Fix potential oops in file unlocking at free time
[fio.git] / fio.1
CommitLineData
65f3c785 1.TH fio 1 "October 2013" "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
b2cecdc2 23.BI \-\-runtime \fR=\fPruntime
24Limit run time to \fIruntime\fR seconds.
d60e92d1
AC
25.TP
26.B \-\-latency\-log
27Generate per-job latency logs.
28.TP
29.B \-\-bandwidth\-log
30Generate per-job bandwidth logs.
31.TP
32.B \-\-minimal
d1429b5c 33Print statistics in a terse, semicolon-delimited format.
d60e92d1 34.TP
49da1240
JA
35.B \-\-version
36Display version information and exit.
37.TP
065248bf 38.BI \-\-terse\-version \fR=\fPversion
4d658652 39Set terse version output format (Current version 3, or older version 2).
49da1240
JA
40.TP
41.B \-\-help
42Display usage information and exit.
43.TP
44.BI \-\-cmdhelp \fR=\fPcommand
45Print help information for \fIcommand\fR. May be `all' for all commands.
46.TP
de890a1e
SL
47.BI \-\-enghelp \fR=\fPioengine[,command]
48List all commands defined by \fIioengine\fR, or print help for \fIcommand\fR defined by \fIioengine\fR.
49.TP
d60e92d1
AC
50.BI \-\-showcmd \fR=\fPjobfile
51Convert \fIjobfile\fR to a set of command-line options.
52.TP
d60e92d1
AC
53.BI \-\-eta \fR=\fPwhen
54Specifies when real-time ETA estimate should be printed. \fIwhen\fR may
55be one of `always', `never' or `auto'.
56.TP
30b5d57f
JA
57.BI \-\-eta\-newline \fR=\fPtime
58Force an ETA newline for every `time` period passed.
59.TP
60.BI \-\-status\-interval \fR=\fPtime
61Report full output status every `time` period passed.
62.TP
49da1240
JA
63.BI \-\-readonly
64Turn on safety read-only checks, preventing any attempted write.
65.TP
c0a5d35e 66.BI \-\-section \fR=\fPsec
49da1240 67Only run section \fIsec\fR from job file. Multiple of these options can be given, adding more sections to run.
c0a5d35e 68.TP
49da1240
JA
69.BI \-\-alloc\-size \fR=\fPkb
70Set the internal smalloc pool size to \fIkb\fP kilobytes.
d60e92d1 71.TP
49da1240
JA
72.BI \-\-warnings\-fatal
73All fio parser warnings are fatal, causing fio to exit with an error.
9183788d 74.TP
49da1240 75.BI \-\-max\-jobs \fR=\fPnr
57e118a2 76Set the maximum allowed number of jobs (threads/processes) to support.
d60e92d1 77.TP
49da1240
JA
78.BI \-\-server \fR=\fPargs
79Start a backend server, with \fIargs\fP specifying what to listen to. See client/server section.
f57a9c59 80.TP
49da1240
JA
81.BI \-\-daemonize \fR=\fPpidfile
82Background a fio server, writing the pid to the given pid file.
83.TP
84.BI \-\-client \fR=\fPhost
85Instead of running the jobs locally, send and run them on the given host.
f2a2ce0e
HL
86.TP
87.BI \-\-idle\-prof \fR=\fPoption
88Report cpu idleness on a system or percpu basis (\fIoption\fP=system,percpu) or run unit work calibration only (\fIoption\fP=calibrate).
d60e92d1
AC
89.SH "JOB FILE FORMAT"
90Job files are in `ini' format. They consist of one or more
91job definitions, which begin with a job name in square brackets and
92extend to the next job name. The job name can be any ASCII string
93except `global', which has a special meaning. Following the job name is
94a sequence of zero or more parameters, one per line, that define the
95behavior of the job. Any line starting with a `;' or `#' character is
d1429b5c 96considered a comment and ignored.
d9956b64
AC
97.P
98If \fIjobfile\fR is specified as `-', the job file will be read from
99standard input.
d60e92d1
AC
100.SS "Global Section"
101The global section contains default parameters for jobs specified in the
102job file. A job is only affected by global sections residing above it,
103and there may be any number of global sections. Specific job definitions
104may override any parameter set in global sections.
105.SH "JOB PARAMETERS"
106.SS Types
107Some parameters may take arguments of a specific type. The types used are:
108.TP
109.I str
110String: a sequence of alphanumeric characters.
111.TP
112.I int
d60e92d1 113SI integer: a whole number, possibly containing a suffix denoting the base unit
b09da8fa
JA
114of the value. Accepted suffixes are `k', 'M', 'G', 'T', and 'P', denoting
115kilo (1024), mega (1024^2), giga (1024^3), tera (1024^4), and peta (1024^5)
116respectively. The suffix is not case sensitive. If prefixed with '0x', the
5982a925
MS
117value is assumed to be base 16 (hexadecimal). A suffix may include a trailing 'b',
118for instance 'kb' is identical to 'k'. You can specify a base 10 value
57fc29fa
JA
119by using 'KiB', 'MiB', 'GiB', etc. This is useful for disk drives where
120values are often given in base 10 values. Specifying '30GiB' will get you
12130*1000^3 bytes.
d60e92d1
AC
122.TP
123.I bool
124Boolean: a true or false value. `0' denotes false, `1' denotes true.
125.TP
126.I irange
127Integer range: a range of integers specified in the format
d1429b5c
AC
128\fIlower\fR:\fIupper\fR or \fIlower\fR\-\fIupper\fR. \fIlower\fR and
129\fIupper\fR may contain a suffix as described above. If an option allows two
130sets of ranges, they are separated with a `,' or `/' character. For example:
131`8\-8k/8M\-4G'.
