105.TP
106.I int
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200107SI integer: a whole number, possibly containing a suffix denoting the base unit
Jens Axboeb09da8f2009-07-17 23:16:17 +0200108of the value. Accepted suffixes are `k', 'M', 'G', 'T', and 'P', denoting
109kilo (1024), mega (1024^2), giga (1024^3), tera (1024^4), and peta (1024^5)
110respectively. The suffix is not case sensitive. If prefixed with '0x', the
Martin Steigerwald5982a922011-06-27 16:07:24 +0200111value is assumed to be base 16 (hexadecimal). A suffix may include a trailing 'b',
112for instance 'kb' is identical to 'k'. You can specify a base 10 value
Jens Axboe57fc29f2010-06-23 22:24:07 +0200113by using 'KiB', 'MiB', 'GiB', etc. This is useful for disk drives where
114values are often given in base 10 values. Specifying '30GiB' will get you
11530*1000^3 bytes.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200116.TP
117.I bool
118Boolean: a true or false value. `0' denotes false, `1' denotes true.
119.TP
120.I irange
121Integer range: a range of integers specified in the format
Aaron Carrolld1429b52007-09-18 08:10:57 +0200122\fIlower\fR:\fIupper\fR or \fIlower\fR\-\fIupper\fR. \fIlower\fR and
123\fIupper\fR may contain a suffix as described above. If an option allows two
124sets of ranges, they are separated with a `,' or `/' character. For example:
125`8\-8k/8M\-4G'.
Yu-ju Hong83349192011-08-13 00:53:44 +0200126.TP
127.I float_list
128List of floating numbers: A list of floating numbers, separated by
129a ':' charcater.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200130.SS "Parameter List"
131.TP
132.BI name \fR=\fPstr
Aaron Carrolld9956b62007-11-16 12:12:45 +0100133May be used to override the job name. On the command line, this parameter
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200134has the special purpose of signalling the start of a new job.
135.TP
136.BI description \fR=\fPstr
137Human-readable description of the job. It is printed when the job is run, but
138otherwise has no special purpose.
139.TP
140.BI directory \fR=\fPstr
141Prefix filenames with this directory. Used to place files in a location other
142than `./'.
143.TP
144.BI filename \fR=\fPstr
145.B fio
146normally makes up a file name based on the job name, thread number, and file
Aaron Carrolld1429b52007-09-18 08:10:57 +0200147number. If you want to share files between threads in a job or several jobs,
Steven Langde890a12011-11-09 14:03:34 +0100148specify a \fIfilename\fR for each of them to override the default.
149If the I/O engine is file-based, you can specify
Aaron Carrolld1429b52007-09-18 08:10:57 +0200150a number of files by separating the names with a `:' character. `\-' is a
151reserved name, meaning stdin or stdout, depending on the read/write direction
152set.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200153.TP
Jens Axboe3ce9dca2009-06-10 08:55:21 +0200154.BI lockfile \fR=\fPstr
155Fio defaults to not locking any files before it does IO to them. If a file or
156file descriptor is shared, fio can serialize IO to that file to make the end
157result consistent. This is usual for emulating real workloads that share files.
158The lock modes are:
159.RS
160.RS
161.TP
162.B none
163No locking. This is the default.
164.TP
165.B exclusive
166Only one thread or process may do IO at the time, excluding all others.
167.TP
168.B readwrite
169Read-write locking on the file. Many readers may access the file at the same
170time, but writes get exclusive access.
171.RE
172.P
173The option may be post-fixed with a lock batch number. If set, then each
174thread/process may do that amount of IOs to the file before giving up the lock.
175Since lock acquisition is expensive, batching the lock/unlocks will speed up IO.
176.RE
177.P
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200178.BI opendir \fR=\fPstr
179Recursively open any files below directory \fIstr\fR.
180.TP
181.BI readwrite \fR=\fPstr "\fR,\fP rw" \fR=\fPstr
182Type of I/O pattern. Accepted values are:
183.RS
184.RS
185.TP
186.B read
Aaron Carrolld1429b52007-09-18 08:10:57 +0200187Sequential reads.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200188.TP
189.B write
Aaron Carrolld1429b52007-09-18 08:10:57 +0200190Sequential writes.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200191.TP
192.B randread
Aaron Carrolld1429b52007-09-18 08:10:57 +0200193Random reads.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200194.TP
195.B randwrite
Aaron Carrolld1429b52007-09-18 08:10:57 +0200196Random writes.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200197.TP
198.B rw
Aaron Carrolld1429b52007-09-18 08:10:57 +0200199Mixed sequential reads and writes.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200200.TP
201.B randrw
Aaron Carrolld1429b52007-09-18 08:10:57 +0200202Mixed random reads and writes.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200203.RE
204.P
Jens Axboe38dad622010-07-20 14:46:00 -0600205For mixed I/O, the default split is 50/50. For certain types of io the result
206may still be skewed a bit, since the speed may be different. It is possible to
207specify a number of IO's to do before getting a new offset, this is one by
208appending a `:\fI<nr>\fR to the end of the string given. For a random read, it
209would look like \fBrw=randread:8\fR for passing in an offset modifier with a
Jens Axboe059b0802011-08-25 09:09:37 +0200210value of 8. If the postfix is used with a sequential IO pattern, then the value
211specified will be added to the generated offset for each IO. For instance,
212using \fBrw=write:4k\fR will skip 4k for every write. It turns sequential IO
213into sequential IO with holes. See the \fBrw_sequencer\fR option.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200214.RE
215.TP
Jens Axboe38dad622010-07-20 14:46:00 -0600216.BI rw_sequencer \fR=\fPstr
217If an offset modifier is given by appending a number to the \fBrw=<str>\fR line,
218then this option controls how that number modifies the IO offset being
219generated. Accepted values are:
220.RS
221.RS
222.TP
223.B sequential
224Generate sequential offset
225.TP
226.B identical
227Generate the same offset
228.RE
229.P
230\fBsequential\fR is only useful for random IO, where fio would normally
231generate a new random offset for every IO. If you append eg 8 to randread, you
232would get a new random offset for every 8 IO's. The result would be a seek for
233only every 8 IO's, instead of for every IO. Use \fBrw=randread:8\fR to specify
234that. As sequential IO is already sequential, setting \fBsequential\fR for that
235would not result in any differences. \fBidentical\fR behaves in a similar
236fashion, except it sends the same offset 8 number of times before generating a
237new offset.
238.RE
239.P
240.TP
Jens Axboe90fef2d2009-07-17 22:33:32 +0200241.BI kb_base \fR=\fPint
242The base unit for a kilobyte. The defacto base is 2^10, 1024. Storage
243manufacturers like to use 10^3 or 1000 as a base ten unit instead, for obvious
244reasons. Allow values are 1024 or 1000, with 1024 being the default.
245.TP
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200246.BI randrepeat \fR=\fPbool
247Seed the random number generator in a predictable way so results are repeatable
Aaron Carrolld1429b52007-09-18 08:10:57 +0200248across runs. Default: true.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200249.TP
Jens Axboe2615cc42011-03-28 09:35:09 +0200250.BI use_os_rand \fR=\fPbool
251Fio can either use the random generator supplied by the OS to generator random
252offsets, or it can use it's own internal generator (based on Tausworthe).
253Default is to use the internal generator, which is often of better quality and
254faster. Default: false.
255.TP
Eric Gourioua596f042011-06-17 09:11:45 +0200256.BI fallocate \fR=\fPstr
257Whether pre-allocation is performed when laying down files. Accepted values
258are:
259.RS
260.RS
261.TP
262.B none
263Do not pre-allocate space.
264.TP
265.B posix
266Pre-allocate via posix_fallocate().
