Spelling fix. Reported by DebianĀ“s lintian.
[fio.git] / fio.1
CommitLineData
f8b8f7da 1.TH fio 1 "December 2014" "User Manual"
d60e92d1
AC
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
49da1240
JA
15.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
d60e92d1
AC
20.BI \-\-output \fR=\fPfilename
21Write output to \fIfilename\fR.
22.TP
e28ee21d 23.BI \-\-output-format \fR=\fPformat
513e37ee
VF
24Set the reporting format to \fInormal\fR, \fIterse\fR, \fIjson\fR, or
25\fIjson+\fR. Multiple formats can be selected, separate by a comma. \fIterse\fR
26is a CSV based format. \fIjson+\fR is like \fIjson\fR, except it adds a full
27dump of the latency buckets.
e28ee21d 28.TP
b2cecdc2 29.BI \-\-runtime \fR=\fPruntime
30Limit run time to \fIruntime\fR seconds.
d60e92d1 31.TP
d60e92d1
AC
32.B \-\-bandwidth\-log
33Generate per-job bandwidth logs.
34.TP
35.B \-\-minimal
d1429b5c 36Print statistics in a terse, semicolon-delimited format.
d60e92d1 37.TP
f6a7df53
JA
38.B \-\-append-terse
39Print statistics in selected mode AND terse, semicolon-delimited format.
40Deprecated, use \-\-output-format instead to select multiple formats.
41.TP
49da1240
JA
42.B \-\-version
43Display version information and exit.
44.TP
065248bf 45.BI \-\-terse\-version \fR=\fPversion
4d658652 46Set terse version output format (Current version 3, or older version 2).
49da1240
JA
47.TP
48.B \-\-help
49Display usage information and exit.
50.TP
fec0f21c
JA
51.B \-\-cpuclock-test
52Perform test and validation of internal CPU clock
53.TP
54.BI \-\-crctest[\fR=\fPtest]
55Test the speed of the builtin checksumming functions. If no argument is given,
56all of them are tested. Or a comma separated list can be passed, in which
57case the given ones are tested.
58.TP
49da1240
JA
59.BI \-\-cmdhelp \fR=\fPcommand
60Print help information for \fIcommand\fR. May be `all' for all commands.
61.TP
de890a1e
SL
62.BI \-\-enghelp \fR=\fPioengine[,command]
63List all commands defined by \fIioengine\fR, or print help for \fIcommand\fR defined by \fIioengine\fR.
64.TP
d60e92d1
AC
65.BI \-\-showcmd \fR=\fPjobfile
66Convert \fIjobfile\fR to a set of command-line options.
67.TP
d60e92d1
AC
68.BI \-\-eta \fR=\fPwhen
69Specifies when real-time ETA estimate should be printed. \fIwhen\fR may
70be one of `always', `never' or `auto'.
71.TP
30b5d57f
JA
72.BI \-\-eta\-newline \fR=\fPtime
73Force an ETA newline for every `time` period passed.
74.TP
75.BI \-\-status\-interval \fR=\fPtime
76Report full output status every `time` period passed.
77.TP
49da1240
JA
78.BI \-\-readonly
79Turn on safety read-only checks, preventing any attempted write.
80.TP
c0a5d35e 81.BI \-\-section \fR=\fPsec
cf145d90 82Only run section \fIsec\fR from job file. This option can be used multiple times to add more sections to run.
c0a5d35e 83.TP
49da1240
JA
84.BI \-\-alloc\-size \fR=\fPkb
85Set the internal smalloc pool size to \fIkb\fP kilobytes.
d60e92d1 86.TP
49da1240
JA
87.BI \-\-warnings\-fatal
88All fio parser warnings are fatal, causing fio to exit with an error.
9183788d 89.TP
49da1240 90.BI \-\-max\-jobs \fR=\fPnr
57e118a2 91Set the maximum allowed number of jobs (threads/processes) to support.
d60e92d1 92.TP
49da1240
JA
93.BI \-\-server \fR=\fPargs
94Start a backend server, with \fIargs\fP specifying what to listen to. See client/server section.
f57a9c59 95.TP
49da1240
JA
96.BI \-\-daemonize \fR=\fPpidfile
97Background a fio server, writing the pid to the given pid file.
98.TP
99.BI \-\-client \fR=\fPhost
39b5f61e 100Instead of running the jobs locally, send and run them on the given host or set of hosts. See client/server section.
f2a2ce0e
HL
101.TP
102.BI \-\-idle\-prof \fR=\fPoption
103Report cpu idleness on a system or percpu basis (\fIoption\fP=system,percpu) or run unit work calibration only (\fIoption\fP=calibrate).
d60e92d1
AC
104.SH "JOB FILE FORMAT"
105Job files are in `ini' format. They consist of one or more
106job definitions, which begin with a job name in square brackets and
107extend to the next job name. The job name can be any ASCII string
108except `global', which has a special meaning. Following the job name is
109a sequence of zero or more parameters, one per line, that define the
110behavior of the job. Any line starting with a `;' or `#' character is
d1429b5c 111considered a comment and ignored.
d9956b64
AC
112.P
113If \fIjobfile\fR is specified as `-', the job file will be read from
114standard input.
d60e92d1
AC
115.SS "Global Section"
116The global section contains default parameters for jobs specified in the
117job file. A job is only affected by global sections residing above it,
118and there may be any number of global sections. Specific job definitions
119may override any parameter set in global sections.
120.SH "JOB PARAMETERS"
121.SS Types
b470a02c
SC
122Some parameters may take arguments of a specific type.
123Anywhere a numeric value is required, an arithmetic expression may be used,
d59aa780
JA
124provided it is surrounded by parentheses. Supported operators are:
125.RS
126.RS
127.TP
128.B addition (+)
129.TP
130.B subtraction (-)
131.TP
132.B multiplication (*)
133.TP
134.B division (/)
135.TP
136.B modulus (%)
137.TP
138.B exponentiation (^)
139.RE
140.RE
141.P
142For time values in expressions, units are microseconds by default. This is
143different than for time values not in expressions (not enclosed in
144parentheses). The types used are:
d60e92d1
AC
145.TP
146.I str
147String: a sequence of alphanumeric characters.
148.TP
149.I int
d60e92d1 150SI integer: a whole number, possibly containing a suffix denoting the base unit
b09da8fa
JA
151of the value. Accepted suffixes are `k', 'M', 'G', 'T', and 'P', denoting
152kilo (1024), mega (1024^2), giga (1024^3), tera (1024^4), and peta (1024^5)
74454ce4
CE
153respectively. If prefixed with '0x', the value is assumed to be base 16
154(hexadecimal). A suffix may include a trailing 'b', for instance 'kb' is
155identical to 'k'. You can specify a base 10 value by using 'KiB', 'MiB','GiB',
156etc. This is useful for disk drives where values are often given in base 10
157values. Specifying '30GiB' will get you 30*1000^3 bytes.
158When specifying times the default suffix meaning changes, still denoting the
159base unit of the value, but accepted suffixes are 'D' (days), 'H' (hours), 'M'
0de5b26f
JA
160(minutes), 'S' Seconds, 'ms' (or msec) milli seconds, 'us' (or 'usec') micro
161seconds. Time values without a unit specify seconds.
74454ce4 162The suffixes are not case sensitive.
d60e92d1
AC
163.TP
164.I bool
165Boolean: a true or false value. `0' denotes false, `1' denotes true.
166.TP
167.I irange
168Integer range: a range of integers specified in the format
d1429b5c
AC
169\fIlower\fR:\fIupper\fR or \fIlower\fR\-\fIupper\fR. \fIlower\fR and
170\fIupper\fR may contain a suffix as described above. If an option allows two
171sets of ranges, they are separated with a `,' or `/' character. For example:
172`8\-8k/8M\-4G'.
83349190
YH
173.TP
174.I float_list
175List of floating numbers: A list of floating numbers, separated by
cecbfd47 176a ':' character.
d60e92d1
AC
177.SS "Parameter List"
178.TP
179.BI name \fR=\fPstr
d9956b64 180May be used to override the job name. On the command line, this parameter
d60e92d1
AC
181has the special purpose of signalling the start of a new job.
182.TP
9cc8cb91
AK
183.BI wait_for \fR=\fPstr
184Specifies the name of the already defined job to wait for. Single waitee name
185only may be specified. If set, the job won't be started until all workers of
186the waitee job are done. Wait_for operates on the job name basis, so there are
187a few limitations. First, the waitee must be defined prior to the waiter job
188(meaning no forward references). Second, if a job is being referenced as a
189waitee, it must have a unique name (no duplicate waitees).
190.TP
d60e92d1
AC
191.BI description \fR=\fPstr
192Human-readable description of the job. It is printed when the job is run, but
193otherwise has no special purpose.
194.TP
195.BI directory \fR=\fPstr
196Prefix filenames with this directory. Used to place files in a location other
197than `./'.
bcbfeefa
CE
198You can specify a number of directories by separating the names with a ':'
199character. These directories will be assigned equally distributed to job clones
200creates with \fInumjobs\fR as long as they are using generated filenames.
201If specific \fIfilename(s)\fR are set fio will use the first listed directory,
202and thereby matching the \fIfilename\fR semantic which generates a file each
67445b63
JA
203clone if not specified, but let all clones use the same if set. See
204\fIfilename\fR for considerations regarding escaping certain characters on
205some platforms.
d60e92d1
AC
206.TP
207.BI filename \fR=\fPstr
208.B fio
209normally makes up a file name based on the job name, thread number, and file
d1429b5c 210number. If you want to share files between threads in a job or several jobs,
de890a1e
SL
211specify a \fIfilename\fR for each of them to override the default.
212If the I/O engine is file-based, you can specify
d1429b5c
AC
213a number of files by separating the names with a `:' character. `\-' is a
214reserved name, meaning stdin or stdout, depending on the read/write direction
67445b63
JA
215set. On Windows, disk devices are accessed as \\.\PhysicalDrive0 for the first
216device, \\.\PhysicalDrive1 for the second etc. Note: Windows and FreeBSD
217prevent write access to areas of the disk containing in-use data
218(e.g. filesystems). If the wanted filename does need to include a colon, then
4904acd5
JM
219escape that with a '\\' character. For instance, if the filename is
220"/dev/dsk/foo@3,0:c", then you would use filename="/dev/dsk/foo@3,0\\:c".
d60e92d1 221.TP
de98bd30 222.BI filename_format \fR=\fPstr
ce594fbe 223If sharing multiple files between jobs, it is usually necessary to have
de98bd30
JA
224fio generate the exact names that you want. By default, fio will name a file
225based on the default file format specification of
226\fBjobname.jobnumber.filenumber\fP. With this option, that can be
227customized. Fio will recognize and replace the following keywords in this
228string:
229.RS
230.RS
231.TP
232.B $jobname
233The name of the worker thread or process.
234.TP
235.B $jobnum
236The incremental number of the worker thread or process.
237.TP
238.B $filenum
239The incremental number of the file for that worker thread or process.
240.RE
241.P
242To have dependent jobs share a set of files, this option can be set to
243have fio generate filenames that are shared between the two. For instance,
244if \fBtestfiles.$filenum\fR is specified, file number 4 for any job will
245be named \fBtestfiles.4\fR. The default of \fB$jobname.$jobnum.$filenum\fR
246will be used if no other format specifier is given.
247.RE
248.P
249.TP
922a5be8
JA
250.BI unique_filename \fR=\fPbool
251To avoid collisions between networked clients, fio defaults to prefixing
252any generated filenames (with a directory specified) with the source of
253the client connecting. To disable this behavior, set this option to 0.
254.TP
3ce9dcaf
JA
255.BI lockfile \fR=\fPstr
256Fio defaults to not locking any files before it does IO to them. If a file or
257file descriptor is shared, fio can serialize IO to that file to make the end
258result consistent. This is usual for emulating real workloads that share files.
259The lock modes are:
260.RS
261.RS
262.TP
263.B none
264No locking. This is the default.
265.TP
266.B exclusive
cf145d90 267Only one thread or process may do IO at a time, excluding all others.
3ce9dcaf
JA
268.TP
269.B readwrite
270Read-write locking on the file. Many readers may access the file at the same
271time, but writes get exclusive access.
272.RE
ce594fbe 273.RE
3ce9dcaf 274.P
d60e92d1
AC
275.BI opendir \fR=\fPstr
276Recursively open any files below directory \fIstr\fR.
277.TP
278.BI readwrite \fR=\fPstr "\fR,\fP rw" \fR=\fPstr
279Type of I/O pattern. Accepted values are:
280.RS
281.RS
282.TP
283.B read
d1429b5c 284Sequential reads.
d60e92d1
AC
285.TP
286.B write
d1429b5c 287Sequential writes.
d60e92d1 288.TP
fa769d44
SW
289.B trim
290Sequential trim (Linux block devices only).
291.TP
d60e92d1 292.B randread
d1429b5c 293Random reads.
d60e92d1
AC
294.TP
295.B randwrite
d1429b5c 296Random writes.
d60e92d1 297.TP
fa769d44
SW
298.B randtrim
299Random trim (Linux block devices only).
300.TP
10b023db 301.B rw, readwrite
d1429b5c 302Mixed sequential reads and writes.
d60e92d1 303.TP
ff6bb260 304.B randrw
d1429b5c 305Mixed random reads and writes.