83349190
YH
132.TP
133.I float_list
134List of floating numbers: A list of floating numbers, separated by
cecbfd47 135a ':' character.
d60e92d1
AC
136.SS "Parameter List"
137.TP
138.BI name \fR=\fPstr
d9956b64 139May be used to override the job name. On the command line, this parameter
d60e92d1
AC
140has the special purpose of signalling the start of a new job.
141.TP
142.BI description \fR=\fPstr
143Human-readable description of the job. It is printed when the job is run, but
144otherwise has no special purpose.
145.TP
146.BI directory \fR=\fPstr
147Prefix filenames with this directory. Used to place files in a location other
148than `./'.
149.TP
150.BI filename \fR=\fPstr
151.B fio
152normally makes up a file name based on the job name, thread number, and file
d1429b5c 153number. If you want to share files between threads in a job or several jobs,
de890a1e
SL
154specify a \fIfilename\fR for each of them to override the default.
155If the I/O engine is file-based, you can specify
d1429b5c
AC
156a number of files by separating the names with a `:' character. `\-' is a
157reserved name, meaning stdin or stdout, depending on the read/write direction
158set.
d60e92d1 159.TP
de98bd30 160.BI filename_format \fR=\fPstr
ce594fbe 161If sharing multiple files between jobs, it is usually necessary to have
de98bd30
JA
162fio generate the exact names that you want. By default, fio will name a file
163based on the default file format specification of
164\fBjobname.jobnumber.filenumber\fP. With this option, that can be
165customized. Fio will recognize and replace the following keywords in this
166string:
167.RS
168.RS
169.TP
170.B $jobname
171The name of the worker thread or process.
172.TP
173.B $jobnum
174The incremental number of the worker thread or process.
175.TP
176.B $filenum
177The incremental number of the file for that worker thread or process.
178.RE
179.P
180To have dependent jobs share a set of files, this option can be set to
181have fio generate filenames that are shared between the two. For instance,
182if \fBtestfiles.$filenum\fR is specified, file number 4 for any job will
183be named \fBtestfiles.4\fR. The default of \fB$jobname.$jobnum.$filenum\fR
184will be used if no other format specifier is given.
185.RE
186.P
187.TP
3ce9dcaf
JA
188.BI lockfile \fR=\fPstr
189Fio defaults to not locking any files before it does IO to them. If a file or
190file descriptor is shared, fio can serialize IO to that file to make the end
191result consistent. This is usual for emulating real workloads that share files.
192The lock modes are:
193.RS
194.RS
195.TP
196.B none
197No locking. This is the default.
198.TP
199.B exclusive
200Only one thread or process may do IO at the time, excluding all others.
201.TP
202.B readwrite
203Read-write locking on the file. Many readers may access the file at the same
204time, but writes get exclusive access.
205.RE
ce594fbe 206.RE
3ce9dcaf 207.P
d60e92d1
AC
208.BI opendir \fR=\fPstr
209Recursively open any files below directory \fIstr\fR.
210.TP
211.BI readwrite \fR=\fPstr "\fR,\fP rw" \fR=\fPstr
212Type of I/O pattern. Accepted values are:
213.RS
214.RS
215.TP
216.B read
d1429b5c 217Sequential reads.
d60e92d1
AC
218.TP
219.B write
d1429b5c 220Sequential writes.
d60e92d1 221.TP
fa769d44
SW
222.B trim
223Sequential trim (Linux block devices only).
224.TP
d60e92d1 225.B randread
d1429b5c 226Random reads.
d60e92d1
AC
227.TP
228.B randwrite
d1429b5c 229Random writes.
d60e92d1 230.TP
fa769d44
SW
231.B randtrim
232Random trim (Linux block devices only).
233.TP
10b023db 234.B rw, readwrite
d1429b5c 235Mixed sequential reads and writes.
d60e92d1
AC
236.TP
237.B randrw
d1429b5c 238Mixed random reads and writes.
d60e92d1
AC
239.RE
240.P
38dad62d
JA
241For mixed I/O, the default split is 50/50. For certain types of io the result
242may still be skewed a bit, since the speed may be different. It is possible to
3b7fa9ec 243specify a number of IO's to do before getting a new offset, this is done by
38dad62d
JA
244appending a `:\fI<nr>\fR to the end of the string given. For a random read, it
245would look like \fBrw=randread:8\fR for passing in an offset modifier with a
059b0802
JA
246value of 8. If the postfix is used with a sequential IO pattern, then the value
247specified will be added to the generated offset for each IO. For instance,
248using \fBrw=write:4k\fR will skip 4k for every write. It turns sequential IO
249into sequential IO with holes. See the \fBrw_sequencer\fR option.
d60e92d1
AC
250.RE
251.TP
38dad62d
JA
252.BI rw_sequencer \fR=\fPstr
253If an offset modifier is given by appending a number to the \fBrw=<str>\fR line,
254then this option controls how that number modifies the IO offset being
255generated. Accepted values are:
256.RS
257.RS
258.TP
259.B sequential
260Generate sequential offset
261.TP
262.B identical
263Generate the same offset
264.RE
265.P
266\fBsequential\fR is only useful for random IO, where fio would normally
267generate a new random offset for every IO. If you append eg 8 to randread, you
268would get a new random offset for every 8 IO's. The result would be a seek for
269only every 8 IO's, instead of for every IO. Use \fBrw=randread:8\fR to specify
270that. As sequential IO is already sequential, setting \fBsequential\fR for that
271would not result in any differences. \fBidentical\fR behaves in a similar
272fashion, except it sends the same offset 8 number of times before generating a
273new offset.