267.TP
268.B keep
269Pre-allocate via fallocate() with FALLOC_FL_KEEP_SIZE set.
270.TP
271.B 0
272Backward-compatible alias for 'none'.
273.TP
274.B 1
275Backward-compatible alias for 'posix'.
276.RE
277.P
278May not be available on all supported platforms. 'keep' is only
279available on Linux. If using ZFS on Solaris this must be set to 'none'
280because ZFS doesn't support it. Default: 'posix'.
281.RE
Jens Axboe7bc8c2c2010-01-28 11:31:31 +0100282.TP
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200283.BI fadvise_hint \fR=\fPbool
Aaron Carrolld1429b52007-09-18 08:10:57 +0200284Disable use of \fIposix_fadvise\fR\|(2) to advise the kernel what I/O patterns
285are likely to be issued. Default: true.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200286.TP
Jens Axboef7fa2652009-03-09 14:20:20 +0100287.BI size \fR=\fPint
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200288Total size of I/O for this job. \fBfio\fR will run until this many bytes have
289been transfered, unless limited by other options (\fBruntime\fR, for instance).
Jens Axboed7c8be02010-11-25 08:21:39 +0100290Unless \fBnrfiles\fR and \fBfilesize\fR options are given, this amount will be
Jens Axboed6667262010-06-25 11:32:48 +0200291divided between the available files for the job. If not set, fio will use the
292full size of the given files or devices. If the the files do not exist, size
Jens Axboe7bb59102011-07-12 19:47:03 +0200293must be given. It is also possible to give size as a percentage between 1 and
294100. If size=20% is given, fio will use 20% of the full size of the given files
295or devices.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200296.TP
Jens Axboe74586c12011-01-20 10:16:03 -0700297.BI fill_device \fR=\fPbool "\fR,\fB fill_fs" \fR=\fPbool
Jens Axboe3ce9dca2009-06-10 08:55:21 +0200298Sets size to something really large and waits for ENOSPC (no space left on
299device) as the terminating condition. Only makes sense with sequential write.
300For a read workload, the mount point will be filled first then IO started on
Jens Axboe4f124322011-01-19 15:35:26 -0700301the result. This option doesn't make sense if operating on a raw device node,
302since the size of that is already known by the file system. Additionally,
303writing beyond end-of-device will not return ENOSPC there.
Jens Axboe3ce9dca2009-06-10 08:55:21 +0200304.TP
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200305.BI filesize \fR=\fPirange
306Individual file sizes. May be a range, in which case \fBfio\fR will select sizes
Aaron Carrolld1429b52007-09-18 08:10:57 +0200307for files at random within the given range, limited to \fBsize\fR in total (if
308that is given). If \fBfilesize\fR is not specified, each created file is the
309same size.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200310.TP
Jens Axboef7fa2652009-03-09 14:20:20 +0100311.BI blocksize \fR=\fPint[,int] "\fR,\fB bs" \fR=\fPint[,int]
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200312Block size for I/O units. Default: 4k. Values for reads and writes can be
Jens Axboe656ebab2010-04-13 10:39:14 +0200313specified separately in the format \fIread\fR,\fIwrite\fR, either of
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200314which may be empty to leave that value at its default.
315.TP
Jens Axboe91837882008-02-05 12:02:07 +0100316.BI blocksize_range \fR=\fPirange[,irange] "\fR,\fB bsrange" \fR=\fPirange[,irange]
Aaron Carrolld1429b52007-09-18 08:10:57 +0200317Specify a range of I/O block sizes. The issued I/O unit will always be a
318multiple of the minimum size, unless \fBblocksize_unaligned\fR is set. Applies
Jens Axboe91837882008-02-05 12:02:07 +0100319to both reads and writes if only one range is given, but can be specified
Jens Axboe656ebab2010-04-13 10:39:14 +0200320separately with a comma seperating the values. Example: bsrange=1k-4k,2k-8k.
Jens Axboe91837882008-02-05 12:02:07 +0100321Also (see \fBblocksize\fR).
322.TP
323.BI bssplit \fR=\fPstr
324This option allows even finer grained control of the block sizes issued,
325not just even splits between them. With this option, you can weight various
326block sizes for exact control of the issued IO for a job that has mixed
327block sizes. The format of the option is bssplit=blocksize/percentage,
Martin Steigerwald5982a922011-06-27 16:07:24 +0200328optionally adding as many definitions as needed separated by a colon.
Jens Axboe91837882008-02-05 12:02:07 +0100329Example: bssplit=4k/10:64k/50:32k/40 would issue 50% 64k blocks, 10% 4k
Jens Axboec83cdd32009-04-24 14:23:59 +0200330blocks and 40% 32k blocks. \fBbssplit\fR also supports giving separate
331splits to reads and writes. The format is identical to what the
332\fBbs\fR option accepts, the read and write parts are separated with a
333comma.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200334.TP
335.B blocksize_unaligned\fR,\fP bs_unaligned
Aaron Carrolld1429b52007-09-18 08:10:57 +0200336If set, any size in \fBblocksize_range\fR may be used. This typically won't
337work with direct I/O, as that normally requires sector alignment.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200338.TP
Jens Axboe2b7a01d2009-03-11 11:00:13 +0100339.BI blockalign \fR=\fPint[,int] "\fR,\fB ba" \fR=\fPint[,int]
Martin Steigerwald639ce0f2009-05-20 11:33:49 +0200340At what boundary to align random IO offsets. Defaults to the same as 'blocksize'
341the minimum blocksize given. Minimum alignment is typically 512b
Jens Axboe2b7a01d2009-03-11 11:00:13 +0100342for using direct IO, though it usually depends on the hardware block size.
343This option is mutually exclusive with using a random map for files, so it
344will turn off that option.
Jens Axboe43602662009-03-14 20:08:47 +0100345.TP
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200346.B zero_buffers
347Initialise buffers with all zeros. Default: fill buffers with random data.
348.TP
Jens Axboe901bb992009-03-14 20:17:36 +0100349.B refill_buffers
350If this option is given, fio will refill the IO buffers on every submit. The
351default is to only fill it at init time and reuse that data. Only makes sense
352if zero_buffers isn't specified, naturally. If data verification is enabled,
353refill_buffers is also automatically enabled.
354.TP
Jens Axboefd684182011-09-19 09:24:44 +0200355.BI scramble_buffers \fR=\fPbool
356If \fBrefill_buffers\fR is too costly and the target is using data
357deduplication, then setting this option will slightly modify the IO buffer
358contents to defeat normal de-dupe attempts. This is not enough to defeat
359more clever block compression attempts, but it will stop naive dedupe
360of blocks. Default: true.
361.TP
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200362.BI nrfiles \fR=\fPint
363Number of files to use for this job. Default: 1.
364.TP
365.BI openfiles \fR=\fPint
366Number of files to keep open at the same time. Default: \fBnrfiles\fR.
367.TP
368.BI file_service_type \fR=\fPstr
369Defines how files to service are selected. The following types are defined:
370.RS
371.RS
372.TP
373.B random
374Choose a file at random
375.TP
376.B roundrobin
377Round robin over open files (default).
Jens Axboe6b7f6852009-03-09 14:22:56 +0100378.B sequential
379Do each file in the set sequentially.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200380.RE
381.P
382The number of I/Os to issue before switching a new file can be specified by
383appending `:\fIint\fR' to the service type.
384.RE
385.TP
386.BI ioengine \fR=\fPstr
387Defines how the job issues I/O. The following types are defined:
388.RS
389.RS
390.TP
391.B sync
392Basic \fIread\fR\|(2) or \fIwrite\fR\|(2) I/O. \fIfseek\fR\|(2) is used to
393position the I/O location.