82a90686
JA
306.TP
307.B trimwrite
308Trim and write mixed workload. Blocks will be trimmed first, then the same
309blocks will be written to.
d60e92d1
AC
310.RE
311.P
38dad62d
JA
312For mixed I/O, the default split is 50/50. For certain types of io the result
313may still be skewed a bit, since the speed may be different. It is possible to
3b7fa9ec 314specify a number of IO's to do before getting a new offset, this is done by
38dad62d
JA
315appending a `:\fI<nr>\fR to the end of the string given. For a random read, it
316would look like \fBrw=randread:8\fR for passing in an offset modifier with a
059b0802
JA
317value of 8. If the postfix is used with a sequential IO pattern, then the value
318specified will be added to the generated offset for each IO. For instance,
319using \fBrw=write:4k\fR will skip 4k for every write. It turns sequential IO
320into sequential IO with holes. See the \fBrw_sequencer\fR option.
d60e92d1
AC
321.RE
322.TP
38dad62d
JA
323.BI rw_sequencer \fR=\fPstr
324If an offset modifier is given by appending a number to the \fBrw=<str>\fR line,
325then this option controls how that number modifies the IO offset being
326generated. Accepted values are:
327.RS
328.RS
329.TP
330.B sequential
331Generate sequential offset
332.TP
333.B identical
334Generate the same offset
335.RE
336.P
337\fBsequential\fR is only useful for random IO, where fio would normally
338generate a new random offset for every IO. If you append eg 8 to randread, you
339would get a new random offset for every 8 IO's. The result would be a seek for
340only every 8 IO's, instead of for every IO. Use \fBrw=randread:8\fR to specify
341that. As sequential IO is already sequential, setting \fBsequential\fR for that
342would not result in any differences. \fBidentical\fR behaves in a similar
343fashion, except it sends the same offset 8 number of times before generating a
344new offset.
345.RE
346.P
347.TP
90fef2d1
JA
348.BI kb_base \fR=\fPint
349The base unit for a kilobyte. The defacto base is 2^10, 1024. Storage
350manufacturers like to use 10^3 or 1000 as a base ten unit instead, for obvious
5c9323fb 351reasons. Allowed values are 1024 or 1000, with 1024 being the default.
90fef2d1 352.TP
771e58be
JA
353.BI unified_rw_reporting \fR=\fPbool
354Fio normally reports statistics on a per data direction basis, meaning that
355read, write, and trim are accounted and reported separately. If this option is
cf145d90 356set fio sums the results and reports them as "mixed" instead.
771e58be 357.TP
d60e92d1 358.BI randrepeat \fR=\fPbool
56e2a5fc
CE
359Seed the random number generator used for random I/O patterns in a predictable
360way so the pattern is repeatable across runs. Default: true.
361.TP
362.BI allrandrepeat \fR=\fPbool
363Seed all random number generators in a predictable way so results are
364repeatable across runs. Default: false.
d60e92d1 365.TP
04778baf
JA
366.BI randseed \fR=\fPint
367Seed the random number generators based on this seed value, to be able to
368control what sequence of output is being generated. If not set, the random
369sequence depends on the \fBrandrepeat\fR setting.
370.TP
a596f047
EG
371.BI fallocate \fR=\fPstr
372Whether pre-allocation is performed when laying down files. Accepted values
373are:
374.RS
375.RS
376.TP
377.B none
378Do not pre-allocate space.
379.TP
380.B posix
ccc2b328 381Pre-allocate via \fBposix_fallocate\fR\|(3).
a596f047
EG
382.TP
383.B keep
ccc2b328 384Pre-allocate via \fBfallocate\fR\|(2) with FALLOC_FL_KEEP_SIZE set.
a596f047
EG
385.TP
386.B 0
387Backward-compatible alias for 'none'.
388.TP
389.B 1
390Backward-compatible alias for 'posix'.
391.RE
392.P
393May not be available on all supported platforms. 'keep' is only
394available on Linux. If using ZFS on Solaris this must be set to 'none'
395because ZFS doesn't support it. Default: 'posix'.
396.RE
7bc8c2cf 397.TP
d60e92d1 398.BI fadvise_hint \fR=\fPbool
cf145d90 399Use \fBposix_fadvise\fR\|(2) to advise the kernel what I/O patterns
d1429b5c 400are likely to be issued. Default: true.
d60e92d1 401.TP
37659335
JA
402.BI fadvise_stream \fR=\fPint
403Use \fBposix_fadvise\fR\|(2) to advise the kernel what stream ID the
404writes issued belong to. Only supported on Linux. Note, this option
405may change going forward.
406.TP
f7fa2653 407.BI size \fR=\fPint
d60e92d1 408Total size of I/O for this job. \fBfio\fR will run until this many bytes have
a4d3b4db
JA
409been transferred, unless limited by other options (\fBruntime\fR, for instance,
410or increased/descreased by \fBio_size\fR). Unless \fBnrfiles\fR and
411\fBfilesize\fR options are given, this amount will be divided between the
412available files for the job. If not set, fio will use the full size of the
413given files or devices. If the files do not exist, size must be given. It is
414also possible to give size as a percentage between 1 and 100. If size=20% is
415given, fio will use 20% of the full size of the given files or devices.
416.TP
417.BI io_size \fR=\fPint "\fR,\fB io_limit \fR=\fPint
77731b29
JA
418Normally fio operates within the region set by \fBsize\fR, which means that
419the \fBsize\fR option sets both the region and size of IO to be performed.
420Sometimes that is not what you want. With this option, it is possible to
421define just the amount of IO that fio should do. For instance, if \fBsize\fR
422is set to 20G and \fBio_limit\fR is set to 5G, fio will perform IO within
a4d3b4db
JA
423the first 20G but exit when 5G have been done. The opposite is also
424possible - if \fBsize\fR is set to 20G, and \fBio_size\fR is set to 40G, then
425fio will do 40G of IO within the 0..20G region.
d60e92d1 426.TP
74586c1e 427.BI fill_device \fR=\fPbool "\fR,\fB fill_fs" \fR=\fPbool
3ce9dcaf
JA
428Sets size to something really large and waits for ENOSPC (no space left on
429device) as the terminating condition. Only makes sense with sequential write.
430For a read workload, the mount point will be filled first then IO started on
4f12432e
JA
431the result. This option doesn't make sense if operating on a raw device node,
432since the size of that is already known by the file system. Additionally,
433writing beyond end-of-device will not return ENOSPC there.
3ce9dcaf 434.TP
d60e92d1
AC
435.BI filesize \fR=\fPirange
436Individual file sizes. May be a range, in which case \fBfio\fR will select sizes
d1429b5c
AC
437for files at random within the given range, limited to \fBsize\fR in total (if
438that is given). If \fBfilesize\fR is not specified, each created file is the
439same size.
d60e92d1 440.TP
bedc9dc2
JA
441.BI file_append \fR=\fPbool
442Perform IO after the end of the file. Normally fio will operate within the
443size of a file. If this option is set, then fio will append to the file
444instead. This has identical behavior to setting \fRoffset\fP to the size
0aae4ce7 445of a file. This option is ignored on non-regular files.
bedc9dc2 446.TP
f7fa2653 447.BI blocksize \fR=\fPint[,int] "\fR,\fB bs" \fR=\fPint[,int]
d9472271
JA
448Block size for I/O units. Default: 4k. Values for reads, writes, and trims
449can be specified separately in the format \fIread\fR,\fIwrite\fR,\fItrim\fR
450either of which may be empty to leave that value at its default. If a trailing
451comma isn't given, the remainder will inherit the last value set.
d60e92d1 452.TP
9183788d 453.BI blocksize_range \fR=\fPirange[,irange] "\fR,\fB bsrange" \fR=\fPirange[,irange]
d1429b5c
AC
454Specify a range of I/O block sizes. The issued I/O unit will always be a
455multiple of the minimum size, unless \fBblocksize_unaligned\fR is set. Applies
9183788d 456to both reads and writes if only one range is given, but can be specified
de8f6de9 457separately with a comma separating the values. Example: bsrange=1k-4k,2k-8k.
9183788d
JA
458Also (see \fBblocksize\fR).
459.TP
460.BI bssplit \fR=\fPstr
461This option allows even finer grained control of the block sizes issued,
462not just even splits between them. With this option, you can weight various
463block sizes for exact control of the issued IO for a job that has mixed
464block sizes. The format of the option is bssplit=blocksize/percentage,
5982a925 465optionally adding as many definitions as needed separated by a colon.
9183788d 466Example: bssplit=4k/10:64k/50:32k/40 would issue 50% 64k blocks, 10% 4k
c83cdd3e
JA
467blocks and 40% 32k blocks. \fBbssplit\fR also supports giving separate
468splits to reads and writes. The format is identical to what the
469\fBbs\fR option accepts, the read and write parts are separated with a
470comma.
d60e92d1
AC
471.TP
472.B blocksize_unaligned\fR,\fP bs_unaligned
d1429b5c
AC
473If set, any size in \fBblocksize_range\fR may be used. This typically won't
474work with direct I/O, as that normally requires sector alignment.
d60e92d1 475.TP
2b7a01d0 476.BI blockalign \fR=\fPint[,int] "\fR,\fB ba" \fR=\fPint[,int]
639ce0f3
MS
477At what boundary to align random IO offsets. Defaults to the same as 'blocksize'
478the minimum blocksize given. Minimum alignment is typically 512b
2b7a01d0
JA
479for using direct IO, though it usually depends on the hardware block size.
480This option is mutually exclusive with using a random map for files, so it
481will turn off that option.
43602667 482.TP
6aca9b3d
JA
483.BI bs_is_seq_rand \fR=\fPbool
484If this option is set, fio will use the normal read,write blocksize settings as
485sequential,random instead. Any random read or write will use the WRITE
486blocksize settings, and any sequential read or write will use the READ
487blocksize setting.
488.TP
d60e92d1 489.B zero_buffers
cf145d90 490Initialize buffers with all zeros. Default: fill buffers with random data.
d60e92d1 491.TP
901bb994
JA
492.B refill_buffers
493If this option is given, fio will refill the IO buffers on every submit. The
494default is to only fill it at init time and reuse that data. Only makes sense
495if zero_buffers isn't specified, naturally. If data verification is enabled,
496refill_buffers is also automatically enabled.
497.TP
fd68418e
JA
498.BI scramble_buffers \fR=\fPbool
499If \fBrefill_buffers\fR is too costly and the target is using data
500deduplication, then setting this option will slightly modify the IO buffer
501contents to defeat normal de-dupe attempts. This is not enough to defeat
502more clever block compression attempts, but it will stop naive dedupe
503of blocks. Default: true.
504.TP
c5751c62
JA
505.BI buffer_compress_percentage \fR=\fPint
506If this is set, then fio will attempt to provide IO buffer content (on WRITEs)
507that compress to the specified level. Fio does this by providing a mix of
d1af2894
JA
508random data and a fixed pattern. The fixed pattern is either zeroes, or the
509pattern specified by \fBbuffer_pattern\fR. If the pattern option is used, it
510might skew the compression ratio slightly. Note that this is per block size
511unit, for file/disk wide compression level that matches this setting. Note
512that this is per block size unit, for file/disk wide compression level that
513matches this setting, you'll also want to set refill_buffers.
c5751c62
JA
514.TP
515.BI buffer_compress_chunk \fR=\fPint
516See \fBbuffer_compress_percentage\fR. This setting allows fio to manage how
517big the ranges of random data and zeroed data is. Without this set, fio will
518provide \fBbuffer_compress_percentage\fR of blocksize random data, followed by
519the remaining zeroed. With this set to some chunk size smaller than the block
520size, fio can alternate random and zeroed data throughout the IO buffer.
521.TP
ce35b1ec 522.BI buffer_pattern \fR=\fPstr
cf145d90
CVB
523If set, fio will fill the IO buffers with this pattern. If not set, the contents
524of IO buffers is defined by the other options related to buffer contents. The
ce35b1ec 525setting can be any pattern of bytes, and can be prefixed with 0x for hex
02975b64 526values. It may also be a string, where the string must then be wrapped with
2fa5a241
RP
527"", e.g.:
528.RS
529.RS
530\fBbuffer_pattern\fR="abcd"
531.RS
532or
533.RE
534\fBbuffer_pattern\fR=-12
535.RS
536or
537.RE
538\fBbuffer_pattern\fR=0xdeadface
539.RE
540.LP
541Also you can combine everything together in any order:
542.LP
543.RS
544\fBbuffer_pattern\fR=0xdeadface"abcd"-12
545.RE
546.RE
ce35b1ec 547.TP
5c94b008
JA
548.BI dedupe_percentage \fR=\fPint
549If set, fio will generate this percentage of identical buffers when writing.
550These buffers will be naturally dedupable. The contents of the buffers depend
551on what other buffer compression settings have been set. It's possible to have
552the individual buffers either fully compressible, or not at all. This option
553only controls the distribution of unique buffers.
554.TP
d60e92d1
AC
555.BI nrfiles \fR=\fPint
556Number of files to use for this job. Default: 1.
557.TP
558.BI openfiles \fR=\fPint
559Number of files to keep open at the same time. Default: \fBnrfiles\fR.
560.TP
561.BI file_service_type \fR=\fPstr
562Defines how files to service are selected. The following types are defined:
563.RS
564.RS
565.TP
566.B random
5c9323fb 567Choose a file at random.
d60e92d1
AC
568.TP
569.B roundrobin
cf145d90 570Round robin over opened files (default).
5c9323fb 571.TP
6b7f6851
JA
572.B sequential
573Do each file in the set sequentially.
8c07860d
JA
574.TP
575.B zipf
576Use a zipfian distribution to decide what file to access.
577.TP
578.B pareto
579Use a pareto distribution to decide what file to access.
580.TP
581.B gauss
582Use a gaussian (normal) distribution to decide what file to access.
d60e92d1
AC
583.RE
584.P
8c07860d
JA
585For \fBrandom\fR, \fBroundrobin\fR, and \fBsequential\fR, a postfix can be
586appended to tell fio how many I/Os to issue before switching to a new file.