274.RE
275.P
276.TP
90fef2d1
JA
277.BI kb_base \fR=\fPint
278The base unit for a kilobyte. The defacto base is 2^10, 1024. Storage
279manufacturers like to use 10^3 or 1000 as a base ten unit instead, for obvious
5c9323fb 280reasons. Allowed values are 1024 or 1000, with 1024 being the default.
90fef2d1 281.TP
771e58be
JA
282.BI unified_rw_reporting \fR=\fPbool
283Fio normally reports statistics on a per data direction basis, meaning that
284read, write, and trim are accounted and reported separately. If this option is
285set, the fio will sum the results and report them as "mixed" instead.
286.TP
d60e92d1
AC
287.BI randrepeat \fR=\fPbool
288Seed the random number generator in a predictable way so results are repeatable
d1429b5c 289across runs. Default: true.
d60e92d1 290.TP
04778baf
JA
291.BI randseed \fR=\fPint
292Seed the random number generators based on this seed value, to be able to
293control what sequence of output is being generated. If not set, the random
294sequence depends on the \fBrandrepeat\fR setting.
295.TP
2615cc4b
JA
296.BI use_os_rand \fR=\fPbool
297Fio can either use the random generator supplied by the OS to generator random
298offsets, or it can use it's own internal generator (based on Tausworthe).
299Default is to use the internal generator, which is often of better quality and
300faster. Default: false.
301.TP
a596f047
EG
302.BI fallocate \fR=\fPstr
303Whether pre-allocation is performed when laying down files. Accepted values
304are:
305.RS
306.RS
307.TP
308.B none
309Do not pre-allocate space.
310.TP
311.B posix
ccc2b328 312Pre-allocate via \fBposix_fallocate\fR\|(3).
a596f047
EG
313.TP
314.B keep
ccc2b328 315Pre-allocate via \fBfallocate\fR\|(2) with FALLOC_FL_KEEP_SIZE set.
a596f047
EG
316.TP
317.B 0
318Backward-compatible alias for 'none'.
319.TP
320.B 1
321Backward-compatible alias for 'posix'.
322.RE
323.P
324May not be available on all supported platforms. 'keep' is only
325available on Linux. If using ZFS on Solaris this must be set to 'none'
326because ZFS doesn't support it. Default: 'posix'.
327.RE
7bc8c2cf 328.TP
d60e92d1 329.BI fadvise_hint \fR=\fPbool
ccc2b328 330Use of \fBposix_fadvise\fR\|(2) to advise the kernel what I/O patterns
d1429b5c 331are likely to be issued. Default: true.
d60e92d1 332.TP
f7fa2653 333.BI size \fR=\fPint
d60e92d1 334Total size of I/O for this job. \fBfio\fR will run until this many bytes have
ca45881f 335been transferred, unless limited by other options (\fBruntime\fR, for instance).
d7c8be03 336Unless \fBnrfiles\fR and \fBfilesize\fR options are given, this amount will be
d6667268 337divided between the available files for the job. If not set, fio will use the
cecbfd47 338full size of the given files or devices. If the files do not exist, size
7bb59102
JA
339must be given. It is also possible to give size as a percentage between 1 and
340100. If size=20% is given, fio will use 20% of the full size of the given files
341or devices.
d60e92d1 342.TP
74586c1e 343.BI fill_device \fR=\fPbool "\fR,\fB fill_fs" \fR=\fPbool
3ce9dcaf
JA
344Sets size to something really large and waits for ENOSPC (no space left on
345device) as the terminating condition. Only makes sense with sequential write.
346For a read workload, the mount point will be filled first then IO started on
4f12432e
JA
347the result. This option doesn't make sense if operating on a raw device node,
348since the size of that is already known by the file system. Additionally,
349writing beyond end-of-device will not return ENOSPC there.
3ce9dcaf 350.TP
d60e92d1
AC
351.BI filesize \fR=\fPirange
352Individual file sizes. May be a range, in which case \fBfio\fR will select sizes
d1429b5c
AC
353for files at random within the given range, limited to \fBsize\fR in total (if
354that is given). If \fBfilesize\fR is not specified, each created file is the
355same size.
d60e92d1 356.TP
f7fa2653 357.BI blocksize \fR=\fPint[,int] "\fR,\fB bs" \fR=\fPint[,int]
d9472271
JA
358Block size for I/O units. Default: 4k. Values for reads, writes, and trims
359can be specified separately in the format \fIread\fR,\fIwrite\fR,\fItrim\fR
360either of which may be empty to leave that value at its default. If a trailing
361comma isn't given, the remainder will inherit the last value set.
d60e92d1 362.TP
9183788d 363.BI blocksize_range \fR=\fPirange[,irange] "\fR,\fB bsrange" \fR=\fPirange[,irange]
d1429b5c
AC
364Specify a range of I/O block sizes. The issued I/O unit will always be a
365multiple of the minimum size, unless \fBblocksize_unaligned\fR is set. Applies
9183788d 366to both reads and writes if only one range is given, but can be specified
de8f6de9 367separately with a comma separating the values. Example: bsrange=1k-4k,2k-8k.
9183788d
JA
368Also (see \fBblocksize\fR).
369.TP
370.BI bssplit \fR=\fPstr
371This option allows even finer grained control of the block sizes issued,
372not just even splits between them. With this option, you can weight various
373block sizes for exact control of the issued IO for a job that has mixed
374block sizes. The format of the option is bssplit=blocksize/percentage,
5982a925 375optionally adding as many definitions as needed separated by a colon.