394.TP
gurudas paia31041e2007-10-23 15:12:30 +0200395.B psync
396Basic \fIpread\fR\|(2) or \fIpwrite\fR\|(2) I/O.
397.TP
Jens Axboe91837882008-02-05 12:02:07 +0100398.B vsync
399Basic \fIreadv\fR\|(2) or \fIwritev\fR\|(2) I/O. Will emulate queuing by
400coalescing adjacents IOs into a single submission.
401.TP
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200402.B libaio
Steven Langde890a12011-11-09 14:03:34 +0100403Linux native asynchronous I/O. This ioengine defines engine specific options.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200404.TP
405.B posixaio
Bruce Cran03e20d62011-01-02 20:14:54 +0100406POSIX asynchronous I/O using \fIaio_read\fR\|(3) and \fIaio_write\fR\|(3).
407.TP
408.B solarisaio
409Solaris native asynchronous I/O.
410.TP
411.B windowsaio
412Windows native asynchronous I/O.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200413.TP
414.B mmap
Aaron Carrolld1429b52007-09-18 08:10:57 +0200415File is memory mapped with \fImmap\fR\|(2) and data copied using
416\fImemcpy\fR\|(3).
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200417.TP
418.B splice
Aaron Carrolld1429b52007-09-18 08:10:57 +0200419\fIsplice\fR\|(2) is used to transfer the data and \fIvmsplice\fR\|(2) to
420transfer data from user-space to the kernel.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200421.TP
422.B syslet-rw
423Use the syslet system calls to make regular read/write asynchronous.
424.TP
425.B sg
426SCSI generic sg v3 I/O. May be either synchronous using the SG_IO ioctl, or if
Aaron Carrolld1429b52007-09-18 08:10:57 +0200427the target is an sg character device, we use \fIread\fR\|(2) and
428\fIwrite\fR\|(2) for asynchronous I/O.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200429.TP
430.B null
431Doesn't transfer any data, just pretends to. Mainly used to exercise \fBfio\fR
432itself and for debugging and testing purposes.
433.TP
434.B net
Steven Langde890a12011-11-09 14:03:34 +0100435Transfer over the network. The protocol to be used can be defined with the
436\fBprotocol\fR parameter. Depending on the protocol, \fBfilename\fR,
437\fBhostname\fR, \fBport\fR, or \fBlisten\fR must be specified.
438This ioengine defines engine specific options.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200439.TP
440.B netsplice
441Like \fBnet\fR, but uses \fIsplice\fR\|(2) and \fIvmsplice\fR\|(2) to map data
Steven Langde890a12011-11-09 14:03:34 +0100442and send/receive. This ioengine defines engine specific options.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200443.TP
gurudas pai53aec0a2007-10-05 13:20:18 +0200444.B cpuio
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200445Doesn't transfer any data, but burns CPU cycles according to \fBcpuload\fR and
446\fBcpucycles\fR parameters.
447.TP
448.B guasi
449The GUASI I/O engine is the Generic Userspace Asynchronous Syscall Interface
450approach to asycnronous I/O.
Aaron Carrolld1429b52007-09-18 08:10:57 +0200451.br
452See <http://www.xmailserver.org/guasi\-lib.html>.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200453.TP
ren yufei21b8aee2011-08-01 10:01:57 +0200454.B rdma
Bart Van Assche85286c52011-08-07 21:50:51 +0200455The RDMA I/O engine supports both RDMA memory semantics (RDMA_WRITE/RDMA_READ)
456and channel semantics (Send/Recv) for the InfiniBand, RoCE and iWARP protocols.
ren yufei21b8aee2011-08-01 10:01:57 +0200457.TP
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200458.B external
459Loads an external I/O engine object file. Append the engine filename as
460`:\fIenginepath\fR'.
461.RE
462.RE
463.TP
464.BI iodepth \fR=\fPint
Sebastian Kayser8489dae2010-12-01 22:28:47 +0100465Number of I/O units to keep in flight against the file. Note that increasing
466iodepth beyond 1 will not affect synchronous ioengines (except for small
Jens Axboeee72ca02010-12-02 20:05:37 +0100467degress when verify_async is in use). Even async engines my impose OS
468restrictions causing the desired depth not to be achieved. This may happen on
469Linux when using libaio and not setting \fBdirect\fR=1, since buffered IO is
470not async on that OS. Keep an eye on the IO depth distribution in the
471fio output to verify that the achieved depth is as expected. Default: 1.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200472.TP
473.BI iodepth_batch \fR=\fPint
474Number of I/Os to submit at once. Default: \fBiodepth\fR.
475.TP
Jens Axboe3ce9dca2009-06-10 08:55:21 +0200476.BI iodepth_batch_complete \fR=\fPint
477This defines how many pieces of IO to retrieve at once. It defaults to 1 which
478 means that we'll ask for a minimum of 1 IO in the retrieval process from the
479kernel. The IO retrieval will go on until we hit the limit set by
480\fBiodepth_low\fR. If this variable is set to 0, then fio will always check for
481completed events before queuing more IO. This helps reduce IO latency, at the
482cost of more retrieval system calls.
483.TP
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200484.BI iodepth_low \fR=\fPint
485Low watermark indicating when to start filling the queue again. Default:
486\fBiodepth\fR.
487.TP
488.BI direct \fR=\fPbool
489If true, use non-buffered I/O (usually O_DIRECT). Default: false.
490.TP
491.BI buffered \fR=\fPbool
492If true, use buffered I/O. This is the opposite of the \fBdirect\fR parameter.
493Default: true.
494.TP
Jens Axboef7fa2652009-03-09 14:20:20 +0100495.BI offset \fR=\fPint
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200496Offset in the file to start I/O. Data before the offset will not be touched.
497.TP
498.BI fsync \fR=\fPint
Aaron Carrolld1429b52007-09-18 08:10:57 +0200499How many I/Os to perform before issuing an \fBfsync\fR\|(2) of dirty data. If
5000, don't sync. Default: 0.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200501.TP
Jens Axboe5f9099e2009-06-16 22:40:26 +0200502.BI fdatasync \fR=\fPint
503Like \fBfsync\fR, but uses \fBfdatasync\fR\|(2) instead to only sync the
504data parts of the file. Default: 0.
505.TP
Jens Axboee76b1da2010-03-09 20:49:54 +0100506.BI sync_file_range \fR=\fPstr:int
507Use sync_file_range() for every \fRval\fP number of write operations. Fio will
508track range of writes that have happened since the last sync_file_range() call.
509\fRstr\fP can currently be one or more of:
510.RS
511.TP
512.B wait_before
513SYNC_FILE_RANGE_WAIT_BEFORE
514.TP
515.B write
516SYNC_FILE_RANGE_WRITE
517.TP
518.B wait_after
519SYNC_FILE_RANGE_WRITE
520.TP
521.RE
522.P
523So if you do sync_file_range=wait_before,write:8, fio would use
524\fBSYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE\fP for every 8 writes.
525Also see the sync_file_range(2) man page. This option is Linux specific.
526.TP
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200527.BI overwrite \fR=\fPbool
Aaron Carrolld1429b52007-09-18 08:10:57 +0200528If writing, setup the file first and do overwrites. Default: false.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200529.TP
530.BI end_fsync \fR=\fPbool
Aaron Carrolld1429b52007-09-18 08:10:57 +0200531Sync file contents when job exits. Default: false.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200532.TP
533.BI fsync_on_close \fR=\fPbool
534If true, sync file contents on close. This differs from \fBend_fsync\fR in that
Aaron Carrolld1429b52007-09-18 08:10:57 +0200535it will happen on every close, not just at the end of the job. Default: false.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200536.TP
537.BI rwmixcycle \fR=\fPint
538How many milliseconds before switching between reads and writes for a mixed
539workload. Default: 500ms.