587For example, specifying \fBfile_service_type=random:8\fR would cause fio to
588issue \fI8\fR I/Os before selecting a new file at random. For the non-uniform
589distributions, a floating point postfix can be given to influence how the
590distribution is skewed. See \fBrandom_distribution\fR for a description of how
591that would work.
d60e92d1
AC
592.RE
593.TP
594.BI ioengine \fR=\fPstr
595Defines how the job issues I/O. The following types are defined:
596.RS
597.RS
598.TP
599.B sync
ccc2b328 600Basic \fBread\fR\|(2) or \fBwrite\fR\|(2) I/O. \fBfseek\fR\|(2) is used to
d60e92d1
AC
601position the I/O location.
602.TP
a31041ea 603.B psync
ccc2b328 604Basic \fBpread\fR\|(2) or \fBpwrite\fR\|(2) I/O.
a31041ea 605.TP
9183788d 606.B vsync
ccc2b328 607Basic \fBreadv\fR\|(2) or \fBwritev\fR\|(2) I/O. Will emulate queuing by
cecbfd47 608coalescing adjacent IOs into a single submission.
9183788d 609.TP
a46c5e01 610.B pvsync
ccc2b328 611Basic \fBpreadv\fR\|(2) or \fBpwritev\fR\|(2) I/O.
a46c5e01 612.TP
2cafffbe
JA
613.B pvsync2
614Basic \fBpreadv2\fR\|(2) or \fBpwritev2\fR\|(2) I/O.
615.TP
d60e92d1 616.B libaio
de890a1e 617Linux native asynchronous I/O. This ioengine defines engine specific options.
d60e92d1
AC
618.TP
619.B posixaio
ccc2b328 620POSIX asynchronous I/O using \fBaio_read\fR\|(3) and \fBaio_write\fR\|(3).
03e20d68
BC
621.TP
622.B solarisaio
623Solaris native asynchronous I/O.
624.TP
625.B windowsaio
626Windows native asynchronous I/O.
d60e92d1
AC
627.TP
628.B mmap
ccc2b328
SW
629File is memory mapped with \fBmmap\fR\|(2) and data copied using
630\fBmemcpy\fR\|(3).
d60e92d1
AC
631.TP
632.B splice
ccc2b328 633\fBsplice\fR\|(2) is used to transfer the data and \fBvmsplice\fR\|(2) to
d1429b5c 634transfer data from user-space to the kernel.
d60e92d1
AC
635.TP
636.B syslet-rw
637Use the syslet system calls to make regular read/write asynchronous.
638.TP
639.B sg
640SCSI generic sg v3 I/O. May be either synchronous using the SG_IO ioctl, or if
ccc2b328
SW
641the target is an sg character device, we use \fBread\fR\|(2) and
642\fBwrite\fR\|(2) for asynchronous I/O.
d60e92d1
AC
643.TP
644.B null
645Doesn't transfer any data, just pretends to. Mainly used to exercise \fBfio\fR
646itself and for debugging and testing purposes.
647.TP
648.B net
de890a1e
SL
649Transfer over the network. The protocol to be used can be defined with the
650\fBprotocol\fR parameter. Depending on the protocol, \fBfilename\fR,
651\fBhostname\fR, \fBport\fR, or \fBlisten\fR must be specified.
652This ioengine defines engine specific options.
d60e92d1
AC
653.TP
654.B netsplice
ccc2b328 655Like \fBnet\fR, but uses \fBsplice\fR\|(2) and \fBvmsplice\fR\|(2) to map data
de890a1e 656and send/receive. This ioengine defines engine specific options.
d60e92d1 657.TP
53aec0a4 658.B cpuio
d60e92d1
AC
659Doesn't transfer any data, but burns CPU cycles according to \fBcpuload\fR and
660\fBcpucycles\fR parameters.
661.TP
662.B guasi
663The GUASI I/O engine is the Generic Userspace Asynchronous Syscall Interface
cecbfd47 664approach to asynchronous I/O.
d1429b5c
AC
665.br
666See <http://www.xmailserver.org/guasi\-lib.html>.
d60e92d1 667.TP
21b8aee8 668.B rdma
85286c5c
BVA
669The RDMA I/O engine supports both RDMA memory semantics (RDMA_WRITE/RDMA_READ)
670and channel semantics (Send/Recv) for the InfiniBand, RoCE and iWARP protocols.
21b8aee8 671.TP
d60e92d1
AC
672.B external
673Loads an external I/O engine object file. Append the engine filename as
674`:\fIenginepath\fR'.
d54fce84
DM
675.TP
676.B falloc
cecbfd47 677 IO engine that does regular linux native fallocate call to simulate data
d54fce84
DM
678transfer as fio ioengine
679.br
680 DDIR_READ does fallocate(,mode = FALLOC_FL_KEEP_SIZE,)
681.br
0981fd71 682 DIR_WRITE does fallocate(,mode = 0)
d54fce84
DM
683.br
684 DDIR_TRIM does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE)
685.TP
686.B e4defrag
687IO engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate defragment activity
688request to DDIR_WRITE event
0d978694
DAG
689.TP
690.B rbd
ff6bb260
SL
691IO engine supporting direct access to Ceph Rados Block Devices (RBD) via librbd
692without the need to use the kernel rbd driver. This ioengine defines engine specific
0d978694 693options.
a7c386f4 694.TP
695.B gfapi
cc47f094 696Using Glusterfs libgfapi sync interface to direct access to Glusterfs volumes without
697having to go through FUSE. This ioengine defines engine specific
698options.
699.TP
700.B gfapi_async
701Using Glusterfs libgfapi async interface to direct access to Glusterfs volumes without
a7c386f4 702having to go through FUSE. This ioengine defines engine specific
703options.
1b10477b 704.TP
b74e419e
MM
705.B libhdfs
706Read and write through Hadoop (HDFS). The \fBfilename\fR option is used to
707specify host,port of the hdfs name-node to connect. This engine interprets
708offsets a little differently. In HDFS, files once created cannot be modified.
709So random writes are not possible. To imitate this, libhdfs engine expects
710bunch of small files to be created over HDFS, and engine will randomly pick a
711file out of those files based on the offset generated by fio backend. (see the
712example job file to create such files, use rw=write option). Please note, you
713might want to set necessary environment variables to work with hdfs/libhdfs
714properly.
65fa28ca
DE
715.TP
716.B mtd
717Read, write and erase an MTD character device (e.g., /dev/mtd0). Discards are
718treated as erases. Depending on the underlying device type, the I/O may have
719to go in a certain pattern, e.g., on NAND, writing sequentially to erase blocks
720and discarding before overwriting. The writetrim mode works well for this
721constraint.
5c4ef02e
JA
722.TP
723.B pmemblk
724Read and write through the NVML libpmemblk interface.
d60e92d1 725.RE
595e1734 726.P
d60e92d1
AC
727.RE
728.TP
729.BI iodepth \fR=\fPint
8489dae4
SK
730Number of I/O units to keep in flight against the file. Note that increasing
731iodepth beyond 1 will not affect synchronous ioengines (except for small
cf145d90 732degress when verify_async is in use). Even async engines may impose OS
ee72ca09
JA
733restrictions causing the desired depth not to be achieved. This may happen on
734Linux when using libaio and not setting \fBdirect\fR=1, since buffered IO is
735not async on that OS. Keep an eye on the IO depth distribution in the
736fio output to verify that the achieved depth is as expected. Default: 1.
d60e92d1 737.TP
e63a0b2f
RP
738.BI iodepth_batch \fR=\fPint "\fR,\fP iodepth_batch_submit" \fR=\fPint
739This defines how many pieces of IO to submit at once. It defaults to 1
740which means that we submit each IO as soon as it is available, but can
741be raised to submit bigger batches of IO at the time. If it is set to 0
742the \fBiodepth\fR value will be used.
d60e92d1 743.TP
82407585 744.BI iodepth_batch_complete_min \fR=\fPint "\fR,\fP iodepth_batch_complete" \fR=\fPint
3ce9dcaf
JA
745This defines how many pieces of IO to retrieve at once. It defaults to 1 which
746 means that we'll ask for a minimum of 1 IO in the retrieval process from the
747kernel. The IO retrieval will go on until we hit the limit set by
748\fBiodepth_low\fR. If this variable is set to 0, then fio will always check for
749completed events before queuing more IO. This helps reduce IO latency, at the
750cost of more retrieval system calls.
751.TP
82407585
RP
752.BI iodepth_batch_complete_max \fR=\fPint
753This defines maximum pieces of IO to
754retrieve at once. This variable should be used along with
755\fBiodepth_batch_complete_min\fR=int variable, specifying the range
756of min and max amount of IO which should be retrieved. By default
757it is equal to \fBiodepth_batch_complete_min\fR value.
758
759Example #1:
760.RS
761.RS
762\fBiodepth_batch_complete_min\fR=1
763.LP
764\fBiodepth_batch_complete_max\fR=<iodepth>
765.RE
766
767which means that we will retrieve at leat 1 IO and up to the
768whole submitted queue depth. If none of IO has been completed
769yet, we will wait.
770
771Example #2:
772.RS
773\fBiodepth_batch_complete_min\fR=0
774.LP
775\fBiodepth_batch_complete_max\fR=<iodepth>
776.RE
777
778which means that we can retrieve up to the whole submitted
779queue depth, but if none of IO has been completed yet, we will
780NOT wait and immediately exit the system call. In this example
781we simply do polling.
782.RE
783.TP
d60e92d1
AC
784.BI iodepth_low \fR=\fPint
785Low watermark indicating when to start filling the queue again. Default:
ff6bb260 786\fBiodepth\fR.
d60e92d1 787.TP
1ad01bd1
JA
788.BI io_submit_mode \fR=\fPstr
789This option controls how fio submits the IO to the IO engine. The default is
790\fBinline\fR, which means that the fio job threads submit and reap IO directly.
791If set to \fBoffload\fR, the job threads will offload IO submission to a
792dedicated pool of IO threads. This requires some coordination and thus has a
793bit of extra overhead, especially for lower queue depth IO where it can
794increase latencies. The benefit is that fio can manage submission rates
795independently of the device completion rates. This avoids skewed latency
796reporting if IO gets back up on the device side (the coordinated omission
797problem).
798.TP
d60e92d1
AC
799.BI direct \fR=\fPbool
800If true, use non-buffered I/O (usually O_DIRECT). Default: false.
801.TP
d01612f3
CM
802.BI atomic \fR=\fPbool
803If value is true, attempt to use atomic direct IO. Atomic writes are guaranteed
804to be stable once acknowledged by the operating system. Only Linux supports
805O_ATOMIC right now.
806.TP
d60e92d1
AC
807.BI buffered \fR=\fPbool
808If true, use buffered I/O. This is the opposite of the \fBdirect\fR parameter.
809Default: true.
810.TP
f7fa2653 811.BI offset \fR=\fPint
d60e92d1
AC
812Offset in the file to start I/O. Data before the offset will not be touched.
813.TP
591e9e06
JA
814.BI offset_increment \fR=\fPint
815If this is provided, then the real offset becomes the
69bdd6ba
JH
816offset + offset_increment * thread_number, where the thread number is a
817counter that starts at 0 and is incremented for each sub-job (i.e. when
818numjobs option is specified). This option is useful if there are several jobs
819which are intended to operate on a file in parallel disjoint segments, with
820even spacing between the starting points.
591e9e06 821.TP
ddf24e42
JA
822.BI number_ios \fR=\fPint
823Fio will normally perform IOs until it has exhausted the size of the region
824set by \fBsize\fR, or if it exhaust the allocated time (or hits an error
825condition). With this setting, the range/size can be set independently of
826the number of IOs to perform. When fio reaches this number, it will exit
be3fec7d
JA
827normally and report status. Note that this does not extend the amount
828of IO that will be done, it will only stop fio if this condition is met
829before other end-of-job criteria.
ddf24e42 830.TP
d60e92d1 831.BI fsync \fR=\fPint
d1429b5c
AC
832How many I/Os to perform before issuing an \fBfsync\fR\|(2) of dirty data. If
8330, don't sync. Default: 0.
d60e92d1 834.TP
5f9099ea
JA
835.BI fdatasync \fR=\fPint
836Like \fBfsync\fR, but uses \fBfdatasync\fR\|(2) instead to only sync the
837data parts of the file. Default: 0.
838.TP
fa769d44
SW
839.BI write_barrier \fR=\fPint
840Make every Nth write a barrier write.
841.TP
e76b1da4 842.BI sync_file_range \fR=\fPstr:int
ccc2b328
SW
843Use \fBsync_file_range\fR\|(2) for every \fRval\fP number of write operations. Fio will
844track range of writes that have happened since the last \fBsync_file_range\fR\|(2) call.
e76b1da4
JA
845\fRstr\fP can currently be one or more of:
846.RS
847.TP
848.B wait_before
849SYNC_FILE_RANGE_WAIT_BEFORE
850.TP
851.B write
852SYNC_FILE_RANGE_WRITE
853.TP
854.B wait_after
855SYNC_FILE_RANGE_WRITE
856.TP
857.RE
858.P
859So if you do sync_file_range=wait_before,write:8, fio would use
860\fBSYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE\fP for every 8 writes.
ccc2b328 861Also see the \fBsync_file_range\fR\|(2) man page. This option is Linux specific.
e76b1da4 862.TP
d60e92d1 863.BI overwrite \fR=\fPbool
d1429b5c 864If writing, setup the file first and do overwrites. Default: false.
d60e92d1
AC
865.TP
866.BI end_fsync \fR=\fPbool
dbd11ead 867Sync file contents when a write stage has completed. Default: false.
d60e92d1
AC
868.TP
869.BI fsync_on_close \fR=\fPbool
870If true, sync file contents on close. This differs from \fBend_fsync\fR in that
d1429b5c 871it will happen on every close, not just at the end of the job. Default: false.
d60e92d1 872.TP
d60e92d1
AC
873.BI rwmixread \fR=\fPint
874Percentage of a mixed workload that should be reads. Default: 50.