9183788d 376Example: bssplit=4k/10:64k/50:32k/40 would issue 50% 64k blocks, 10% 4k
c83cdd3e
JA
377blocks and 40% 32k blocks. \fBbssplit\fR also supports giving separate
378splits to reads and writes. The format is identical to what the
379\fBbs\fR option accepts, the read and write parts are separated with a
380comma.
d60e92d1
AC
381.TP
382.B blocksize_unaligned\fR,\fP bs_unaligned
d1429b5c
AC
383If set, any size in \fBblocksize_range\fR may be used. This typically won't
384work with direct I/O, as that normally requires sector alignment.
d60e92d1 385.TP
2b7a01d0 386.BI blockalign \fR=\fPint[,int] "\fR,\fB ba" \fR=\fPint[,int]
639ce0f3
MS
387At what boundary to align random IO offsets. Defaults to the same as 'blocksize'
388the minimum blocksize given. Minimum alignment is typically 512b
2b7a01d0
JA
389for using direct IO, though it usually depends on the hardware block size.
390This option is mutually exclusive with using a random map for files, so it
391will turn off that option.
43602667 392.TP
6aca9b3d
JA
393.BI bs_is_seq_rand \fR=\fPbool
394If this option is set, fio will use the normal read,write blocksize settings as
395sequential,random instead. Any random read or write will use the WRITE
396blocksize settings, and any sequential read or write will use the READ
397blocksize setting.
398.TP
d60e92d1
AC
399.B zero_buffers
400Initialise buffers with all zeros. Default: fill buffers with random data.
401.TP
901bb994
JA
402.B refill_buffers
403If this option is given, fio will refill the IO buffers on every submit. The
404default is to only fill it at init time and reuse that data. Only makes sense
405if zero_buffers isn't specified, naturally. If data verification is enabled,
406refill_buffers is also automatically enabled.
407.TP
fd68418e
JA
408.BI scramble_buffers \fR=\fPbool
409If \fBrefill_buffers\fR is too costly and the target is using data
410deduplication, then setting this option will slightly modify the IO buffer
411contents to defeat normal de-dupe attempts. This is not enough to defeat
412more clever block compression attempts, but it will stop naive dedupe
413of blocks. Default: true.
414.TP
c5751c62
JA
415.BI buffer_compress_percentage \fR=\fPint
416If this is set, then fio will attempt to provide IO buffer content (on WRITEs)
417that compress to the specified level. Fio does this by providing a mix of
418random data and zeroes. Note that this is per block size unit, for file/disk
419wide compression level that matches this setting, you'll also want to set
420\fBrefill_buffers\fR.
421.TP
422.BI buffer_compress_chunk \fR=\fPint
423See \fBbuffer_compress_percentage\fR. This setting allows fio to manage how
424big the ranges of random data and zeroed data is. Without this set, fio will
425provide \fBbuffer_compress_percentage\fR of blocksize random data, followed by
426the remaining zeroed. With this set to some chunk size smaller than the block
427size, fio can alternate random and zeroed data throughout the IO buffer.
428.TP
ce35b1ec
JA
429.BI buffer_pattern \fR=\fPstr
430If set, fio will fill the io buffers with this pattern. If not set, the contents
431of io buffers is defined by the other options related to buffer contents. The
432setting can be any pattern of bytes, and can be prefixed with 0x for hex
433values.
434.TP
d60e92d1
AC
435.BI nrfiles \fR=\fPint
436Number of files to use for this job. Default: 1.
437.TP
438.BI openfiles \fR=\fPint
439Number of files to keep open at the same time. Default: \fBnrfiles\fR.
440.TP
441.BI file_service_type \fR=\fPstr
442Defines how files to service are selected. The following types are defined:
443.RS
444.RS
445.TP
446.B random
5c9323fb 447Choose a file at random.
d60e92d1
AC
448.TP
449.B roundrobin
450Round robin over open files (default).
5c9323fb 451.TP
6b7f6851
JA
452.B sequential
453Do each file in the set sequentially.
d60e92d1
AC
454.RE
455.P
456The number of I/Os to issue before switching a new file can be specified by
457appending `:\fIint\fR' to the service type.
458.RE
459.TP
460.BI ioengine \fR=\fPstr
461Defines how the job issues I/O. The following types are defined:
462.RS
463.RS
464.TP
465.B sync
ccc2b328 466Basic \fBread\fR\|(2) or \fBwrite\fR\|(2) I/O. \fBfseek\fR\|(2) is used to
d60e92d1
AC
467position the I/O location.
468.TP
a31041ea 469.B psync
ccc2b328 470Basic \fBpread\fR\|(2) or \fBpwrite\fR\|(2) I/O.
a31041ea 471.TP
9183788d 472.B vsync
ccc2b328 473Basic \fBreadv\fR\|(2) or \fBwritev\fR\|(2) I/O. Will emulate queuing by
cecbfd47 474coalescing adjacent IOs into a single submission.
9183788d 475.TP
a46c5e01 476.B pvsync
ccc2b328 477Basic \fBpreadv\fR\|(2) or \fBpwritev\fR\|(2) I/O.
a46c5e01 478.TP
d60e92d1 479.B libaio
de890a1e 480Linux native asynchronous I/O. This ioengine defines engine specific options.
d60e92d1
AC
481.TP
482.B posixaio
ccc2b328 483POSIX asynchronous I/O using \fBaio_read\fR\|(3) and \fBaio_write\fR\|(3).
03e20d68
BC
484.TP
485.B solarisaio
486Solaris native asynchronous I/O.