540.TP
541.BI rwmixread \fR=\fPint
542Percentage of a mixed workload that should be reads. Default: 50.
543.TP
544.BI rwmixwrite \fR=\fPint
Aaron Carrolld1429b52007-09-18 08:10:57 +0200545Percentage of a mixed workload that should be writes. If \fBrwmixread\fR and
Jens Axboec35dd7a2009-06-10 08:39:16 +0200546\fBrwmixwrite\fR are given and do not sum to 100%, the latter of the two
547overrides the first. This may interfere with a given rate setting, if fio is
548asked to limit reads or writes to a certain rate. If that is the case, then
549the distribution may be skewed. Default: 50.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200550.TP
551.B norandommap
552Normally \fBfio\fR will cover every block of the file when doing random I/O. If
553this parameter is given, a new offset will be chosen without looking at past
554I/O history. This parameter is mutually exclusive with \fBverify\fR.
555.TP
Jens Axboe744492c2011-08-08 09:47:13 +0200556.BI softrandommap \fR=\fPbool
Jens Axboe3ce9dca2009-06-10 08:55:21 +0200557See \fBnorandommap\fR. If fio runs with the random block map enabled and it
558fails to allocate the map, if this option is set it will continue without a
559random block map. As coverage will not be as complete as with random maps, this
560option is disabled by default.
561.TP
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200562.BI nice \fR=\fPint
563Run job with given nice value. See \fInice\fR\|(2).
564.TP
565.BI prio \fR=\fPint
566Set I/O priority value of this job between 0 (highest) and 7 (lowest). See
567\fIionice\fR\|(1).
568.TP
569.BI prioclass \fR=\fPint
570Set I/O priority class. See \fIionice\fR\|(1).
571.TP
572.BI thinktime \fR=\fPint
573Stall job for given number of microseconds between issuing I/Os.
574.TP
575.BI thinktime_spin \fR=\fPint
576Pretend to spend CPU time for given number of microseconds, sleeping the rest
577of the time specified by \fBthinktime\fR. Only valid if \fBthinktime\fR is set.
578.TP
579.BI thinktime_blocks \fR=\fPint
580Number of blocks to issue before waiting \fBthinktime\fR microseconds.
581Default: 1.
582.TP
583.BI rate \fR=\fPint
Jens Axboec35dd7a2009-06-10 08:39:16 +0200584Cap bandwidth used by this job. The number is in bytes/sec, the normal postfix
585rules apply. You can use \fBrate\fR=500k to limit reads and writes to 500k each,
586or you can specify read and writes separately. Using \fBrate\fR=1m,500k would
587limit reads to 1MB/sec and writes to 500KB/sec. Capping only reads or writes
588can be done with \fBrate\fR=,500k or \fBrate\fR=500k,. The former will only
589limit writes (to 500KB/sec), the latter will only limit reads.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200590.TP
591.BI ratemin \fR=\fPint
592Tell \fBfio\fR to do whatever it can to maintain at least the given bandwidth.
Jens Axboec35dd7a2009-06-10 08:39:16 +0200593Failing to meet this requirement will cause the job to exit. The same format
594as \fBrate\fR is used for read vs write separation.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200595.TP
596.BI rate_iops \fR=\fPint
Jens Axboec35dd7a2009-06-10 08:39:16 +0200597Cap the bandwidth to this number of IOPS. Basically the same as rate, just
598specified independently of bandwidth. The same format as \fBrate\fR is used for
599read vs write seperation. If \fBblocksize\fR is a range, the smallest block
600size is used as the metric.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200601.TP
602.BI rate_iops_min \fR=\fPint
Jens Axboec35dd7a2009-06-10 08:39:16 +0200603If this rate of I/O is not met, the job will exit. The same format as \fBrate\fR
604is used for read vs write seperation.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200605.TP
606.BI ratecycle \fR=\fPint
607Average bandwidth for \fBrate\fR and \fBratemin\fR over this number of
608milliseconds. Default: 1000ms.
609.TP
610.BI cpumask \fR=\fPint
611Set CPU affinity for this job. \fIint\fR is a bitmask of allowed CPUs the job
612may run on. See \fBsched_setaffinity\fR\|(2).
613.TP
614.BI cpus_allowed \fR=\fPstr
615Same as \fBcpumask\fR, but allows a comma-delimited list of CPU numbers.
616.TP
617.BI startdelay \fR=\fPint
618Delay start of job for the specified number of seconds.
619.TP
620.BI runtime \fR=\fPint
621Terminate processing after the specified number of seconds.
622.TP
623.B time_based
624If given, run for the specified \fBruntime\fR duration even if the files are
625completely read or written. The same workload will be repeated as many times
626as \fBruntime\fR allows.
627.TP
Jens Axboe901bb992009-03-14 20:17:36 +0100628.BI ramp_time \fR=\fPint
629If set, fio will run the specified workload for this amount of time before
630logging any performance numbers. Useful for letting performance settle before
631logging results, thus minimizing the runtime required for stable results. Note
Jens Axboec35dd7a2009-06-10 08:39:16 +0200632that the \fBramp_time\fR is considered lead in time for a job, thus it will
633increase the total runtime if a special timeout or runtime is specified.
Jens Axboe901bb992009-03-14 20:17:36 +0100634.TP
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200635.BI invalidate \fR=\fPbool
636Invalidate buffer-cache for the file prior to starting I/O. Default: true.
637.TP
638.BI sync \fR=\fPbool
639Use synchronous I/O for buffered writes. For the majority of I/O engines,
Aaron Carrolld1429b52007-09-18 08:10:57 +0200640this means using O_SYNC. Default: false.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200641.TP
642.BI iomem \fR=\fPstr "\fR,\fP mem" \fR=\fPstr
643Allocation method for I/O unit buffer. Allowed values are:
644.RS
645.RS
646.TP
647.B malloc
648Allocate memory with \fImalloc\fR\|(3).
649.TP
650.B shm
651Use shared memory buffers allocated through \fIshmget\fR\|(2).
652.TP
653.B shmhuge
654Same as \fBshm\fR, but use huge pages as backing.
655.TP
656.B mmap
657Use \fImmap\fR\|(2) for allocation. Uses anonymous memory unless a filename
658is given after the option in the format `:\fIfile\fR'.
659.TP
660.B mmaphuge
661Same as \fBmmap\fR, but use huge files as backing.
662.RE
663.P
664The amount of memory allocated is the maximum allowed \fBblocksize\fR for the
665job multiplied by \fBiodepth\fR. For \fBshmhuge\fR or \fBmmaphuge\fR to work,
666the system must have free huge pages allocated. \fBmmaphuge\fR also needs to
Jens Axboe2e266ba2009-09-14 08:56:53 +0200667have hugetlbfs mounted, and \fIfile\fR must point there. At least on Linux,
668huge pages must be manually allocated. See \fB/proc/sys/vm/nr_hugehages\fR
669and the documentation for that. Normally you just need to echo an appropriate
670number, eg echoing 8 will ensure that the OS has 8 huge pages ready for
671use.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200672.RE
673.TP
Jens Axboed3923652011-08-03 12:38:39 +0200674.BI iomem_align \fR=\fPint "\fR,\fP mem_align" \fR=\fPint
Jens Axboed529ee12009-07-01 10:33:03 +0200675This indiciates the memory alignment of the IO memory buffers. Note that the
676given alignment is applied to the first IO unit buffer, if using \fBiodepth\fR
677the alignment of the following buffers are given by the \fBbs\fR used. In
678other words, if using a \fBbs\fR that is a multiple of the page sized in the
679system, all buffers will be aligned to this value. If using a \fBbs\fR that
680is not page aligned, the alignment of subsequent IO memory buffers is the
681sum of the \fBiomem_align\fR and \fBbs\fR used.