875.TP
876.BI rwmixwrite \fR=\fPint
d1429b5c 877Percentage of a mixed workload that should be writes. If \fBrwmixread\fR and
c35dd7a6
JA
878\fBrwmixwrite\fR are given and do not sum to 100%, the latter of the two
879overrides the first. This may interfere with a given rate setting, if fio is
880asked to limit reads or writes to a certain rate. If that is the case, then
881the distribution may be skewed. Default: 50.
d60e92d1 882.TP
92d42d69
JA
883.BI random_distribution \fR=\fPstr:float
884By default, fio will use a completely uniform random distribution when asked
885to perform random IO. Sometimes it is useful to skew the distribution in
886specific ways, ensuring that some parts of the data is more hot than others.
887Fio includes the following distribution models:
888.RS
889.TP
890.B random
891Uniform random distribution
892.TP
893.B zipf
894Zipf distribution
895.TP
896.B pareto
897Pareto distribution
898.TP
8116fd24
JA
899.B gauss
900Normal (gaussian) distribution
901.TP
e0a04ac1
JA
902.B zoned
903Zoned random distribution
904.TP
92d42d69 905.RE
8116fd24
JA
906When using a \fBzipf\fR or \fBpareto\fR distribution, an input value is also
907needed to define the access pattern. For \fBzipf\fR, this is the zipf theta.
908For \fBpareto\fR, it's the pareto power. Fio includes a test program, genzipf,
909that can be used visualize what the given input values will yield in terms of
910hit rates. If you wanted to use \fBzipf\fR with a theta of 1.2, you would use
92d42d69 911random_distribution=zipf:1.2 as the option. If a non-uniform model is used,
8116fd24
JA
912fio will disable use of the random map. For the \fBgauss\fR distribution, a
913normal deviation is supplied as a value between 0 and 100.
e0a04ac1
JA
914.P
915.RS
916For a \fBzoned\fR distribution, fio supports specifying percentages of IO
917access that should fall within what range of the file or device. For example,
918given a criteria of:
919.P
920.RS
92160% of accesses should be to the first 10%
922.RE
923.RS
92430% of accesses should be to the next 20%
925.RE
926.RS
9278% of accesses should be to to the next 30%
928.RE
929.RS
9302% of accesses should be to the next 40%
931.RE
932.P
933we can define that through zoning of the random accesses. For the above
934example, the user would do:
935.P
936.RS
937.B random_distribution=zoned:60/10:30/20:8/30:2/40
938.RE
939.P
940similarly to how \fBbssplit\fR works for setting ranges and percentages of block
941sizes. Like \fBbssplit\fR, it's possible to specify separate zones for reads,
942writes, and trims. If just one set is given, it'll apply to all of them.
943.RE
92d42d69 944.TP
211c9b89
JA
945.BI percentage_random \fR=\fPint
946For a random workload, set how big a percentage should be random. This defaults
947to 100%, in which case the workload is fully random. It can be set from
948anywhere from 0 to 100. Setting it to 0 would make the workload fully
d9472271
JA
949sequential. It is possible to set different values for reads, writes, and
950trim. To do so, simply use a comma separated list. See \fBblocksize\fR.
211c9b89 951.TP
d60e92d1
AC
952.B norandommap
953Normally \fBfio\fR will cover every block of the file when doing random I/O. If
954this parameter is given, a new offset will be chosen without looking at past
955I/O history. This parameter is mutually exclusive with \fBverify\fR.
956.TP
744492c9 957.BI softrandommap \fR=\fPbool
3ce9dcaf
JA
958See \fBnorandommap\fR. If fio runs with the random block map enabled and it
959fails to allocate the map, if this option is set it will continue without a
960random block map. As coverage will not be as complete as with random maps, this
961option is disabled by default.
962.TP
e8b1961d
JA
963.BI random_generator \fR=\fPstr
964Fio supports the following engines for generating IO offsets for random IO:
965.RS
966.TP
967.B tausworthe
968Strong 2^88 cycle random number generator
969.TP
970.B lfsr
971Linear feedback shift register generator
972.TP
c3546b53
JA
973.B tausworthe64
974Strong 64-bit 2^258 cycle random number generator
975.TP
e8b1961d
JA
976.RE
977.P
978Tausworthe is a strong random number generator, but it requires tracking on the
979side if we want to ensure that blocks are only read or written once. LFSR
980guarantees that we never generate the same offset twice, and it's also less
981computationally expensive. It's not a true random generator, however, though
982for IO purposes it's typically good enough. LFSR only works with single block
983sizes, not with workloads that use multiple block sizes. If used with such a
3bb85e84
JA
984workload, fio may read or write some blocks multiple times. The default
985value is tausworthe, unless the required space exceeds 2^32 blocks. If it does,
986then tausworthe64 is selected automatically.
e8b1961d 987.TP
d60e92d1 988.BI nice \fR=\fPint
ccc2b328 989Run job with given nice value. See \fBnice\fR\|(2).
d60e92d1
AC
990.TP
991.BI prio \fR=\fPint
992Set I/O priority value of this job between 0 (highest) and 7 (lowest). See
ccc2b328 993\fBionice\fR\|(1).
d60e92d1
AC
994.TP
995.BI prioclass \fR=\fPint
ccc2b328 996Set I/O priority class. See \fBionice\fR\|(1).
d60e92d1
AC
997.TP
998.BI thinktime \fR=\fPint
999Stall job for given number of microseconds between issuing I/Os.
1000.TP
1001.BI thinktime_spin \fR=\fPint
1002Pretend to spend CPU time for given number of microseconds, sleeping the rest
1003of the time specified by \fBthinktime\fR. Only valid if \fBthinktime\fR is set.
1004.TP
1005.BI thinktime_blocks \fR=\fPint
4d01ece6
JA
1006Only valid if thinktime is set - control how many blocks to issue, before
1007waiting \fBthinktime\fR microseconds. If not set, defaults to 1 which will
1008make fio wait \fBthinktime\fR microseconds after every block. This
1009effectively makes any queue depth setting redundant, since no more than 1 IO
1010will be queued before we have to complete it and do our thinktime. In other
1011words, this setting effectively caps the queue depth if the latter is larger.
d60e92d1
AC
1012Default: 1.
1013.TP
1014.BI rate \fR=\fPint
c35dd7a6
JA
1015Cap bandwidth used by this job. The number is in bytes/sec, the normal postfix
1016rules apply. You can use \fBrate\fR=500k to limit reads and writes to 500k each,
1017or you can specify read and writes separately. Using \fBrate\fR=1m,500k would
1018limit reads to 1MB/sec and writes to 500KB/sec. Capping only reads or writes
1019can be done with \fBrate\fR=,500k or \fBrate\fR=500k,. The former will only
1020limit writes (to 500KB/sec), the latter will only limit reads.
d60e92d1 1021.TP
6d428bcd 1022.BI rate_min \fR=\fPint
d60e92d1 1023Tell \fBfio\fR to do whatever it can to maintain at least the given bandwidth.
c35dd7a6
JA
1024Failing to meet this requirement will cause the job to exit. The same format
1025as \fBrate\fR is used for read vs write separation.
d60e92d1
AC
1026.TP
1027.BI rate_iops \fR=\fPint
c35dd7a6
JA
1028Cap the bandwidth to this number of IOPS. Basically the same as rate, just
1029specified independently of bandwidth. The same format as \fBrate\fR is used for
de8f6de9 1030read vs write separation. If \fBblocksize\fR is a range, the smallest block
c35dd7a6 1031size is used as the metric.
d60e92d1
AC
1032.TP
1033.BI rate_iops_min \fR=\fPint
c35dd7a6 1034If this rate of I/O is not met, the job will exit. The same format as \fBrate\fR
de8f6de9 1035is used for read vs write separation.
d60e92d1 1036.TP
6de65959
JA
1037.BI rate_process \fR=\fPstr
1038This option controls how fio manages rated IO submissions. The default is
1039\fBlinear\fR, which submits IO in a linear fashion with fixed delays between
1040IOs that gets adjusted based on IO completion rates. If this is set to
1041\fBpoisson\fR, fio will submit IO based on a more real world random request
1042flow, known as the Poisson process
5d02b083
JA
1043(https://en.wikipedia.org/wiki/Poisson_process). The lambda will be
104410^6 / IOPS for the given workload.
ff6bb260 1045.TP
6d428bcd
JA
1046.BI rate_cycle \fR=\fPint
1047Average bandwidth for \fBrate\fR and \fBrate_min\fR over this number of
d60e92d1
AC
1048milliseconds. Default: 1000ms.
1049.TP
3e260a46
JA
1050.BI latency_target \fR=\fPint
1051If set, fio will attempt to find the max performance point that the given
1052workload will run at while maintaining a latency below this target. The
1053values is given in microseconds. See \fBlatency_window\fR and
1054\fBlatency_percentile\fR.
1055.TP
1056.BI latency_window \fR=\fPint
1057Used with \fBlatency_target\fR to specify the sample window that the job
1058is run at varying queue depths to test the performance. The value is given
1059in microseconds.
1060.TP
1061.BI latency_percentile \fR=\fPfloat
1062The percentage of IOs that must fall within the criteria specified by
1063\fBlatency_target\fR and \fBlatency_window\fR. If not set, this defaults
1064to 100.0, meaning that all IOs must be equal or below to the value set
1065by \fBlatency_target\fR.
1066.TP
15501535
JA
1067.BI max_latency \fR=\fPint
1068If set, fio will exit the job if it exceeds this maximum latency. It will exit
1069with an ETIME error.
1070.TP
d60e92d1
AC
1071.BI cpumask \fR=\fPint
1072Set CPU affinity for this job. \fIint\fR is a bitmask of allowed CPUs the job
1073may run on. See \fBsched_setaffinity\fR\|(2).
1074.TP
1075.BI cpus_allowed \fR=\fPstr
1076Same as \fBcpumask\fR, but allows a comma-delimited list of CPU numbers.
1077.TP
c2acfbac
JA
1078.BI cpus_allowed_policy \fR=\fPstr
1079Set the policy of how fio distributes the CPUs specified by \fBcpus_allowed\fR
1080or \fBcpumask\fR. Two policies are supported:
1081.RS
1082.RS
1083.TP
1084.B shared
1085All jobs will share the CPU set specified.
1086.TP
1087.B split
1088Each job will get a unique CPU from the CPU set.
1089.RE
1090.P
1091\fBshared\fR is the default behaviour, if the option isn't specified. If
ada083cd
JA
1092\fBsplit\fR is specified, then fio will assign one cpu per job. If not enough
1093CPUs are given for the jobs listed, then fio will roundrobin the CPUs in
1094the set.
c2acfbac
JA
1095.RE
1096.P
1097.TP
d0b937ed 1098.BI numa_cpu_nodes \fR=\fPstr
cecbfd47 1099Set this job running on specified NUMA nodes' CPUs. The arguments allow
d0b937ed
YR
1100comma delimited list of cpu numbers, A-B ranges, or 'all'.
1101.TP
1102.BI numa_mem_policy \fR=\fPstr
1103Set this job's memory policy and corresponding NUMA nodes. Format of
cecbfd47 1104the arguments:
d0b937ed
YR
1105.RS
1106.TP
1107.B <mode>[:<nodelist>]
1108.TP
1109.B mode
1110is one of the following memory policy:
1111.TP
1112.B default, prefer, bind, interleave, local
1113.TP
1114.RE
1115For \fBdefault\fR and \fBlocal\fR memory policy, no \fBnodelist\fR is
1116needed to be specified. For \fBprefer\fR, only one node is
1117allowed. For \fBbind\fR and \fBinterleave\fR, \fBnodelist\fR allows
1118comma delimited list of numbers, A-B ranges, or 'all'.
1119.TP
23ed19b0
CE
1120.BI startdelay \fR=\fPirange
1121Delay start of job for the specified number of seconds. Supports all time
1122suffixes to allow specification of hours, minutes, seconds and
bd66aa2c 1123milliseconds - seconds are the default if a unit is omitted.
23ed19b0
CE
1124Can be given as a range which causes each thread to choose randomly out of the
1125range.
d60e92d1
AC
1126.TP
1127.BI runtime \fR=\fPint
1128Terminate processing after the specified number of seconds.
1129.TP
1130.B time_based
1131If given, run for the specified \fBruntime\fR duration even if the files are
1132completely read or written. The same workload will be repeated as many times
1133as \fBruntime\fR allows.
1134.TP
901bb994
JA
1135.BI ramp_time \fR=\fPint
1136If set, fio will run the specified workload for this amount of time before
1137logging any performance numbers. Useful for letting performance settle before
1138logging results, thus minimizing the runtime required for stable results. Note
c35dd7a6
JA
1139that the \fBramp_time\fR is considered lead in time for a job, thus it will
1140increase the total runtime if a special timeout or runtime is specified.
901bb994 1141.TP
d60e92d1
AC
1142.BI invalidate \fR=\fPbool
1143Invalidate buffer-cache for the file prior to starting I/O. Default: true.
1144.TP
1145.BI sync \fR=\fPbool
1146Use synchronous I/O for buffered writes. For the majority of I/O engines,
d1429b5c 1147this means using O_SYNC. Default: false.
d60e92d1
AC
1148.TP
1149.BI iomem \fR=\fPstr "\fR,\fP mem" \fR=\fPstr
1150Allocation method for I/O unit buffer. Allowed values are:
1151.RS
1152.RS
1153.TP
1154.B malloc
ccc2b328 1155Allocate memory with \fBmalloc\fR\|(3).
d60e92d1
AC
1156.TP
1157.B shm
ccc2b328 1158Use shared memory buffers allocated through \fBshmget\fR\|(2).
d60e92d1
AC
1159.TP
1160.B shmhuge
1161Same as \fBshm\fR, but use huge pages as backing.