487.TP
488.B windowsaio
489Windows native asynchronous I/O.
d60e92d1
AC
490.TP
491.B mmap
ccc2b328
SW
492File is memory mapped with \fBmmap\fR\|(2) and data copied using
493\fBmemcpy\fR\|(3).
d60e92d1
AC
494.TP
495.B splice
ccc2b328 496\fBsplice\fR\|(2) is used to transfer the data and \fBvmsplice\fR\|(2) to
d1429b5c 497transfer data from user-space to the kernel.
d60e92d1
AC
498.TP
499.B syslet-rw
500Use the syslet system calls to make regular read/write asynchronous.
501.TP
502.B sg
503SCSI generic sg v3 I/O. May be either synchronous using the SG_IO ioctl, or if
ccc2b328
SW
504the target is an sg character device, we use \fBread\fR\|(2) and
505\fBwrite\fR\|(2) for asynchronous I/O.
d60e92d1
AC
506.TP
507.B null
508Doesn't transfer any data, just pretends to. Mainly used to exercise \fBfio\fR
509itself and for debugging and testing purposes.
510.TP
511.B net
de890a1e
SL
512Transfer over the network. The protocol to be used can be defined with the
513\fBprotocol\fR parameter. Depending on the protocol, \fBfilename\fR,
514\fBhostname\fR, \fBport\fR, or \fBlisten\fR must be specified.
515This ioengine defines engine specific options.
d60e92d1
AC
516.TP
517.B netsplice
ccc2b328 518Like \fBnet\fR, but uses \fBsplice\fR\|(2) and \fBvmsplice\fR\|(2) to map data
de890a1e 519and send/receive. This ioengine defines engine specific options.
d60e92d1 520.TP
53aec0a4 521.B cpuio
d60e92d1
AC
522Doesn't transfer any data, but burns CPU cycles according to \fBcpuload\fR and
523\fBcpucycles\fR parameters.
524.TP
525.B guasi
526The GUASI I/O engine is the Generic Userspace Asynchronous Syscall Interface
cecbfd47 527approach to asynchronous I/O.
d1429b5c
AC
528.br
529See <http://www.xmailserver.org/guasi\-lib.html>.
d60e92d1 530.TP
21b8aee8 531.B rdma
85286c5c
BVA
532The RDMA I/O engine supports both RDMA memory semantics (RDMA_WRITE/RDMA_READ)
533and channel semantics (Send/Recv) for the InfiniBand, RoCE and iWARP protocols.
21b8aee8 534.TP
d60e92d1
AC
535.B external
536Loads an external I/O engine object file. Append the engine filename as
537`:\fIenginepath\fR'.
d54fce84
DM
538.TP
539.B falloc
cecbfd47 540 IO engine that does regular linux native fallocate call to simulate data
d54fce84
DM
541transfer as fio ioengine
542.br
543 DDIR_READ does fallocate(,mode = FALLOC_FL_KEEP_SIZE,)
544.br
0981fd71 545 DIR_WRITE does fallocate(,mode = 0)
d54fce84
DM
546.br
547 DDIR_TRIM does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE)
548.TP
549.B e4defrag
550IO engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate defragment activity
551request to DDIR_WRITE event
d60e92d1 552.RE
595e1734 553.P
d60e92d1
AC
554.RE
555.TP
556.BI iodepth \fR=\fPint
8489dae4
SK
557Number of I/O units to keep in flight against the file. Note that increasing
558iodepth beyond 1 will not affect synchronous ioengines (except for small
ee72ca09
JA
559degress when verify_async is in use). Even async engines my impose OS
560restrictions causing the desired depth not to be achieved. This may happen on
561Linux when using libaio and not setting \fBdirect\fR=1, since buffered IO is
562not async on that OS. Keep an eye on the IO depth distribution in the
563fio output to verify that the achieved depth is as expected. Default: 1.
d60e92d1
AC
564.TP
565.BI iodepth_batch \fR=\fPint
566Number of I/Os to submit at once. Default: \fBiodepth\fR.
567.TP
3ce9dcaf
JA
568.BI iodepth_batch_complete \fR=\fPint
569This defines how many pieces of IO to retrieve at once. It defaults to 1 which
570 means that we'll ask for a minimum of 1 IO in the retrieval process from the
571kernel. The IO retrieval will go on until we hit the limit set by
572\fBiodepth_low\fR. If this variable is set to 0, then fio will always check for
573completed events before queuing more IO. This helps reduce IO latency, at the
574cost of more retrieval system calls.
575.TP
d60e92d1
AC
576.BI iodepth_low \fR=\fPint
577Low watermark indicating when to start filling the queue again. Default:
578\fBiodepth\fR.
579.TP
580.BI direct \fR=\fPbool
581If true, use non-buffered I/O (usually O_DIRECT). Default: false.
582.TP
d01612f3
CM
583.BI atomic \fR=\fPbool
584If value is true, attempt to use atomic direct IO. Atomic writes are guaranteed
585to be stable once acknowledged by the operating system. Only Linux supports
586O_ATOMIC right now.
587.TP
d60e92d1
AC
588.BI buffered \fR=\fPbool
589If true, use buffered I/O. This is the opposite of the \fBdirect\fR parameter.
590Default: true.
591.TP
f7fa2653 592.BI offset \fR=\fPint
d60e92d1
AC
593Offset in the file to start I/O. Data before the offset will not be touched.