682.TP
Jens Axboef7fa2652009-03-09 14:20:20 +0100683.BI hugepage\-size \fR=\fPint
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200684Defines the size of a huge page. Must be at least equal to the system setting.
Jens Axboeb22989b2009-07-17 22:29:23 +0200685Should be a multiple of 1MB. Default: 4MB.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200686.TP
687.B exitall
688Terminate all jobs when one finishes. Default: wait for each job to finish.
689.TP
690.BI bwavgtime \fR=\fPint
691Average bandwidth calculations over the given time in milliseconds. Default:
692500ms.
693.TP
Jens Axboec8eeb9d2011-10-05 14:02:22 +0200694.BI iopsavgtime \fR=\fPint
695Average IOPS calculations over the given time in milliseconds. Default:
696500ms.
697.TP
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200698.BI create_serialize \fR=\fPbool
Aaron Carrolld1429b52007-09-18 08:10:57 +0200699If true, serialize file creation for the jobs. Default: true.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200700.TP
701.BI create_fsync \fR=\fPbool
702\fIfsync\fR\|(2) data file after creation. Default: true.
703.TP
Jens Axboe6b7f6852009-03-09 14:22:56 +0100704.BI create_on_open \fR=\fPbool
705If true, the files are not created until they are opened for IO by the job.
706.TP
Jens Axboee9f48472009-06-03 12:14:08 +0200707.BI pre_read \fR=\fPbool
708If this is given, files will be pre-read into memory before starting the given
709IO operation. This will also clear the \fR \fBinvalidate\fR flag, since it is
Jens Axboe9c0d2242009-07-01 12:26:28 +0200710pointless to pre-read and then drop the cache. This will only work for IO
711engines that are seekable, since they allow you to read the same data
712multiple times. Thus it will not work on eg network or splice IO.
Jens Axboee9f48472009-06-03 12:14:08 +0200713.TP
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200714.BI unlink \fR=\fPbool
715Unlink job files when done. Default: false.
716.TP
717.BI loops \fR=\fPint
718Specifies the number of iterations (runs of the same workload) of this job.
719Default: 1.
720.TP
721.BI do_verify \fR=\fPbool
722Run the verify phase after a write phase. Only valid if \fBverify\fR is set.
723Default: true.
724.TP
725.BI verify \fR=\fPstr
726Method of verifying file contents after each iteration of the job. Allowed
727values are:
728.RS
729.RS
730.TP
Jens Axboeb892dc02009-09-05 20:37:35 +0200731.B md5 crc16 crc32 crc32c crc32c-intel crc64 crc7 sha256 sha512 sha1
Jens Axboe0539d752010-06-21 15:22:56 +0200732Store appropriate checksum in the header of each block. crc32c-intel is
733hardware accelerated SSE4.2 driven, falls back to regular crc32c if
734not supported by the system.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200735.TP
736.B meta
737Write extra information about each I/O (timestamp, block number, etc.). The
Jens Axboe996093b2010-06-24 08:37:13 +0200738block number is verified. See \fBverify_pattern\fR as well.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200739.TP
740.B null
741Pretend to verify. Used for testing internals.
742.RE
Jens Axboeb892dc02009-09-05 20:37:35 +0200743
744This option can be used for repeated burn-in tests of a system to make sure
745that the written data is also correctly read back. If the data direction given
746is a read or random read, fio will assume that it should verify a previously
747written file. If the data direction includes any form of write, the verify will
748be of the newly written data.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200749.RE
750.TP
751.BI verify_sort \fR=\fPbool
752If true, written verify blocks are sorted if \fBfio\fR deems it to be faster to
753read them back in a sorted manner. Default: true.
754.TP
Jens Axboef7fa2652009-03-09 14:20:20 +0100755.BI verify_offset \fR=\fPint
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200756Swap the verification header with data somewhere else in the block before
Aaron Carrolld1429b52007-09-18 08:10:57 +0200757writing. It is swapped back before verifying.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200758.TP
Jens Axboef7fa2652009-03-09 14:20:20 +0100759.BI verify_interval \fR=\fPint
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200760Write the verification header for this number of bytes, which should divide
761\fBblocksize\fR. Default: \fBblocksize\fR.
762.TP
Jens Axboe996093b2010-06-24 08:37:13 +0200763.BI verify_pattern \fR=\fPstr
764If set, fio will fill the io buffers with this pattern. Fio defaults to filling
765with totally random bytes, but sometimes it's interesting to fill with a known
766pattern for io verification purposes. Depending on the width of the pattern,
767fio will fill 1/2/3/4 bytes of the buffer at the time(it can be either a
768decimal or a hex number). The verify_pattern if larger than a 32-bit quantity
769has to be a hex number that starts with either "0x" or "0X". Use with
770\fBverify\fP=meta.
771.TP
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200772.BI verify_fatal \fR=\fPbool
773If true, exit the job on the first observed verification failure. Default:
774false.
775.TP
Jens Axboeb463e932011-01-12 09:03:23 +0100776.BI verify_dump \fR=\fPbool
777If set, dump the contents of both the original data block and the data block we
778read off disk to files. This allows later analysis to inspect just what kind of
Jens Axboeef71e312011-10-25 22:43:36 +0200779data corruption occurred. Off by default.
Jens Axboeb463e932011-01-12 09:03:23 +0100780.TP
Jens Axboee8462bd2009-07-06 12:59:04 +0200781.BI verify_async \fR=\fPint
782Fio will normally verify IO inline from the submitting thread. This option
783takes an integer describing how many async offload threads to create for IO
784verification instead, causing fio to offload the duty of verifying IO contents
Jens Axboec85c3242009-07-06 14:12:57 +0200785to one or more separate threads. If using this offload option, even sync IO
786engines can benefit from using an \fBiodepth\fR setting higher than 1, as it
787allows them to have IO in flight while verifies are running.
Jens Axboee8462bd2009-07-06 12:59:04 +0200788.TP
789.BI verify_async_cpus \fR=\fPstr
790Tell fio to set the given CPU affinity on the async IO verification threads.
791See \fBcpus_allowed\fP for the format used.
792.TP
Jens Axboe6f874182010-06-21 12:53:26 +0200793.BI verify_backlog \fR=\fPint
794Fio will normally verify the written contents of a job that utilizes verify
795once that job has completed. In other words, everything is written then
796everything is read back and verified. You may want to verify continually
797instead for a variety of reasons. Fio stores the meta data associated with an
798IO block in memory, so for large verify workloads, quite a bit of memory would
David Nellans092f7072010-10-26 08:08:42 -0600799be used up holding this meta data. If this option is enabled, fio will write
800only N blocks before verifying these blocks.
Jens Axboe6f874182010-06-21 12:53:26 +0200801.TP
802.BI verify_backlog_batch \fR=\fPint
803Control how many blocks fio will verify if verify_backlog is set. If not set,
804will default to the value of \fBverify_backlog\fR (meaning the entire queue is
David Nellans092f7072010-10-26 08:08:42 -0600805read back and verified). If \fBverify_backlog_batch\fR is less than
806\fBverify_backlog\fR then not all blocks will be verified, if
807\fBverify_backlog_batch\fR is larger than \fBverify_backlog\fR, some blocks
808will be verified more than once.
Jens Axboe6f874182010-06-21 12:53:26 +0200809.TP
Jens Axboed3923652011-08-03 12:38:39 +0200810.B stonewall "\fR,\fP wait_for_previous"
Martin Steigerwald5982a922011-06-27 16:07:24 +0200811Wait for preceding jobs in the job file to exit before starting this one.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200812\fBstonewall\fR implies \fBnew_group\fR.