1162.TP
1163.B mmap
ccc2b328 1164Use \fBmmap\fR\|(2) for allocation. Uses anonymous memory unless a filename
d60e92d1
AC
1165is given after the option in the format `:\fIfile\fR'.
1166.TP
1167.B mmaphuge
1168Same as \fBmmap\fR, but use huge files as backing.
09c782bb
JA
1169.TP
1170.B mmapshared
1171Same as \fBmmap\fR, but use a MMAP_SHARED mapping.
d60e92d1
AC
1172.RE
1173.P
1174The amount of memory allocated is the maximum allowed \fBblocksize\fR for the
1175job multiplied by \fBiodepth\fR. For \fBshmhuge\fR or \fBmmaphuge\fR to work,
1176the system must have free huge pages allocated. \fBmmaphuge\fR also needs to
2e266ba6
JA
1177have hugetlbfs mounted, and \fIfile\fR must point there. At least on Linux,
1178huge pages must be manually allocated. See \fB/proc/sys/vm/nr_hugehages\fR
1179and the documentation for that. Normally you just need to echo an appropriate
1180number, eg echoing 8 will ensure that the OS has 8 huge pages ready for
1181use.
d60e92d1
AC
1182.RE
1183.TP
d392365e 1184.BI iomem_align \fR=\fPint "\fR,\fP mem_align" \fR=\fPint
cecbfd47 1185This indicates the memory alignment of the IO memory buffers. Note that the
d529ee19
JA
1186given alignment is applied to the first IO unit buffer, if using \fBiodepth\fR
1187the alignment of the following buffers are given by the \fBbs\fR used. In
1188other words, if using a \fBbs\fR that is a multiple of the page sized in the
1189system, all buffers will be aligned to this value. If using a \fBbs\fR that
1190is not page aligned, the alignment of subsequent IO memory buffers is the
1191sum of the \fBiomem_align\fR and \fBbs\fR used.
1192.TP
f7fa2653 1193.BI hugepage\-size \fR=\fPint
d60e92d1 1194Defines the size of a huge page. Must be at least equal to the system setting.
b22989b9 1195Should be a multiple of 1MB. Default: 4MB.
d60e92d1
AC
1196.TP
1197.B exitall
1198Terminate all jobs when one finishes. Default: wait for each job to finish.
1199.TP
f9cafb12
JA
1200.B exitall_on_error \fR=\fPbool
1201Terminate all jobs if one job finishes in error. Default: wait for each job
1202to finish.
1203.TP
d60e92d1 1204.BI bwavgtime \fR=\fPint
a47591e4
JA
1205Average bandwidth calculations over the given time in milliseconds. If the job
1206also does bandwidth logging through \fBwrite_bw_log\fR, then the minimum of
1207this option and \fBlog_avg_msec\fR will be used. Default: 500ms.
d60e92d1 1208.TP
c8eeb9df 1209.BI iopsavgtime \fR=\fPint
a47591e4
JA
1210Average IOPS calculations over the given time in milliseconds. If the job
1211also does IOPS logging through \fBwrite_iops_log\fR, then the minimum of
1212this option and \fBlog_avg_msec\fR will be used. Default: 500ms.
c8eeb9df 1213.TP
d60e92d1 1214.BI create_serialize \fR=\fPbool
d1429b5c 1215If true, serialize file creation for the jobs. Default: true.
d60e92d1
AC
1216.TP
1217.BI create_fsync \fR=\fPbool
ccc2b328 1218\fBfsync\fR\|(2) data file after creation. Default: true.
d60e92d1 1219.TP
6b7f6851
JA
1220.BI create_on_open \fR=\fPbool
1221If true, the files are not created until they are opened for IO by the job.
1222.TP
25460cf6
JA
1223.BI create_only \fR=\fPbool
1224If true, fio will only run the setup phase of the job. If files need to be
1225laid out or updated on disk, only that will be done. The actual job contents
1226are not executed.
1227.TP
2378826d
JA
1228.BI allow_file_create \fR=\fPbool
1229If true, fio is permitted to create files as part of its workload. This is
1230the default behavior. If this option is false, then fio will error out if the
1231files it needs to use don't already exist. Default: true.
1232.TP
e81ecca3
JA
1233.BI allow_mounted_write \fR=\fPbool
1234If this isn't set, fio will abort jobs that are destructive (eg that write)
1235to what appears to be a mounted device or partition. This should help catch
1236creating inadvertently destructive tests, not realizing that the test will
1237destroy data on the mounted file system. Default: false.
1238.TP
e9f48479
JA
1239.BI pre_read \fR=\fPbool
1240If this is given, files will be pre-read into memory before starting the given
1241IO operation. This will also clear the \fR \fBinvalidate\fR flag, since it is
9c0d2241
JA
1242pointless to pre-read and then drop the cache. This will only work for IO
1243engines that are seekable, since they allow you to read the same data
1244multiple times. Thus it will not work on eg network or splice IO.
e9f48479 1245.TP
d60e92d1
AC
1246.BI unlink \fR=\fPbool
1247Unlink job files when done. Default: false.
1248.TP
1249.BI loops \fR=\fPint
1250Specifies the number of iterations (runs of the same workload) of this job.
1251Default: 1.
1252.TP
5e4c7118
JA
1253.BI verify_only \fR=\fPbool
1254Do not perform the specified workload, only verify data still matches previous
1255invocation of this workload. This option allows one to check data multiple
1256times at a later date without overwriting it. This option makes sense only for
1257workloads that write data, and does not support workloads with the
1258\fBtime_based\fR option set.
1259.TP
d60e92d1
AC
1260.BI do_verify \fR=\fPbool
1261Run the verify phase after a write phase. Only valid if \fBverify\fR is set.
1262Default: true.
1263.TP
1264.BI verify \fR=\fPstr
b638d82f
RP
1265Method of verifying file contents after each iteration of the job. Each
1266verification method also implies verification of special header, which is
1267written to the beginning of each block. This header also includes meta
1268information, like offset of the block, block number, timestamp when block
1269was written, etc. \fBverify\fR=str can be combined with \fBverify_pattern\fR=str
1270option. The allowed values are:
d60e92d1
AC
1271.RS
1272.RS
1273.TP
844ea602 1274.B md5 crc16 crc32 crc32c crc32c-intel crc64 crc7 sha256 sha512 sha1 xxhash
0539d758
JA
1275Store appropriate checksum in the header of each block. crc32c-intel is
1276hardware accelerated SSE4.2 driven, falls back to regular crc32c if
1277not supported by the system.
d60e92d1
AC
1278.TP
1279.B meta
b638d82f
RP
1280This option is deprecated, since now meta information is included in generic
1281verification header and meta verification happens by default. For detailed
1282information see the description of the \fBverify\fR=str setting. This option
1283is kept because of compatibility's sake with old configurations. Do not use it.
d60e92d1 1284.TP
59245381
JA
1285.B pattern
1286Verify a strict pattern. Normally fio includes a header with some basic
1287information and checksumming, but if this option is set, only the
1288specific pattern set with \fBverify_pattern\fR is verified.
1289.TP
d60e92d1
AC
1290.B null
1291Pretend to verify. Used for testing internals.
1292.RE
b892dc08
JA
1293
1294This option can be used for repeated burn-in tests of a system to make sure
1295that the written data is also correctly read back. If the data direction given
1296is a read or random read, fio will assume that it should verify a previously
1297written file. If the data direction includes any form of write, the verify will
1298be of the newly written data.
d60e92d1
AC
1299.RE
1300.TP
5c9323fb 1301.BI verifysort \fR=\fPbool
d60e92d1
AC
1302If true, written verify blocks are sorted if \fBfio\fR deems it to be faster to
1303read them back in a sorted manner. Default: true.
1304.TP
fa769d44
SW
1305.BI verifysort_nr \fR=\fPint
1306Pre-load and sort verify blocks for a read workload.
1307.TP
f7fa2653 1308.BI verify_offset \fR=\fPint
d60e92d1 1309Swap the verification header with data somewhere else in the block before
d1429b5c 1310writing. It is swapped back before verifying.
d60e92d1 1311.TP
f7fa2653 1312.BI verify_interval \fR=\fPint
d60e92d1
AC
1313Write the verification header for this number of bytes, which should divide
1314\fBblocksize\fR. Default: \fBblocksize\fR.
1315.TP
996093bb
JA
1316.BI verify_pattern \fR=\fPstr
1317If set, fio will fill the io buffers with this pattern. Fio defaults to filling
1318with totally random bytes, but sometimes it's interesting to fill with a known
1319pattern for io verification purposes. Depending on the width of the pattern,
1320fio will fill 1/2/3/4 bytes of the buffer at the time(it can be either a
1321decimal or a hex number). The verify_pattern if larger than a 32-bit quantity
1322has to be a hex number that starts with either "0x" or "0X". Use with
b638d82f
RP
1323\fBverify\fP=str. Also, verify_pattern supports %o format, which means that for
1324each block offset will be written and then verifyied back, e.g.:
2fa5a241
RP
1325.RS
1326.RS
1327\fBverify_pattern\fR=%o
1328.RE
1329Or use combination of everything:
1330.LP
1331.RS
1332\fBverify_pattern\fR=0xff%o"abcd"-21
1333.RE
1334.RE
996093bb 1335.TP
d60e92d1
AC
1336.BI verify_fatal \fR=\fPbool
1337If true, exit the job on the first observed verification failure. Default:
1338false.
1339.TP
b463e936
JA
1340.BI verify_dump \fR=\fPbool
1341If set, dump the contents of both the original data block and the data block we
1342read off disk to files. This allows later analysis to inspect just what kind of
ef71e317 1343data corruption occurred. Off by default.
b463e936 1344.TP
e8462bd8
JA
1345.BI verify_async \fR=\fPint
1346Fio will normally verify IO inline from the submitting thread. This option
1347takes an integer describing how many async offload threads to create for IO
1348verification instead, causing fio to offload the duty of verifying IO contents
c85c324c
JA
1349to one or more separate threads. If using this offload option, even sync IO
1350engines can benefit from using an \fBiodepth\fR setting higher than 1, as it
1351allows them to have IO in flight while verifies are running.
e8462bd8
JA
1352.TP
1353.BI verify_async_cpus \fR=\fPstr
1354Tell fio to set the given CPU affinity on the async IO verification threads.
1355See \fBcpus_allowed\fP for the format used.
1356.TP
6f87418f
JA
1357.BI verify_backlog \fR=\fPint
1358Fio will normally verify the written contents of a job that utilizes verify
1359once that job has completed. In other words, everything is written then
1360everything is read back and verified. You may want to verify continually
1361instead for a variety of reasons. Fio stores the meta data associated with an
1362IO block in memory, so for large verify workloads, quite a bit of memory would
092f707f
DN
1363be used up holding this meta data. If this option is enabled, fio will write
1364only N blocks before verifying these blocks.
6f87418f
JA
1365.TP
1366.BI verify_backlog_batch \fR=\fPint
1367Control how many blocks fio will verify if verify_backlog is set. If not set,
1368will default to the value of \fBverify_backlog\fR (meaning the entire queue is
ff6bb260
SL
1369read back and verified). If \fBverify_backlog_batch\fR is less than
1370\fBverify_backlog\fR then not all blocks will be verified, if
092f707f
DN
1371\fBverify_backlog_batch\fR is larger than \fBverify_backlog\fR, some blocks
1372will be verified more than once.
6f87418f 1373.TP
fa769d44
SW
1374.BI trim_percentage \fR=\fPint
1375Number of verify blocks to discard/trim.
1376.TP
1377.BI trim_verify_zero \fR=\fPbool
1378Verify that trim/discarded blocks are returned as zeroes.
1379.TP
1380.BI trim_backlog \fR=\fPint
1381Trim after this number of blocks are written.
1382.TP
1383.BI trim_backlog_batch \fR=\fPint
1384Trim this number of IO blocks.
1385.TP
1386.BI experimental_verify \fR=\fPbool
1387Enable experimental verification.
1388.TP
ca09be4b
JA
1389.BI verify_state_save \fR=\fPbool
1390When a job exits during the write phase of a verify workload, save its
1391current state. This allows fio to replay up until that point, if the
1392verify state is loaded for the verify read phase.
1393.TP
1394.BI verify_state_load \fR=\fPbool
1395If a verify termination trigger was used, fio stores the current write
1396state of each thread. This can be used at verification time so that fio
1397knows how far it should verify. Without this information, fio will run
1398a full verification pass, according to the settings in the job file used.
1399.TP
d392365e 1400.B stonewall "\fR,\fP wait_for_previous"
5982a925 1401Wait for preceding jobs in the job file to exit before starting this one.
d60e92d1
AC
1402\fBstonewall\fR implies \fBnew_group\fR.
1403.TP
1404.B new_group
1405Start a new reporting group. If not given, all jobs in a file will be part
1406of the same reporting group, unless separated by a stonewall.
1407.TP
1408.BI numjobs \fR=\fPint
ff6bb260 1409Number of clones (processes/threads performing the same workload) of this job.
d60e92d1
AC
1410Default: 1.
1411.TP
1412.B group_reporting
1413If set, display per-group reports instead of per-job when \fBnumjobs\fR is
1414specified.
1415.TP
1416.B thread
1417Use threads created with \fBpthread_create\fR\|(3) instead of processes created
1418with \fBfork\fR\|(2).
1419.TP
f7fa2653 1420.BI zonesize \fR=\fPint
d60e92d1
AC
1421Divide file into zones of the specified size in bytes. See \fBzoneskip\fR.
1422.TP
fa769d44
SW
1423.BI zonerange \fR=\fPint
1424Give size of an IO zone. See \fBzoneskip\fR.