594.TP
591e9e06
JA
595.BI offset_increment \fR=\fPint
596If this is provided, then the real offset becomes the
597offset + offset_increment * thread_number, where the thread number is a counter
598that starts at 0 and is incremented for each job. This option is useful if
599there are several jobs which are intended to operate on a file in parallel in
600disjoint segments, with even spacing between the starting points.
601.TP
ddf24e42
JA
602.BI number_ios \fR=\fPint
603Fio will normally perform IOs until it has exhausted the size of the region
604set by \fBsize\fR, or if it exhaust the allocated time (or hits an error
605condition). With this setting, the range/size can be set independently of
606the number of IOs to perform. When fio reaches this number, it will exit
607normally and report status.
608.TP
d60e92d1 609.BI fsync \fR=\fPint
d1429b5c
AC
610How many I/Os to perform before issuing an \fBfsync\fR\|(2) of dirty data. If
6110, don't sync. Default: 0.
d60e92d1 612.TP
5f9099ea
JA
613.BI fdatasync \fR=\fPint
614Like \fBfsync\fR, but uses \fBfdatasync\fR\|(2) instead to only sync the
615data parts of the file. Default: 0.
616.TP
fa769d44
SW
617.BI write_barrier \fR=\fPint
618Make every Nth write a barrier write.
619.TP
e76b1da4 620.BI sync_file_range \fR=\fPstr:int
ccc2b328
SW
621Use \fBsync_file_range\fR\|(2) for every \fRval\fP number of write operations. Fio will
622track range of writes that have happened since the last \fBsync_file_range\fR\|(2) call.
e76b1da4
JA
623\fRstr\fP can currently be one or more of:
624.RS
625.TP
626.B wait_before
627SYNC_FILE_RANGE_WAIT_BEFORE
628.TP
629.B write
630SYNC_FILE_RANGE_WRITE
631.TP
632.B wait_after
633SYNC_FILE_RANGE_WRITE
634.TP
635.RE
636.P
637So if you do sync_file_range=wait_before,write:8, fio would use
638\fBSYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE\fP for every 8 writes.
ccc2b328 639Also see the \fBsync_file_range\fR\|(2) man page. This option is Linux specific.
e76b1da4 640.TP
d60e92d1 641.BI overwrite \fR=\fPbool
d1429b5c 642If writing, setup the file first and do overwrites. Default: false.
d60e92d1
AC
643.TP
644.BI end_fsync \fR=\fPbool
dbd11ead 645Sync file contents when a write stage has completed. Default: false.
d60e92d1
AC
646.TP
647.BI fsync_on_close \fR=\fPbool
648If true, sync file contents on close. This differs from \fBend_fsync\fR in that
d1429b5c 649it will happen on every close, not just at the end of the job. Default: false.
d60e92d1 650.TP
d60e92d1
AC
651.BI rwmixread \fR=\fPint
652Percentage of a mixed workload that should be reads. Default: 50.
653.TP
654.BI rwmixwrite \fR=\fPint
d1429b5c 655Percentage of a mixed workload that should be writes. If \fBrwmixread\fR and
c35dd7a6
JA
656\fBrwmixwrite\fR are given and do not sum to 100%, the latter of the two
657overrides the first. This may interfere with a given rate setting, if fio is
658asked to limit reads or writes to a certain rate. If that is the case, then
659the distribution may be skewed. Default: 50.
d60e92d1 660.TP
92d42d69
JA
661.BI random_distribution \fR=\fPstr:float
662By default, fio will use a completely uniform random distribution when asked
663to perform random IO. Sometimes it is useful to skew the distribution in
664specific ways, ensuring that some parts of the data is more hot than others.
665Fio includes the following distribution models:
666.RS
667.TP
668.B random
669Uniform random distribution
670.TP
671.B zipf
672Zipf distribution
673.TP
674.B pareto
675Pareto distribution
676.TP
677.RE
678.P
679When using a zipf or pareto distribution, an input value is also needed to
680define the access pattern. For zipf, this is the zipf theta. For pareto,
681it's the pareto power. Fio includes a test program, genzipf, that can be
682used visualize what the given input values will yield in terms of hit rates.
683If you wanted to use zipf with a theta of 1.2, you would use
684random_distribution=zipf:1.2 as the option. If a non-uniform model is used,
685fio will disable use of the random map.
686.TP
211c9b89
JA
687.BI percentage_random \fR=\fPint
688For a random workload, set how big a percentage should be random. This defaults
689to 100%, in which case the workload is fully random. It can be set from
690anywhere from 0 to 100. Setting it to 0 would make the workload fully
d9472271
JA
691sequential. It is possible to set different values for reads, writes, and
692trim. To do so, simply use a comma separated list. See \fBblocksize\fR.
211c9b89 693.TP
d60e92d1
AC
694.B norandommap
695Normally \fBfio\fR will cover every block of the file when doing random I/O. If
696this parameter is given, a new offset will be chosen without looking at past
697I/O history. This parameter is mutually exclusive with \fBverify\fR.
698.TP
744492c9 699.BI softrandommap \fR=\fPbool
3ce9dcaf
JA
700See \fBnorandommap\fR. If fio runs with the random block map enabled and it
701fails to allocate the map, if this option is set it will continue without a
702random block map. As coverage will not be as complete as with random maps, this
703option is disabled by default.
704.TP
e8b1961d
JA
705.BI random_generator \fR=\fPstr
706Fio supports the following engines for generating IO offsets for random IO:
707.RS
708.TP
709.B tausworthe
710Strong 2^88 cycle random number generator
711.TP
712.B lfsr
713Linear feedback shift register generator
714.TP
715.RE
716.P
717Tausworthe is a strong random number generator, but it requires tracking on the
718side if we want to ensure that blocks are only read or written once. LFSR
719guarantees that we never generate the same offset twice, and it's also less
720computationally expensive. It's not a true random generator, however, though
721for IO purposes it's typically good enough. LFSR only works with single block
722sizes, not with workloads that use multiple block sizes. If used with such a
723workload, fio may read or write some blocks multiple times.