813.TP
814.B new_group
815Start a new reporting group. If not given, all jobs in a file will be part
816of the same reporting group, unless separated by a stonewall.
817.TP
818.BI numjobs \fR=\fPint
819Number of clones (processes/threads performing the same workload) of this job.
820Default: 1.
821.TP
822.B group_reporting
823If set, display per-group reports instead of per-job when \fBnumjobs\fR is
824specified.
825.TP
826.B thread
827Use threads created with \fBpthread_create\fR\|(3) instead of processes created
828with \fBfork\fR\|(2).
829.TP
Jens Axboef7fa2652009-03-09 14:20:20 +0100830.BI zonesize \fR=\fPint
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200831Divide file into zones of the specified size in bytes. See \fBzoneskip\fR.
832.TP
Jens Axboef7fa2652009-03-09 14:20:20 +0100833.BI zoneskip \fR=\fPint
Aaron Carrolld1429b52007-09-18 08:10:57 +0200834Skip the specified number of bytes when \fBzonesize\fR bytes of data have been
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200835read.
836.TP
837.BI write_iolog \fR=\fPstr
Stefan Hajnoczi5b42a482011-01-08 20:28:41 +0100838Write the issued I/O patterns to the specified file. Specify a separate file
839for each job, otherwise the iologs will be interspersed and the file may be
840corrupt.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200841.TP
842.BI read_iolog \fR=\fPstr
843Replay the I/O patterns contained in the specified file generated by
844\fBwrite_iolog\fR, or may be a \fBblktrace\fR binary file.
845.TP
David Nellans64bbb862010-08-24 22:13:30 +0200846.BI replay_no_stall \fR=\fPint
847While replaying I/O patterns using \fBread_iolog\fR the default behavior
848attempts to respect timing information between I/Os. Enabling
849\fBreplay_no_stall\fR causes I/Os to be replayed as fast as possible while
850still respecting ordering.
851.TP
David Nellansd1c46c02010-08-31 21:20:47 +0200852.BI replay_redirect \fR=\fPstr
853While replaying I/O patterns using \fBread_iolog\fR the default behavior
854is to replay the IOPS onto the major/minor device that each IOP was recorded
855from. Setting \fBreplay_redirect\fR causes all IOPS to be replayed onto the
856single specified device regardless of the device it was recorded from.
857.TP
Steven Noonan836bad52011-09-14 09:21:33 +0200858.BI write_bw_log \fR=\fPstr
Jens Axboe901bb992009-03-14 20:17:36 +0100859If given, write a bandwidth log of the jobs in this job file. Can be used to
860store data of the bandwidth of the jobs in their lifetime. The included
861fio_generate_plots script uses gnuplot to turn these text files into nice
862graphs. See \fBwrite_log_log\fR for behaviour of given filename. For this
863option, the postfix is _bw.log.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200864.TP
Steven Noonan836bad52011-09-14 09:21:33 +0200865.BI write_lat_log \fR=\fPstr
Jens Axboe901bb992009-03-14 20:17:36 +0100866Same as \fBwrite_bw_log\fR, but writes I/O completion latencies. If no
867filename is given with this option, the default filename of "jobname_type.log"
868is used. Even if the filename is given, fio will still append the type of log.
869.TP
Jens Axboec8eeb9d2011-10-05 14:02:22 +0200870.BI write_iops_log \fR=\fPstr
871Same as \fBwrite_bw_log\fR, but writes IOPS. If no filename is given with this
872option, the default filename of "jobname_type.log" is used. Even if the
873filename is given, fio will still append the type of log.
874.TP
Steven Noonan836bad52011-09-14 09:21:33 +0200875.BI disable_lat \fR=\fPbool
Jens Axboe02af0982010-06-24 09:59:34 +0200876Disable measurements of total latency numbers. Useful only for cutting
Jens Axboe901bb992009-03-14 20:17:36 +0100877back the number of calls to gettimeofday, as that does impact performance at
878really high IOPS rates. Note that to really get rid of a large amount of these
879calls, this option must be used with disable_slat and disable_bw as well.
880.TP
Steven Noonan836bad52011-09-14 09:21:33 +0200881.BI disable_clat \fR=\fPbool
Steven Noonanc95f9da2011-06-22 09:47:09 +0200882Disable measurements of completion latency numbers. See \fBdisable_lat\fR.
Jens Axboe02af0982010-06-24 09:59:34 +0200883.TP
Steven Noonan836bad52011-09-14 09:21:33 +0200884.BI disable_slat \fR=\fPbool
Jens Axboe02af0982010-06-24 09:59:34 +0200885Disable measurements of submission latency numbers. See \fBdisable_lat\fR.
Jens Axboe901bb992009-03-14 20:17:36 +0100886.TP
Steven Noonan836bad52011-09-14 09:21:33 +0200887.BI disable_bw_measurement \fR=\fPbool
Jens Axboe02af0982010-06-24 09:59:34 +0200888Disable measurements of throughput/bandwidth numbers. See \fBdisable_lat\fR.
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200889.TP
Jens Axboef7fa2652009-03-09 14:20:20 +0100890.BI lockmem \fR=\fPint
Aaron Carrolld60e92d2007-09-17 10:32:59 +0200891Pin the specified amount of memory with \fBmlock\fR\|(2). Can be used to
892simulate a smaller amount of memory.
893.TP
894.BI exec_prerun \fR=\fPstr
895Before running the job, execute the specified command with \fBsystem\fR\|(3).
896.TP
897.BI exec_postrun \fR=\fPstr
898Same as \fBexec_prerun\fR, but the command is executed after the job completes.
899.TP
900.BI ioscheduler \fR=\fPstr
901Attempt to switch the device hosting the file to the specified I/O scheduler.
902.TP
903.BI cpuload \fR=\fPint
904If the job is a CPU cycle-eater, attempt to use the specified percentage of
905CPU cycles.
906.TP
907.BI cpuchunks \fR=\fPint
908If the job is a CPU cycle-eater, split the load into cycles of the
909given time in milliseconds.
910.TP
911.BI disk_util \fR=\fPbool
Aaron Carrolld1429b52007-09-18 08:10:57 +0200912Generate disk utilization statistics if the platform supports it. Default: true.
Jens Axboe901bb992009-03-14 20:17:36 +0100913.TP
914.BI gtod_reduce \fR=\fPbool
915Enable all of the gettimeofday() reducing options (disable_clat, disable_slat,
916disable_bw) plus reduce precision of the timeout somewhat to really shrink the
917gettimeofday() call count. With this option enabled, we only do about 0.4% of
918the gtod() calls we would have done if all time keeping was enabled.
919.TP
920.BI gtod_cpu \fR=\fPint
921Sometimes it's cheaper to dedicate a single thread of execution to just getting
922the current time. Fio (and databases, for instance) are very intensive on
923gettimeofday() calls. With this option, you can set one CPU aside for doing
924nothing but logging current time to a shared memory location. Then the other
925threads/processes that run IO workloads need only copy that segment, instead of
926entering the kernel with a gettimeofday() call. The CPU set aside for doing
927these time calls will be excluded from other uses. Fio will manually clear it
928from the CPU mask of other jobs.
Radha Ramachandranf2bba182009-06-15 08:40:16 +0200929.TP
Jens Axboea696fa22009-12-04 10:05:02 +0100930.BI cgroup \fR=\fPstr
931Add job to this control group. If it doesn't exist, it will be created.
Jens Axboe6adb38a2009-12-07 08:01:26 +0100932The system must have a mounted cgroup blkio mount point for this to work. If
933your system doesn't have it mounted, you can do so with:
934
Martin Steigerwald5982a922011-06-27 16:07:24 +0200935# mount \-t cgroup \-o blkio none /cgroup
Jens Axboea696fa22009-12-04 10:05:02 +0100936.TP
937.BI cgroup_weight \fR=\fPint
938Set the weight of the cgroup to this value. See the documentation that comes
939with the kernel, allowed values are in the range of 100..1000.