1425.TP
f7fa2653 1426.BI zoneskip \fR=\fPint
d1429b5c 1427Skip the specified number of bytes when \fBzonesize\fR bytes of data have been
d60e92d1
AC
1428read.
1429.TP
1430.BI write_iolog \fR=\fPstr
5b42a488
SH
1431Write the issued I/O patterns to the specified file. Specify a separate file
1432for each job, otherwise the iologs will be interspersed and the file may be
1433corrupt.
d60e92d1
AC
1434.TP
1435.BI read_iolog \fR=\fPstr
1436Replay the I/O patterns contained in the specified file generated by
1437\fBwrite_iolog\fR, or may be a \fBblktrace\fR binary file.
1438.TP
64bbb865
DN
1439.BI replay_no_stall \fR=\fPint
1440While replaying I/O patterns using \fBread_iolog\fR the default behavior
1441attempts to respect timing information between I/Os. Enabling
1442\fBreplay_no_stall\fR causes I/Os to be replayed as fast as possible while
1443still respecting ordering.
1444.TP
d1c46c04
DN
1445.BI replay_redirect \fR=\fPstr
1446While replaying I/O patterns using \fBread_iolog\fR the default behavior
1447is to replay the IOPS onto the major/minor device that each IOP was recorded
1448from. Setting \fBreplay_redirect\fR causes all IOPS to be replayed onto the
1449single specified device regardless of the device it was recorded from.
1450.TP
0c63576e
JA
1451.BI replay_align \fR=\fPint
1452Force alignment of IO offsets and lengths in a trace to this power of 2 value.
1453.TP
1454.BI replay_scale \fR=\fPint
1455Scale sector offsets down by this factor when replaying traces.
1456.TP
3a5db920
JA
1457.BI per_job_logs \fR=\fPbool
1458If set, this generates bw/clat/iops log with per file private filenames. If
1459not set, jobs with identical names will share the log filename. Default: true.
1460.TP
836bad52 1461.BI write_bw_log \fR=\fPstr
901bb994
JA
1462If given, write a bandwidth log of the jobs in this job file. Can be used to
1463store data of the bandwidth of the jobs in their lifetime. The included
1464fio_generate_plots script uses gnuplot to turn these text files into nice
26b26fca 1465graphs. See \fBwrite_lat_log\fR for behaviour of given filename. For this
8ad3b3dd 1466option, the postfix is _bw.x.log, where x is the index of the job (1..N,
3a5db920 1467where N is the number of jobs). If \fBper_job_logs\fR is false, then the
a3ae5b05
JA
1468filename will not include the job index. See the \fBLOG FILE FORMATS\fR
1469section.
d60e92d1 1470.TP
836bad52 1471.BI write_lat_log \fR=\fPstr
901bb994 1472Same as \fBwrite_bw_log\fR, but writes I/O completion latencies. If no
8ad3b3dd
JA
1473filename is given with this option, the default filename of
1474"jobname_type.x.log" is used, where x is the index of the job (1..N, where
1475N is the number of jobs). Even if the filename is given, fio will still
3a5db920 1476append the type of log. If \fBper_job_logs\fR is false, then the filename will
a3ae5b05 1477not include the job index. See the \fBLOG FILE FORMATS\fR section.
901bb994 1478.TP
c8eeb9df
JA
1479.BI write_iops_log \fR=\fPstr
1480Same as \fBwrite_bw_log\fR, but writes IOPS. If no filename is given with this
8ad3b3dd
JA
1481option, the default filename of "jobname_type.x.log" is used, where x is the
1482index of the job (1..N, where N is the number of jobs). Even if the filename
3a5db920 1483is given, fio will still append the type of log. If \fBper_job_logs\fR is false,
a3ae5b05
JA
1484then the filename will not include the job index. See the \fBLOG FILE FORMATS\fR
1485section.
c8eeb9df 1486.TP
b8bc8cba
JA
1487.BI log_avg_msec \fR=\fPint
1488By default, fio will log an entry in the iops, latency, or bw log for every
1489IO that completes. When writing to the disk log, that can quickly grow to a
1490very large size. Setting this option makes fio average the each log entry
e6989e10 1491over the specified period of time, reducing the resolution of the log. See
4b1ddb7a 1492\fBlog_max_value\fR as well. Defaults to 0, logging all entries.
e6989e10 1493.TP
4b1ddb7a 1494.BI log_max_value \fR=\fPbool
e6989e10
JA
1495If \fBlog_avg_msec\fR is set, fio logs the average over that window. If you
1496instead want to log the maximum value, set this option to 1. Defaults to
14970, meaning that averaged values are logged.
b8bc8cba 1498.TP
ae588852
JA
1499.BI log_offset \fR=\fPbool
1500If this is set, the iolog options will include the byte offset for the IO
1501entry as well as the other data values.
1502.TP
aee2ab67
JA
1503.BI log_compression \fR=\fPint
1504If this is set, fio will compress the IO logs as it goes, to keep the memory
1505footprint lower. When a log reaches the specified size, that chunk is removed
1506and compressed in the background. Given that IO logs are fairly highly
1507compressible, this yields a nice memory savings for longer runs. The downside
1508is that the compression will consume some background CPU cycles, so it may
1509impact the run. This, however, is also true if the logging ends up consuming
1510most of the system memory. So pick your poison. The IO logs are saved
1511normally at the end of a run, by decompressing the chunks and storing them
1512in the specified log file. This feature depends on the availability of zlib.
1513.TP
c08f9fe2
JA
1514.BI log_compression_cpus \fR=\fPstr
1515Define the set of CPUs that are allowed to handle online log compression
1516for the IO jobs. This can provide better isolation between performance
1517sensitive jobs, and background compression work.
1518.TP
b26317c9 1519.BI log_store_compressed \fR=\fPbool
c08f9fe2
JA
1520If set, fio will store the log files in a compressed format. They can be
1521decompressed with fio, using the \fB\-\-inflate-log\fR command line parameter.
1522The files will be stored with a \fB\.fz\fR suffix.
b26317c9 1523.TP
66347cfa
DE
1524.BI block_error_percentiles \fR=\fPbool
1525If set, record errors in trim block-sized units from writes and trims and output
1526a histogram of how many trims it took to get to errors, and what kind of error
1527was encountered.
1528.TP
836bad52 1529.BI disable_lat \fR=\fPbool
02af0988 1530Disable measurements of total latency numbers. Useful only for cutting
ccc2b328 1531back the number of calls to \fBgettimeofday\fR\|(2), as that does impact performance at
901bb994
JA
1532really high IOPS rates. Note that to really get rid of a large amount of these
1533calls, this option must be used with disable_slat and disable_bw as well.
1534.TP
836bad52 1535.BI disable_clat \fR=\fPbool
c95f9daf 1536Disable measurements of completion latency numbers. See \fBdisable_lat\fR.
02af0988 1537.TP
836bad52 1538.BI disable_slat \fR=\fPbool
02af0988 1539Disable measurements of submission latency numbers. See \fBdisable_lat\fR.
901bb994 1540.TP
836bad52 1541.BI disable_bw_measurement \fR=\fPbool
02af0988 1542Disable measurements of throughput/bandwidth numbers. See \fBdisable_lat\fR.
d60e92d1 1543.TP
f7fa2653 1544.BI lockmem \fR=\fPint
d60e92d1 1545Pin the specified amount of memory with \fBmlock\fR\|(2). Can be used to
81c6b6cd 1546simulate a smaller amount of memory. The amount specified is per worker.
d60e92d1
AC
1547.TP
1548.BI exec_prerun \fR=\fPstr
1549Before running the job, execute the specified command with \fBsystem\fR\|(3).
ce486495
EV
1550.RS
1551Output is redirected in a file called \fBjobname.prerun.txt\fR
1552.RE
d60e92d1
AC
1553.TP
1554.BI exec_postrun \fR=\fPstr
1555Same as \fBexec_prerun\fR, but the command is executed after the job completes.
ce486495
EV
1556.RS
1557Output is redirected in a file called \fBjobname.postrun.txt\fR
1558.RE
d60e92d1
AC
1559.TP
1560.BI ioscheduler \fR=\fPstr
1561Attempt to switch the device hosting the file to the specified I/O scheduler.
1562.TP
d60e92d1 1563.BI disk_util \fR=\fPbool
d1429b5c 1564Generate disk utilization statistics if the platform supports it. Default: true.
901bb994 1565.TP
23893646
JA
1566.BI clocksource \fR=\fPstr
1567Use the given clocksource as the base of timing. The supported options are:
1568.RS
1569.TP
1570.B gettimeofday
ccc2b328 1571\fBgettimeofday\fR\|(2)
23893646
JA
1572.TP
1573.B clock_gettime
ccc2b328 1574\fBclock_gettime\fR\|(2)
23893646
JA
1575.TP
1576.B cpu
1577Internal CPU clock source
1578.TP
1579.RE
1580.P
1581\fBcpu\fR is the preferred clocksource if it is reliable, as it is very fast
1582(and fio is heavy on time calls). Fio will automatically use this clocksource
1583if it's supported and considered reliable on the system it is running on,
1584unless another clocksource is specifically set. For x86/x86-64 CPUs, this
1585means supporting TSC Invariant.
1586.TP
901bb994 1587.BI gtod_reduce \fR=\fPbool
ccc2b328 1588Enable all of the \fBgettimeofday\fR\|(2) reducing options (disable_clat, disable_slat,
901bb994 1589disable_bw) plus reduce precision of the timeout somewhat to really shrink the
ccc2b328 1590\fBgettimeofday\fR\|(2) call count. With this option enabled, we only do about 0.4% of
901bb994
JA
1591the gtod() calls we would have done if all time keeping was enabled.
1592.TP
1593.BI gtod_cpu \fR=\fPint
1594Sometimes it's cheaper to dedicate a single thread of execution to just getting
1595the current time. Fio (and databases, for instance) are very intensive on
ccc2b328 1596\fBgettimeofday\fR\|(2) calls. With this option, you can set one CPU aside for doing
901bb994
JA
1597nothing but logging current time to a shared memory location. Then the other
1598threads/processes that run IO workloads need only copy that segment, instead of
ccc2b328 1599entering the kernel with a \fBgettimeofday\fR\|(2) call. The CPU set aside for doing
901bb994
JA
1600these time calls will be excluded from other uses. Fio will manually clear it
1601from the CPU mask of other jobs.
f2bba182 1602.TP
8b28bd41
DM
1603.BI ignore_error \fR=\fPstr
1604Sometimes you want to ignore some errors during test in that case you can specify
1605error list for each error type.
1606.br
1607ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST
1608.br
1609errors for given error type is separated with ':'.
1610Error may be symbol ('ENOSPC', 'ENOMEM') or an integer.
1611.br
1612Example: ignore_error=EAGAIN,ENOSPC:122 .
ff6bb260
SL
1613.br
1614This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from WRITE.
8b28bd41
DM
1615.TP
1616.BI error_dump \fR=\fPbool
1617If set dump every error even if it is non fatal, true by default. If disabled
1618only fatal error will be dumped
1619.TP
fa769d44
SW
1620.BI profile \fR=\fPstr
1621Select a specific builtin performance test.
1622.TP
a696fa2a
JA
1623.BI cgroup \fR=\fPstr
1624Add job to this control group. If it doesn't exist, it will be created.
6adb38a1
JA
1625The system must have a mounted cgroup blkio mount point for this to work. If
1626your system doesn't have it mounted, you can do so with:
1627
5982a925 1628# mount \-t cgroup \-o blkio none /cgroup
a696fa2a
JA
1629.TP
1630.BI cgroup_weight \fR=\fPint
1631Set the weight of the cgroup to this value. See the documentation that comes
1632with the kernel, allowed values are in the range of 100..1000.
e0b0d892 1633.TP
7de87099
VG
1634.BI cgroup_nodelete \fR=\fPbool
1635Normally fio will delete the cgroups it has created after the job completion.
1636To override this behavior and to leave cgroups around after the job completion,
1637set cgroup_nodelete=1. This can be useful if one wants to inspect various
1638cgroup files after job completion. Default: false
1639.TP
e0b0d892
JA
1640.BI uid \fR=\fPint
1641Instead of running as the invoking user, set the user ID to this value before
1642the thread/process does any work.
1643.TP
1644.BI gid \fR=\fPint
1645Set group ID, see \fBuid\fR.
83349190 1646.TP
fa769d44
SW
1647.BI unit_base \fR=\fPint
1648Base unit for reporting. Allowed values are:
1649.RS
1650.TP
1651.B 0
1652Use auto-detection (default).
1653.TP
1654.B 8
1655Byte based.
1656.TP
1657.B 1
1658Bit based.
1659.RE
1660.P
1661.TP
9e684a49
DE
1662.BI flow_id \fR=\fPint
1663The ID of the flow. If not specified, it defaults to being a global flow. See
1664\fBflow\fR.
1665.TP
1666.BI flow \fR=\fPint
1667Weight in token-based flow control. If this value is used, then there is a
1668\fBflow counter\fR which is used to regulate the proportion of activity between
1669two or more jobs. fio attempts to keep this flow counter near zero. The
1670\fBflow\fR parameter stands for how much should be added or subtracted to the
1671flow counter on each iteration of the main I/O loop. That is, if one job has
1672\fBflow=8\fR and another job has \fBflow=-1\fR, then there will be a roughly
16731:8 ratio in how much one runs vs the other.
1674.TP
1675.BI flow_watermark \fR=\fPint
1676The maximum value that the absolute value of the flow counter is allowed to
1677reach before the job must wait for a lower value of the counter.