724.TP
d60e92d1 725.BI nice \fR=\fPint
ccc2b328 726Run job with given nice value. See \fBnice\fR\|(2).
d60e92d1
AC
727.TP
728.BI prio \fR=\fPint
729Set I/O priority value of this job between 0 (highest) and 7 (lowest). See
ccc2b328 730\fBionice\fR\|(1).
d60e92d1
AC
731.TP
732.BI prioclass \fR=\fPint
ccc2b328 733Set I/O priority class. See \fBionice\fR\|(1).
d60e92d1
AC
734.TP
735.BI thinktime \fR=\fPint
736Stall job for given number of microseconds between issuing I/Os.
737.TP
738.BI thinktime_spin \fR=\fPint
739Pretend to spend CPU time for given number of microseconds, sleeping the rest
740of the time specified by \fBthinktime\fR. Only valid if \fBthinktime\fR is set.
741.TP
742.BI thinktime_blocks \fR=\fPint
4d01ece6
JA
743Only valid if thinktime is set - control how many blocks to issue, before
744waiting \fBthinktime\fR microseconds. If not set, defaults to 1 which will
745make fio wait \fBthinktime\fR microseconds after every block. This
746effectively makes any queue depth setting redundant, since no more than 1 IO
747will be queued before we have to complete it and do our thinktime. In other
748words, this setting effectively caps the queue depth if the latter is larger.
d60e92d1
AC
749Default: 1.
750.TP
751.BI rate \fR=\fPint
c35dd7a6
JA
752Cap bandwidth used by this job. The number is in bytes/sec, the normal postfix
753rules apply. You can use \fBrate\fR=500k to limit reads and writes to 500k each,
754or you can specify read and writes separately. Using \fBrate\fR=1m,500k would
755limit reads to 1MB/sec and writes to 500KB/sec. Capping only reads or writes
756can be done with \fBrate\fR=,500k or \fBrate\fR=500k,. The former will only
757limit writes (to 500KB/sec), the latter will only limit reads.
d60e92d1
AC
758.TP
759.BI ratemin \fR=\fPint
760Tell \fBfio\fR to do whatever it can to maintain at least the given bandwidth.
c35dd7a6
JA
761Failing to meet this requirement will cause the job to exit. The same format
762as \fBrate\fR is used for read vs write separation.
d60e92d1
AC
763.TP
764.BI rate_iops \fR=\fPint
c35dd7a6
JA
765Cap the bandwidth to this number of IOPS. Basically the same as rate, just
766specified independently of bandwidth. The same format as \fBrate\fR is used for
de8f6de9 767read vs write separation. If \fBblocksize\fR is a range, the smallest block
c35dd7a6 768size is used as the metric.
d60e92d1
AC
769.TP
770.BI rate_iops_min \fR=\fPint
c35dd7a6 771If this rate of I/O is not met, the job will exit. The same format as \fBrate\fR
de8f6de9 772is used for read vs write separation.
d60e92d1
AC
773.TP
774.BI ratecycle \fR=\fPint
775Average bandwidth for \fBrate\fR and \fBratemin\fR over this number of
776milliseconds. Default: 1000ms.
777.TP
3e260a46
JA
778.BI latency_target \fR=\fPint
779If set, fio will attempt to find the max performance point that the given
780workload will run at while maintaining a latency below this target. The
781values is given in microseconds. See \fBlatency_window\fR and
782\fBlatency_percentile\fR.
783.TP
784.BI latency_window \fR=\fPint
785Used with \fBlatency_target\fR to specify the sample window that the job
786is run at varying queue depths to test the performance. The value is given
787in microseconds.
788.TP
789.BI latency_percentile \fR=\fPfloat
790The percentage of IOs that must fall within the criteria specified by
791\fBlatency_target\fR and \fBlatency_window\fR. If not set, this defaults
792to 100.0, meaning that all IOs must be equal or below to the value set
793by \fBlatency_target\fR.
794.TP
15501535
JA
795.BI max_latency \fR=\fPint
796If set, fio will exit the job if it exceeds this maximum latency. It will exit
797with an ETIME error.
798.TP
d60e92d1
AC
799.BI cpumask \fR=\fPint
800Set CPU affinity for this job. \fIint\fR is a bitmask of allowed CPUs the job
801may run on. See \fBsched_setaffinity\fR\|(2).
802.TP
803.BI cpus_allowed \fR=\fPstr
804Same as \fBcpumask\fR, but allows a comma-delimited list of CPU numbers.
805.TP
d0b937ed 806.BI numa_cpu_nodes \fR=\fPstr
cecbfd47 807Set this job running on specified NUMA nodes' CPUs. The arguments allow
d0b937ed
YR
808comma delimited list of cpu numbers, A-B ranges, or 'all'.
809.TP
810.BI numa_mem_policy \fR=\fPstr
811Set this job's memory policy and corresponding NUMA nodes. Format of
cecbfd47 812the arguments:
d0b937ed
YR
813.RS
814.TP
815.B <mode>[:<nodelist>]
816.TP
817.B mode
818is one of the following memory policy:
819.TP
820.B default, prefer, bind, interleave, local
821.TP
822.RE
823For \fBdefault\fR and \fBlocal\fR memory policy, no \fBnodelist\fR is
824needed to be specified. For \fBprefer\fR, only one node is
825allowed. For \fBbind\fR and \fBinterleave\fR, \fBnodelist\fR allows
826comma delimited list of numbers, A-B ranges, or 'all'.