Jens Axboee0b0d892009-12-08 10:10:14 +0100940.TP
Vivek Goyal7de87092010-03-31 22:55:15 +0200941.BI cgroup_nodelete \fR=\fPbool
942Normally fio will delete the cgroups it has created after the job completion.
943To override this behavior and to leave cgroups around after the job completion,
944set cgroup_nodelete=1. This can be useful if one wants to inspect various
945cgroup files after job completion. Default: false
946.TP
Jens Axboee0b0d892009-12-08 10:10:14 +0100947.BI uid \fR=\fPint
948Instead of running as the invoking user, set the user ID to this value before
949the thread/process does any work.
950.TP
951.BI gid \fR=\fPint
952Set group ID, see \fBuid\fR.
Yu-ju Hong83349192011-08-13 00:53:44 +0200953.TP
954.BI clat_percentiles \fR=\fPbool
955Enable the reporting of percentiles of completion latencies.
956.TP
957.BI percentile_list \fR=\fPfloat_list
958Overwrite the default list of percentiles for completion
959latencies. Each number is a floating number in the range (0,100], and
960the maximum length of the list is 20. Use ':' to separate the
Martin Steigerwald3eb07282011-10-05 11:41:54 +0200961numbers. For example, \-\-percentile_list=99.5:99.9 will cause fio to
Yu-ju Hong83349192011-08-13 00:53:44 +0200962report the values of completion latency below which 99.5% and 99.9% of
963the observed latencies fell, respectively.
Steven Langde890a12011-11-09 14:03:34 +0100964.SS "Ioengine Parameters List"
965Some parameters are only valid when a specific ioengine is in use. These are
966used identically to normal parameters, with the caveat that when used on the
967command line, the must come after the ioengine that defines them is selected.
968.TP
969.BI (libaio)userspace_reap
970Normally, with the libaio engine in use, fio will use
971the io_getevents system call to reap newly returned events.
972With this flag turned on, the AIO ring will be read directly
973from user-space to reap events. The reaping mode is only
974enabled when polling for a minimum of 0 events (eg when
975iodepth_batch_complete=0).
976.TP
977.BI (net,netsplice)hostname \fR=\fPstr
978The host name or IP address to use for TCP or UDP based IO.
979If the job is a TCP listener or UDP reader, the hostname is not
980used and must be omitted.
981.TP
982.BI (net,netsplice)port \fR=\fPint
983The TCP or UDP port to bind to or connect to.
984.TP
985.BI (net,netsplice)protocol \fR=\fPstr "\fR,\fP proto" \fR=\fPstr
986The network protocol to use. Accepted values are:
987.RS
988.RS
989.TP
990.B tcp
991Transmission control protocol
992.TP
993.B udp
994Unreliable datagram protocol
995.TP
996.B unix
997UNIX domain socket
998.RE
999.P
1000When the protocol is TCP or UDP, the port must also be given,
1001as well as the hostname if the job is a TCP listener or UDP
1002reader. For unix sockets, the normal filename option should be
1003used and the port is invalid.
1004.RE
1005.TP
1006.BI (net,netsplice)listen
1007For TCP network connections, tell fio to listen for incoming
1008connections rather than initiating an outgoing connection. The
1009hostname must be omitted if this option is used.
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001010.SH OUTPUT
Aaron Carrolld1429b52007-09-18 08:10:57 +02001011While running, \fBfio\fR will display the status of the created jobs. For
1012example:
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001013.RS
Aaron Carrolld1429b52007-09-18 08:10:57 +02001014.P
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001015Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
1016.RE
1017.P
Aaron Carrolld1429b52007-09-18 08:10:57 +02001018The characters in the first set of brackets denote the current status of each
1019threads. The possible values are:
1020.P
1021.PD 0
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001022.RS
1023.TP
1024.B P
1025Setup but not started.
1026.TP
1027.B C
1028Thread created.
1029.TP
1030.B I
1031Initialized, waiting.
1032.TP
1033.B R
1034Running, doing sequential reads.
1035.TP
1036.B r
1037Running, doing random reads.
1038.TP
1039.B W
1040Running, doing sequential writes.
1041.TP
1042.B w
1043Running, doing random writes.
1044.TP
1045.B M
1046Running, doing mixed sequential reads/writes.
1047.TP
1048.B m
1049Running, doing mixed random reads/writes.
1050.TP
1051.B F
1052Running, currently waiting for \fBfsync\fR\|(2).
1053.TP
1054.B V
1055Running, verifying written data.
1056.TP
1057.B E
1058Exited, not reaped by main thread.
1059.TP
1060.B \-
1061Exited, thread reaped.
1062.RE
Aaron Carrolld1429b52007-09-18 08:10:57 +02001063.PD
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001064.P
1065The second set of brackets shows the estimated completion percentage of
1066the current group. The third set shows the read and write I/O rate,
1067respectively. Finally, the estimated run time of the job is displayed.
1068.P
1069When \fBfio\fR completes (or is interrupted by Ctrl-C), it will show data
1070for each thread, each group of threads, and each disk, in that order.
1071.P
1072Per-thread statistics first show the threads client number, group-id, and
1073error code. The remaining figures are as follows:
1074.RS
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001075.TP
1076.B io
1077Number of megabytes of I/O performed.
1078.TP
1079.B bw
1080Average data rate (bandwidth).
1081.TP
1082.B runt
1083Threads run time.
1084.TP
1085.B slat
1086Submission latency minimum, maximum, average and standard deviation. This is
1087the time it took to submit the I/O.
1088.TP
1089.B clat
1090Completion latency minimum, maximum, average and standard deviation. This
1091is the time between submission and completion.
1092.TP
1093.B bw
1094Bandwidth minimum, maximum, percentage of aggregate bandwidth received, average
1095and standard deviation.
1096.TP
1097.B cpu
1098CPU usage statistics. Includes user and system time, number of context switches
1099this thread went through and number of major and minor page faults.
1100.TP
1101.B IO depths
1102Distribution of I/O depths. Each depth includes everything less than (or equal)
1103to it, but greater than the previous depth.
1104.TP
1105.B IO issued
1106Number of read/write requests issued, and number of short read/write requests.
1107.TP
1108.B IO latencies
1109Distribution of I/O completion latencies. The numbers follow the same pattern
1110as \fBIO depths\fR.
1111.RE
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001112.P
1113The group statistics show:
Aaron Carrolld1429b52007-09-18 08:10:57 +02001114.PD 0
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001115.RS
1116.TP
1117.B io
1118Number of megabytes I/O performed.
1119.TP
1120.B aggrb
1121Aggregate bandwidth of threads in the group.
1122.TP
1123.B minb
1124Minimum average bandwidth a thread saw.
1125.TP
1126.B maxb
1127Maximum average bandwidth a thread saw.
1128.TP
1129.B mint
Aaron Carrolld1429b52007-09-18 08:10:57 +02001130Shortest runtime of threads in the group.
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001131.TP
1132.B maxt
1133Longest runtime of threads in the group.
1134.RE
Aaron Carrolld1429b52007-09-18 08:10:57 +02001135.PD
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001136.P
1137Finally, disk statistics are printed with reads first:
Aaron Carrolld1429b52007-09-18 08:10:57 +02001138.PD 0
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001139.RS
1140.TP
1141.B ios
1142Number of I/Os performed by all groups.
1143.TP
1144.B merge
1145Number of merges in the I/O scheduler.
1146.TP
1147.B ticks
1148Number of ticks we kept the disk busy.
1149.TP
1150.B io_queue
1151Total time spent in the disk queue.