1678.TP
1679.BI flow_sleep \fR=\fPint
1680The period of time, in microseconds, to wait after the flow watermark has been
1681exceeded before retrying operations
1682.TP
83349190
YH
1683.BI clat_percentiles \fR=\fPbool
1684Enable the reporting of percentiles of completion latencies.
1685.TP
1686.BI percentile_list \fR=\fPfloat_list
66347cfa
DE
1687Overwrite the default list of percentiles for completion latencies and the
1688block error histogram. Each number is a floating number in the range (0,100],
1689and the maximum length of the list is 20. Use ':' to separate the
3eb07285 1690numbers. For example, \-\-percentile_list=99.5:99.9 will cause fio to
83349190
YH
1691report the values of completion latency below which 99.5% and 99.9% of
1692the observed latencies fell, respectively.
de890a1e
SL
1693.SS "Ioengine Parameters List"
1694Some parameters are only valid when a specific ioengine is in use. These are
1695used identically to normal parameters, with the caveat that when used on the
cf145d90 1696command line, they must come after the ioengine.
de890a1e 1697.TP
e4585935
JA
1698.BI (cpu)cpuload \fR=\fPint
1699Attempt to use the specified percentage of CPU cycles.
1700.TP
1701.BI (cpu)cpuchunks \fR=\fPint
1702Split the load into cycles of the given time. In microseconds.
1703.TP
046395d7
JA
1704.BI (cpu)exit_on_io_done \fR=\fPbool
1705Detect when IO threads are done, then exit.
1706.TP
de890a1e
SL
1707.BI (libaio)userspace_reap
1708Normally, with the libaio engine in use, fio will use
1709the io_getevents system call to reap newly returned events.
1710With this flag turned on, the AIO ring will be read directly
1711from user-space to reap events. The reaping mode is only
1712enabled when polling for a minimum of 0 events (eg when
1713iodepth_batch_complete=0).
1714.TP
2cafffbe
JA
1715.BI (psyncv2)hipri
1716Set RWF_HIPRI on IO, indicating to the kernel that it's of
1717higher priority than normal.
1718.TP
de890a1e
SL
1719.BI (net,netsplice)hostname \fR=\fPstr
1720The host name or IP address to use for TCP or UDP based IO.
1721If the job is a TCP listener or UDP reader, the hostname is not
b511c9aa 1722used and must be omitted unless it is a valid UDP multicast address.
de890a1e
SL
1723.TP
1724.BI (net,netsplice)port \fR=\fPint
6315af9d
JA
1725The TCP or UDP port to bind to or connect to. If this is used with
1726\fBnumjobs\fR to spawn multiple instances of the same job type, then
1727this will be the starting port number since fio will use a range of ports.
de890a1e 1728.TP
b93b6a2e
SB
1729.BI (net,netsplice)interface \fR=\fPstr
1730The IP address of the network interface used to send or receive UDP multicast
1731packets.
1732.TP
d3a623de
SB
1733.BI (net,netsplice)ttl \fR=\fPint
1734Time-to-live value for outgoing UDP multicast packets. Default: 1
1735.TP
1d360ffb
JA
1736.BI (net,netsplice)nodelay \fR=\fPbool
1737Set TCP_NODELAY on TCP connections.
1738.TP
de890a1e
SL
1739.BI (net,netsplice)protocol \fR=\fPstr "\fR,\fP proto" \fR=\fPstr
1740The network protocol to use. Accepted values are:
1741.RS
1742.RS
1743.TP
1744.B tcp
1745Transmission control protocol
1746.TP
49ccb8c1
JA
1747.B tcpv6
1748Transmission control protocol V6
1749.TP
de890a1e 1750.B udp
f5cc3d0e 1751User datagram protocol
de890a1e 1752.TP
49ccb8c1
JA
1753.B udpv6
1754User datagram protocol V6
1755.TP
de890a1e
SL
1756.B unix
1757UNIX domain socket
1758.RE
1759.P
1760When the protocol is TCP or UDP, the port must also be given,
1761as well as the hostname if the job is a TCP listener or UDP
1762reader. For unix sockets, the normal filename option should be
1763used and the port is invalid.
1764.RE
1765.TP
1766.BI (net,netsplice)listen
1767For TCP network connections, tell fio to listen for incoming
1768connections rather than initiating an outgoing connection. The
1769hostname must be omitted if this option is used.
d54fce84 1770.TP
7aeb1e94 1771.BI (net, pingpong) \fR=\fPbool
cecbfd47 1772Normally a network writer will just continue writing data, and a network reader
cf145d90 1773will just consume packets. If pingpong=1 is set, a writer will send its normal
7aeb1e94
JA
1774payload to the reader, then wait for the reader to send the same payload back.
1775This allows fio to measure network latencies. The submission and completion
1776latencies then measure local time spent sending or receiving, and the
1777completion latency measures how long it took for the other end to receive and
b511c9aa
SB
1778send back. For UDP multicast traffic pingpong=1 should only be set for a single
1779reader when multiple readers are listening to the same address.
7aeb1e94 1780.TP
1008602c
JA
1781.BI (net, window_size) \fR=\fPint
1782Set the desired socket buffer size for the connection.
1783.TP
e5f34d95
JA
1784.BI (net, mss) \fR=\fPint
1785Set the TCP maximum segment size (TCP_MAXSEG).
1786.TP
d54fce84
DM
1787.BI (e4defrag,donorname) \fR=\fPstr
1788File will be used as a block donor (swap extents between files)
1789.TP
1790.BI (e4defrag,inplace) \fR=\fPint
ff6bb260 1791Configure donor file block allocation strategy
d54fce84
DM
1792.RS
1793.BI 0(default) :
1794Preallocate donor's file on init
1795.TP
1796.BI 1:
cecbfd47 1797allocate space immediately inside defragment event, and free right after event
d54fce84 1798.RE
6e20c6e7
T
1799.TP
1800.BI (rbd)clustername \fR=\fPstr
1801Specifies the name of the ceph cluster.
0d978694
DAG
1802.TP
1803.BI (rbd)rbdname \fR=\fPstr
1804Specifies the name of the RBD.
1805.TP
1806.BI (rbd)pool \fR=\fPstr
1807Specifies the name of the Ceph pool containing the RBD.
1808.TP
1809.BI (rbd)clientname \fR=\fPstr
6e20c6e7 1810Specifies the username (without the 'client.' prefix) used to access the Ceph
08a2cbf6
JA
1811cluster. If the clustername is specified, the clientname shall be the full
1812type.id string. If no type. prefix is given, fio will add 'client.' by default.
65fa28ca
DE
1813.TP
1814.BI (mtd)skipbad \fR=\fPbool
1815Skip operations against known bad blocks.
d60e92d1 1816.SH OUTPUT
d1429b5c
AC
1817While running, \fBfio\fR will display the status of the created jobs. For
1818example:
d60e92d1 1819.RS
d1429b5c 1820.P
d60e92d1
AC
1821Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
1822.RE
1823.P
d1429b5c
AC
1824The characters in the first set of brackets denote the current status of each
1825threads. The possible values are:
1826.P
1827.PD 0
d60e92d1
AC
1828.RS
1829.TP
1830.B P
1831Setup but not started.
1832.TP
1833.B C
1834Thread created.
1835.TP
1836.B I
1837Initialized, waiting.
1838.TP
1839.B R
1840Running, doing sequential reads.
1841.TP
1842.B r
1843Running, doing random reads.
1844.TP
1845.B W
1846Running, doing sequential writes.
1847.TP
1848.B w
1849Running, doing random writes.
1850.TP
1851.B M
1852Running, doing mixed sequential reads/writes.
1853.TP
1854.B m
1855Running, doing mixed random reads/writes.
1856.TP
1857.B F
1858Running, currently waiting for \fBfsync\fR\|(2).
1859.TP
1860.B V
1861Running, verifying written data.
1862.TP
1863.B E
1864Exited, not reaped by main thread.
1865.TP
1866.B \-
1867Exited, thread reaped.
1868.RE
d1429b5c 1869.PD
d60e92d1
AC
1870.P
1871The second set of brackets shows the estimated completion percentage of
1872the current group. The third set shows the read and write I/O rate,
1873respectively. Finally, the estimated run time of the job is displayed.
1874.P
1875When \fBfio\fR completes (or is interrupted by Ctrl-C), it will show data
1876for each thread, each group of threads, and each disk, in that order.
1877.P
1878Per-thread statistics first show the threads client number, group-id, and
1879error code. The remaining figures are as follows:
1880.RS
d60e92d1
AC
1881.TP
1882.B io
1883Number of megabytes of I/O performed.
1884.TP
1885.B bw
1886Average data rate (bandwidth).
1887.TP
1888.B runt
1889Threads run time.
1890.TP
1891.B slat
1892Submission latency minimum, maximum, average and standard deviation. This is
1893the time it took to submit the I/O.
1894.TP
1895.B clat
1896Completion latency minimum, maximum, average and standard deviation. This
1897is the time between submission and completion.
1898.TP
1899.B bw
1900Bandwidth minimum, maximum, percentage of aggregate bandwidth received, average
1901and standard deviation.
1902.TP
1903.B cpu
1904CPU usage statistics. Includes user and system time, number of context switches
23a8e176
JA
1905this thread went through and number of major and minor page faults. The CPU
1906utilization numbers are averages for the jobs in that reporting group, while
1907the context and fault counters are summed.
d60e92d1
AC
1908.TP
1909.B IO depths
1910Distribution of I/O depths. Each depth includes everything less than (or equal)
1911to it, but greater than the previous depth.
1912.TP
1913.B IO issued
1914Number of read/write requests issued, and number of short read/write requests.
1915.TP
1916.B IO latencies
1917Distribution of I/O completion latencies. The numbers follow the same pattern
1918as \fBIO depths\fR.
1919.RE
d60e92d1
AC
1920.P
1921The group statistics show:
d1429b5c 1922.PD 0
d60e92d1
AC
1923.RS
1924.TP
1925.B io
1926Number of megabytes I/O performed.
1927.TP
1928.B aggrb
1929Aggregate bandwidth of threads in the group.
1930.TP
1931.B minb
1932Minimum average bandwidth a thread saw.
1933.TP
1934.B maxb
1935Maximum average bandwidth a thread saw.
1936.TP
1937.B mint
d1429b5c 1938Shortest runtime of threads in the group.
d60e92d1
AC
1939.TP
1940.B maxt
1941Longest runtime of threads in the group.
1942.RE
d1429b5c 1943.PD
d60e92d1
AC
1944.P
1945Finally, disk statistics are printed with reads first:
d1429b5c 1946.PD 0
d60e92d1
AC
1947.RS
1948.TP
1949.B ios
1950Number of I/Os performed by all groups.
1951.TP
1952.B merge
1953Number of merges in the I/O scheduler.
1954.TP
1955.B ticks
1956Number of ticks we kept the disk busy.
1957.TP
1958.B io_queue
1959Total time spent in the disk queue.
1960.TP
1961.B util
1962Disk utilization.
1963.RE
d1429b5c 1964.PD
8423bd11
JA
1965.P
1966It is also possible to get fio to dump the current output while it is
1967running, without terminating the job. To do that, send fio the \fBUSR1\fR
1968signal.
d60e92d1 1969.SH TERSE OUTPUT
2b8c71b0
CE
1970If the \fB\-\-minimal\fR / \fB\-\-append-terse\fR options are given, the
1971results will be printed/appended in a semicolon-delimited format suitable for
1972scripted use.
1973A job description (if provided) follows on a new line. Note that the first
525c2bfa
JA
1974number in the line is the version number. If the output has to be changed
1975for some reason, this number will be incremented by 1 to signify that
1976change. The fields are:
d60e92d1
AC
1977.P
1978.RS
5e726d0a 1979.B terse version, fio version, jobname, groupid, error
d60e92d1
AC
1980.P
1981Read status:
1982.RS
312b4af2 1983.B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, IOPS, runtime \fR(ms)\fP
d60e92d1
AC
1984.P
1985Submission latency:
1986.RS
1987.B min, max, mean, standard deviation
1988.RE
1989Completion latency:
1990.RS
1991.B min, max, mean, standard deviation
1992.RE
1db92cb6
JA
1993Completion latency percentiles (20 fields):
1994.RS
1995.B Xth percentile=usec
1996.RE
525c2bfa
JA
1997Total latency:
1998.RS
1999.B min, max, mean, standard deviation
2000.RE
d60e92d1
AC
2001Bandwidth:
2002.RS
2003.B min, max, aggregate percentage of total, mean, standard deviation
2004.RE
2005.RE
2006.P
2007Write status:
2008.RS
312b4af2 2009.B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, IOPS, runtime \fR(ms)\fP
d60e92d1
AC
2010.P
2011Submission latency:
2012.RS
2013.B min, max, mean, standard deviation
2014.RE
2015Completion latency:
2016.RS
2017.B min, max, mean, standard deviation
2018.RE
1db92cb6
JA
2019Completion latency percentiles (20 fields):
2020.RS
2021.B Xth percentile=usec
2022.RE
525c2bfa
JA
2023Total latency:
2024.RS
2025.B min, max, mean, standard deviation
2026.RE
d60e92d1
AC
2027Bandwidth:
2028.RS
2029.B min, max, aggregate percentage of total, mean, standard deviation
2030.RE
2031.RE
2032.P
d1429b5c 2033CPU usage:
d60e92d1 2034.RS
bd2626f0 2035.B user, system, context switches, major page faults, minor page faults
d60e92d1
AC
2036.RE
2037.P
2038IO depth distribution:
2039.RS
2040.B <=1, 2, 4, 8, 16, 32, >=64
2041.RE
2042.P
562c2d2f 2043IO latency distribution:
d60e92d1 2044.RS
562c2d2f
DN
2045Microseconds:
2046.RS
2047.B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
2048.RE
2049Milliseconds:
2050.RS
2051.B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
2052.RE
2053.RE
2054.P
f2f788dd
JA
2055Disk utilization (1 for each disk used):
2056.RS
2057.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
2058.RE
2059.P
5982a925 2060Error Info (dependent on continue_on_error, default off):
562c2d2f 2061.RS
ff6bb260 2062.B total # errors, first error code
d60e92d1
AC
2063.RE
2064.P
562c2d2f 2065.B text description (if provided in config - appears on newline)
d60e92d1 2066.RE
29dbd1e5
JA
2067.SH TRACE FILE FORMAT
2068There are two trace file format that you can encounter. The older (v1) format
2069is unsupported since version 1.20-rc3 (March 2008). It will still be described
2070below in case that you get an old trace and want to understand it.