1152.TP
1153.B util
1154Disk utilization.
1155.RE
Aaron Carrolld1429b52007-09-18 08:10:57 +02001156.PD
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001157.SH TERSE OUTPUT
1158If the \fB\-\-minimal\fR option is given, the results will be printed in a
David Nellans562c2d22010-09-23 08:38:17 +02001159semicolon-delimited format suitable for scripted use - a job description
1160(if provided) follows on a new line. Note that the first
Jens Axboe525c2bf2010-06-30 15:22:21 +02001161number in the line is the version number. If the output has to be changed
1162for some reason, this number will be incremented by 1 to signify that
1163change. The fields are:
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001164.P
1165.RS
Jens Axboe5e726d02011-10-14 08:08:10 +02001166.B terse version, fio version, jobname, groupid, error
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001167.P
1168Read status:
1169.RS
Jens Axboe312b4af2011-10-13 13:11:42 +02001170.B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, IOPS, runtime \fR(ms)\fP
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001171.P
1172Submission latency:
1173.RS
1174.B min, max, mean, standard deviation
1175.RE
1176Completion latency:
1177.RS
1178.B min, max, mean, standard deviation
1179.RE
Jens Axboe1db92cb2011-10-13 13:43:36 +02001180Completion latency percentiles (20 fields):
1181.RS
1182.B Xth percentile=usec
1183.RE
Jens Axboe525c2bf2010-06-30 15:22:21 +02001184Total latency:
1185.RS
1186.B min, max, mean, standard deviation
1187.RE
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001188Bandwidth:
1189.RS
1190.B min, max, aggregate percentage of total, mean, standard deviation
1191.RE
1192.RE
1193.P
1194Write status:
1195.RS
Jens Axboe312b4af2011-10-13 13:11:42 +02001196.B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, IOPS, runtime \fR(ms)\fP
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001197.P
1198Submission latency:
1199.RS
1200.B min, max, mean, standard deviation
1201.RE
1202Completion latency:
1203.RS
1204.B min, max, mean, standard deviation
1205.RE
Jens Axboe1db92cb2011-10-13 13:43:36 +02001206Completion latency percentiles (20 fields):
1207.RS
1208.B Xth percentile=usec
1209.RE
Jens Axboe525c2bf2010-06-30 15:22:21 +02001210Total latency:
1211.RS
1212.B min, max, mean, standard deviation
1213.RE
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001214Bandwidth:
1215.RS
1216.B min, max, aggregate percentage of total, mean, standard deviation
1217.RE
1218.RE
1219.P
Aaron Carrolld1429b52007-09-18 08:10:57 +02001220CPU usage:
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001221.RS
Carl Henrik Lundebd2626f2008-06-12 09:17:46 +02001222.B user, system, context switches, major page faults, minor page faults
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001223.RE
1224.P
1225IO depth distribution:
1226.RS
1227.B <=1, 2, 4, 8, 16, 32, >=64
1228.RE
1229.P
David Nellans562c2d22010-09-23 08:38:17 +02001230IO latency distribution:
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001231.RS
David Nellans562c2d22010-09-23 08:38:17 +02001232Microseconds:
1233.RS
1234.B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
1235.RE
1236Milliseconds:
1237.RS
1238.B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
1239.RE
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001240.RE
1241.P
Jens Axboef2f788d2011-10-13 14:03:52 +02001242Disk utilization (1 for each disk used):
1243.RS
1244.B name, read ios, write ios, read merges, write merges, read ticks, write ticks, read in-queue time, write in-queue time, disk utilization percentage
1245.RE
1246.P
Martin Steigerwald5982a922011-06-27 16:07:24 +02001247Error Info (dependent on continue_on_error, default off):
David Nellans562c2d22010-09-23 08:38:17 +02001248.RS
1249.B total # errors, first error code
1250.RE
1251.P
1252.B text description (if provided in config - appears on newline)
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001253.RE
Jens Axboe49da1242011-10-13 20:17:02 +02001254.SH CLIENT / SERVER
1255Normally you would run fio as a stand-alone application on the machine
1256where the IO workload should be generated. However, it is also possible to
1257run the frontend and backend of fio separately. This makes it possible to
1258have a fio server running on the machine(s) where the IO workload should
1259be running, while controlling it from another machine.
1260
1261To start the server, you would do:
1262
1263\fBfio \-\-server=args\fR
1264
1265on that machine, where args defines what fio listens to. The arguments
Jens Axboe811826b2011-10-24 09:11:50 +02001266are of the form 'type:hostname or IP:port'. 'type' is either 'ip' (or ip4)
1267for TCP/IP v4, 'ip6' for TCP/IP v6, or 'sock' for a local unix domain socket.
1268'hostname' is either a hostname or IP address, and 'port' is the port to
1269listen to (only valid for TCP/IP, not a local socket). Some examples:
Jens Axboe49da1242011-10-13 20:17:02 +02001270
12711) fio --server
1272
1273 Start a fio server, listening on all interfaces on the default port (8765).
1274
Jens Axboe811826b2011-10-24 09:11:50 +020012752) fio --server=ip:hostname,4444
Jens Axboe49da1242011-10-13 20:17:02 +02001276
1277 Start a fio server, listening on IP belonging to hostname and on port 4444.
1278
Jens Axboe811826b2011-10-24 09:11:50 +020012793) fio --server=ip6:::1,4444
1280
1281 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
1282
12834) fio --server=,4444
Jens Axboe49da1242011-10-13 20:17:02 +02001284
1285 Start a fio server, listening on all interfaces on port 4444.
1286
Jens Axboe811826b2011-10-24 09:11:50 +020012875) fio --server=1.2.3.4
Jens Axboe49da1242011-10-13 20:17:02 +02001288
1289 Start a fio server, listening on IP 1.2.3.4 on the default port.
1290
Jens Axboe811826b2011-10-24 09:11:50 +020012916) fio --server=sock:/tmp/fio.sock
Jens Axboe49da1242011-10-13 20:17:02 +02001292
1293 Start a fio server, listening on the local socket /tmp/fio.sock.
1294
1295When a server is running, you can connect to it from a client. The client
1296is run with:
1297
1298fio --local-args --client=server --remote-args <job file(s)>
1299
1300where --local-args are arguments that are local to the client where it is
1301running, 'server' is the connect string, and --remote-args and <job file(s)>
1302are sent to the server. The 'server' string follows the same format as it
1303does on the server side, to allow IP/hostname/socket and port strings.
1304You can connect to multiple clients as well, to do that you could run:
1305
1306fio --client=server2 --client=server2 <job file(s)>
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001307.SH AUTHORS
Jens Axboe49da1242011-10-13 20:17:02 +02001308
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001309.B fio
Jens Axboeaa58d252010-06-09 09:49:38 +02001310was written by Jens Axboe <jens.axboe@oracle.com>,
1311now Jens Axboe <jaxboe@fusionio.com>.
Aaron Carrolld1429b52007-09-18 08:10:57 +02001312.br
1313This man page was written by Aaron Carroll <aaronc@cse.unsw.edu.au> based
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001314on documentation by Jens Axboe.
1315.SH "REPORTING BUGS"
Jens Axboe482900c2009-06-02 12:15:51 +02001316Report bugs to the \fBfio\fR mailing list <fio@vger.kernel.org>.
Aaron Carrolld1429b52007-09-18 08:10:57 +02001317See \fBREADME\fR.
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001318.SH "SEE ALSO"
Aaron Carrolld1429b52007-09-18 08:10:57 +02001319For further documentation see \fBHOWTO\fR and \fBREADME\fR.
1320.br
1321Sample jobfiles are available in the \fBexamples\fR directory.
Aaron Carrolld60e92d2007-09-17 10:32:59 +02001322