2071
2072In any case the trace is a simple text file with a single action per line.
2073
2074.P
2075.B Trace file format v1
2076.RS
2077Each line represents a single io action in the following format:
2078
2079rw, offset, length
2080
2081where rw=0/1 for read/write, and the offset and length entries being in bytes.
2082
2083This format is not supported in Fio versions => 1.20-rc3.
2084
2085.RE
2086.P
2087.B Trace file format v2
2088.RS
2089The second version of the trace file format was added in Fio version 1.17.
8fb5444d 2090It allows one to access more then one file per trace and has a bigger set of
29dbd1e5
JA
2091possible file actions.
2092
2093The first line of the trace file has to be:
2094
2095\fBfio version 2 iolog\fR
2096
2097Following this can be lines in two different formats, which are described below.
2098The file management format:
2099
2100\fBfilename action\fR
2101
2102The filename is given as an absolute path. The action can be one of these:
2103
2104.P
2105.PD 0
2106.RS
2107.TP
2108.B add
2109Add the given filename to the trace
2110.TP
2111.B open
2112Open the file with the given filename. The filename has to have been previously
2113added with the \fBadd\fR action.
2114.TP
2115.B close
2116Close the file with the given filename. The file must have previously been
2117opened.
2118.RE
2119.PD
2120.P
2121
2122The file io action format:
2123
2124\fBfilename action offset length\fR
2125
2126The filename is given as an absolute path, and has to have been added and opened
2127before it can be used with this format. The offset and length are given in
2128bytes. The action can be one of these:
2129
2130.P
2131.PD 0
2132.RS
2133.TP
2134.B wait
2135Wait for 'offset' microseconds. Everything below 100 is discarded. The time is
2136relative to the previous wait statement.
2137.TP
2138.B read
2139Read \fBlength\fR bytes beginning from \fBoffset\fR
2140.TP
2141.B write
2142Write \fBlength\fR bytes beginning from \fBoffset\fR
2143.TP
2144.B sync
2145fsync() the file
2146.TP
2147.B datasync
2148fdatasync() the file
2149.TP
2150.B trim
2151trim the given file from the given \fBoffset\fR for \fBlength\fR bytes
2152.RE
2153.PD
2154.P
2155
2156.SH CPU IDLENESS PROFILING
2157In some cases, we want to understand CPU overhead in a test. For example,
2158we test patches for the specific goodness of whether they reduce CPU usage.
2159fio implements a balloon approach to create a thread per CPU that runs at
2160idle priority, meaning that it only runs when nobody else needs the cpu.
2161By measuring the amount of work completed by the thread, idleness of each
2162CPU can be derived accordingly.
2163
2164An unit work is defined as touching a full page of unsigned characters. Mean
2165and standard deviation of time to complete an unit work is reported in "unit
2166work" section. Options can be chosen to report detailed percpu idleness or
2167overall system idleness by aggregating percpu stats.
2168
2169.SH VERIFICATION AND TRIGGERS
2170Fio is usually run in one of two ways, when data verification is done. The
2171first is a normal write job of some sort with verify enabled. When the
2172write phase has completed, fio switches to reads and verifies everything
2173it wrote. The second model is running just the write phase, and then later
2174on running the same job (but with reads instead of writes) to repeat the
2175same IO patterns and verify the contents. Both of these methods depend
2176on the write phase being completed, as fio otherwise has no idea how much
2177data was written.
2178
2179With verification triggers, fio supports dumping the current write state
2180to local files. Then a subsequent read verify workload can load this state
2181and know exactly where to stop. This is useful for testing cases where
2182power is cut to a server in a managed fashion, for instance.
2183
2184A verification trigger consists of two things:
2185
2186.RS
2187Storing the write state of each job
2188.LP
2189Executing a trigger command
2190.RE
2191
2192The write state is relatively small, on the order of hundreds of bytes
2193to single kilobytes. It contains information on the number of completions
2194done, the last X completions, etc.
2195
2196A trigger is invoked either through creation (\fBtouch\fR) of a specified
2197file in the system, or through a timeout setting. If fio is run with
2198\fB\-\-trigger\-file=/tmp/trigger-file\fR, then it will continually check for
2199the existence of /tmp/trigger-file. When it sees this file, it will
2200fire off the trigger (thus saving state, and executing the trigger
2201command).
2202
2203For client/server runs, there's both a local and remote trigger. If
2204fio is running as a server backend, it will send the job states back
2205to the client for safe storage, then execute the remote trigger, if
2206specified. If a local trigger is specified, the server will still send
2207back the write state, but the client will then execute the trigger.
2208
2209.RE
2210.P
2211.B Verification trigger example
2212.RS
2213
2214Lets say we want to run a powercut test on the remote machine 'server'.
2215Our write workload is in write-test.fio. We want to cut power to 'server'
2216at some point during the run, and we'll run this test from the safety
2217or our local machine, 'localbox'. On the server, we'll start the fio
2218backend normally:
2219
2220server# \fBfio \-\-server\fR
2221
2222and on the client, we'll fire off the workload:
2223
e0ee7a8b 2224localbox$ \fBfio \-\-client=server \-\-trigger\-file=/tmp/my\-trigger \-\-trigger-remote="bash \-c "echo b > /proc/sysrq-triger""\fR
29dbd1e5
JA
2225
2226We set \fB/tmp/my-trigger\fR as the trigger file, and we tell fio to execute
2227
2228\fBecho b > /proc/sysrq-trigger\fR
2229
2230on the server once it has received the trigger and sent us the write
2231state. This will work, but it's not \fIreally\fR cutting power to the server,
2232it's merely abruptly rebooting it. If we have a remote way of cutting
2233power to the server through IPMI or similar, we could do that through
2234a local trigger command instead. Lets assume we have a script that does
2235IPMI reboot of a given hostname, ipmi-reboot. On localbox, we could
2236then have run fio with a local trigger instead:
2237
2238localbox$ \fBfio \-\-client=server \-\-trigger\-file=/tmp/my\-trigger \-\-trigger="ipmi-reboot server"\fR
2239
2240For this case, fio would wait for the server to send us the write state,
2241then execute 'ipmi-reboot server' when that happened.
2242
2243.RE
2244.P
2245.B Loading verify state
2246.RS
2247To load store write state, read verification job file must contain
2248the verify_state_load option. If that is set, fio will load the previously
2249stored state. For a local fio run this is done by loading the files directly,
2250and on a client/server run, the server backend will ask the client to send
2251the files over and load them from there.
2252
2253.RE
2254
a3ae5b05
JA
2255.SH LOG FILE FORMATS
2256
2257Fio supports a variety of log file formats, for logging latencies, bandwidth,
2258and IOPS. The logs share a common format, which looks like this:
2259
2260.B time (msec), value, data direction, offset
2261
2262Time for the log entry is always in milliseconds. The value logged depends
2263on the type of log, it will be one of the following:
2264
2265.P
2266.PD 0
2267.TP
2268.B Latency log
2269Value is in latency in usecs
2270.TP
2271.B Bandwidth log
2272Value is in KB/sec
2273.TP
2274.B IOPS log
2275Value is in IOPS
2276.PD
2277.P
2278
2279Data direction is one of the following:
2280
2281.P
2282.PD 0
2283.TP
2284.B 0
2285IO is a READ
2286.TP
2287.B 1
2288IO is a WRITE
2289.TP
2290.B 2
2291IO is a TRIM
2292.PD
2293.P
2294
2295The \fIoffset\fR is the offset, in bytes, from the start of the file, for that
2296particular IO. The logging of the offset can be toggled with \fBlog_offset\fR.
2297
2298If windowed logging is enabled though \fBlog_avg_msec\fR, then fio doesn't log
2299individual IOs. Instead of logs the average values over the specified
2300period of time. Since \fIdata direction\fR and \fIoffset\fR are per-IO values,
2301they aren't applicable if windowed logging is enabled. If windowed logging
2302is enabled and \fBlog_max_value\fR is set, then fio logs maximum values in
2303that window instead of averages.
2304
2305.RE
2306
49da1240
JA
2307.SH CLIENT / SERVER
2308Normally you would run fio as a stand-alone application on the machine
2309where the IO workload should be generated. However, it is also possible to
2310run the frontend and backend of fio separately. This makes it possible to
2311have a fio server running on the machine(s) where the IO workload should
2312be running, while controlling it from another machine.
2313
2314To start the server, you would do:
2315
2316\fBfio \-\-server=args\fR
2317
2318on that machine, where args defines what fio listens to. The arguments
811826be 2319are of the form 'type:hostname or IP:port'. 'type' is either 'ip' (or ip4)
20c67f10
MS
2320for TCP/IP v4, 'ip6' for TCP/IP v6, or 'sock' for a local unix domain
2321socket. 'hostname' is either a hostname or IP address, and 'port' is the port to
811826be 2322listen to (only valid for TCP/IP, not a local socket). Some examples:
49da1240 2323
e0ee7a8b 23241) \fBfio \-\-server\fR
49da1240
JA
2325
2326 Start a fio server, listening on all interfaces on the default port (8765).
2327
e0ee7a8b 23282) \fBfio \-\-server=ip:hostname,4444\fR
49da1240
JA
2329
2330 Start a fio server, listening on IP belonging to hostname and on port 4444.
2331
e0ee7a8b 23323) \fBfio \-\-server=ip6:::1,4444\fR
811826be
JA
2333
2334 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
2335
e0ee7a8b 23364) \fBfio \-\-server=,4444\fR
49da1240
JA
2337
2338 Start a fio server, listening on all interfaces on port 4444.
2339
e0ee7a8b 23405) \fBfio \-\-server=1.2.3.4\fR
49da1240
JA
2341
2342 Start a fio server, listening on IP 1.2.3.4 on the default port.
2343
e0ee7a8b 23446) \fBfio \-\-server=sock:/tmp/fio.sock\fR
49da1240
JA
2345
2346 Start a fio server, listening on the local socket /tmp/fio.sock.
2347
2348When a server is running, you can connect to it from a client. The client
2349is run with:
2350
e0ee7a8b 2351\fBfio \-\-local-args \-\-client=server \-\-remote-args <job file(s)>\fR
49da1240 2352
e01e9745
MS
2353where \-\-local-args are arguments that are local to the client where it is
2354running, 'server' is the connect string, and \-\-remote-args and <job file(s)>
49da1240
JA
2355are sent to the server. The 'server' string follows the same format as it
2356does on the server side, to allow IP/hostname/socket and port strings.
2357You can connect to multiple clients as well, to do that you could run:
2358
e0ee7a8b 2359\fBfio \-\-client=server2 \-\-client=server2 <job file(s)>\fR
323255cc
JA
2360
2361If the job file is located on the fio server, then you can tell the server
2362to load a local file as well. This is done by using \-\-remote-config:
2363
e0ee7a8b 2364\fBfio \-\-client=server \-\-remote-config /path/to/file.fio\fR
323255cc 2365
39b5f61e 2366Then fio will open this local (to the server) job file instead
323255cc 2367of being passed one from the client.
39b5f61e 2368
ff6bb260 2369If you have many servers (example: 100 VMs/containers), you can input a pathname
39b5f61e
BE
2370of a file containing host IPs/names as the parameter value for the \-\-client option.
2371For example, here is an example "host.list" file containing 2 hostnames:
2372
2373host1.your.dns.domain
2374.br
2375host2.your.dns.domain
2376
2377The fio command would then be:
2378
e0ee7a8b 2379\fBfio \-\-client=host.list <job file>\fR
39b5f61e
BE
2380
2381In this mode, you cannot input server-specific parameters or job files, and all
2382servers receive the same job file.
2383
2384In order to enable fio \-\-client runs utilizing a shared filesystem from multiple hosts,
ff6bb260
SL
2385fio \-\-client now prepends the IP address of the server to the filename. For example,
2386if fio is using directory /mnt/nfs/fio and is writing filename fileio.tmp,
39b5f61e
BE
2387with a \-\-client hostfile
2388containing two hostnames h1 and h2 with IP addresses 192.168.10.120 and 192.168.10.121, then
2389fio will create two files:
2390
2391/mnt/nfs/fio/192.168.10.120.fileio.tmp
2392.br
2393/mnt/nfs/fio/192.168.10.121.fileio.tmp
2394
d60e92d1 2395.SH AUTHORS
49da1240 2396
d60e92d1 2397.B fio
aa58d252 2398was written by Jens Axboe <jens.axboe@oracle.com>,
f8b8f7da 2399now Jens Axboe <axboe@fb.com>.
d1429b5c
AC
2400.br
2401This man page was written by Aaron Carroll <aaronc@cse.unsw.edu.au> based
d60e92d1
AC
2402on documentation by Jens Axboe.
2403.SH "REPORTING BUGS"
482900c9 2404Report bugs to the \fBfio\fR mailing list <fio@vger.kernel.org>.
d1429b5c 2405See \fBREADME\fR.
d60e92d1 2406.SH "SEE ALSO"
d1429b5c
AC
2407For further documentation see \fBHOWTO\fR and \fBREADME\fR.
2408.br
2409Sample jobfiles are available in the \fBexamples\fR directory.