827.TP
d60e92d1
AC
828.BI startdelay \fR=\fPint
829Delay start of job for the specified number of seconds.
830.TP
831.BI runtime \fR=\fPint
832Terminate processing after the specified number of seconds.
833.TP
834.B time_based
835If given, run for the specified \fBruntime\fR duration even if the files are
836completely read or written. The same workload will be repeated as many times
837as \fBruntime\fR allows.
838.TP
901bb994
JA
839.BI ramp_time \fR=\fPint
840If set, fio will run the specified workload for this amount of time before
841logging any performance numbers. Useful for letting performance settle before
842logging results, thus minimizing the runtime required for stable results. Note
c35dd7a6
JA
843that the \fBramp_time\fR is considered lead in time for a job, thus it will
844increase the total runtime if a special timeout or runtime is specified.
901bb994 845.TP
d60e92d1
AC
846.BI invalidate \fR=\fPbool
847Invalidate buffer-cache for the file prior to starting I/O. Default: true.
848.TP
849.BI sync \fR=\fPbool
850Use synchronous I/O for buffered writes. For the majority of I/O engines,
d1429b5c 851this means using O_SYNC. Default: false.
d60e92d1
AC
852.TP
853.BI iomem \fR=\fPstr "\fR,\fP mem" \fR=\fPstr
854Allocation method for I/O unit buffer. Allowed values are:
855.RS
856.RS
857.TP
858.B malloc
ccc2b328 859Allocate memory with \fBmalloc\fR\|(3).
d60e92d1
AC
860.TP
861.B shm
ccc2b328 862Use shared memory buffers allocated through \fBshmget\fR\|(2).
d60e92d1
AC
863.TP
864.B shmhuge
865Same as \fBshm\fR, but use huge pages as backing.
866.TP
867.B mmap
ccc2b328 868Use \fBmmap\fR\|(2) for allocation. Uses anonymous memory unless a filename
d60e92d1
AC
869is given after the option in the format `:\fIfile\fR'.
870.TP
871.B mmaphuge
872Same as \fBmmap\fR, but use huge files as backing.
873.RE
874.P
875The amount of memory allocated is the maximum allowed \fBblocksize\fR for the
876job multiplied by \fBiodepth\fR. For \fBshmhuge\fR or \fBmmaphuge\fR to work,
877the system must have free huge pages allocated. \fBmmaphuge\fR also needs to
2e266ba6
JA
878have hugetlbfs mounted, and \fIfile\fR must point there. At least on Linux,
879huge pages must be manually allocated. See \fB/proc/sys/vm/nr_hugehages\fR
880and the documentation for that. Normally you just need to echo an appropriate
881number, eg echoing 8 will ensure that the OS has 8 huge pages ready for
882use.
d60e92d1
AC
883.RE
884.TP
d392365e 885.BI iomem_align \fR=\fPint "\fR,\fP mem_align" \fR=\fPint
cecbfd47 886This indicates the memory alignment of the IO memory buffers. Note that the
d529ee19
JA
887given alignment is applied to the first IO unit buffer, if using \fBiodepth\fR
888the alignment of the following buffers are given by the \fBbs\fR used. In
889other words, if using a \fBbs\fR that is a multiple of the page sized in the
890system, all buffers will be aligned to this value. If using a \fBbs\fR that
891is not page aligned, the alignment of subsequent IO memory buffers is the
892sum of the \fBiomem_align\fR and \fBbs\fR used.
893.TP
f7fa2653 894.BI hugepage\-size \fR=\fPint
d60e92d1 895Defines the size of a huge page. Must be at least equal to the system setting.
b22989b9 896Should be a multiple of 1MB. Default: 4MB.
d60e92d1
AC
897.TP
898.B exitall
899Terminate all jobs when one finishes. Default: wait for each job to finish.
900.TP
901.BI bwavgtime \fR=\fPint
902Average bandwidth calculations over the given time in milliseconds. Default:
903500ms.
904.TP
c8eeb9df
JA
905.BI iopsavgtime \fR=\fPint
906Average IOPS calculations over the given time in milliseconds. Default:
907500ms.
908.TP
d60e92d1 909.BI create_serialize \fR=\fPbool
d1429b5c 910If true, serialize file creation for the jobs. Default: true.
d60e92d1
AC
911.TP
912.BI create_fsync \fR=\fPbool
ccc2b328 913\fBfsync\fR\|(2) data file after creation. Default: true.
d60e92d1 914.TP
6b7f6851
JA
915.BI create_on_open \fR=\fPbool
916If true, the files are not created until they are opened for IO by the job.
917.TP
25460cf6
JA
918.BI create_only \fR=\fPbool
919If true, fio will only run the setup phase of the job. If files need to be
920laid out or updated on disk, only that will be done. The actual job contents
921are not executed.
922.TP
e9f48479
JA
923.BI pre_read \fR=\fPbool
924If this is given, files will be pre-read into memory before starting the given
925IO operation. This will also clear the \fR \fBinvalidate\fR flag, since it is
9c0d2241
JA
926pointless to pre-read and then drop the cache. This will only work for IO
927engines that are seekable, since they allow you to read the same data
928multiple times. Thus it will not work on eg network or splice IO.
e9f48479 929.TP
d60e92d1
AC
930.BI unlink \fR=\fPbool
931Unlink job files when done. Default: false.
932.TP
933.BI loops \fR=\fPint
934Specifies the number of iterations (runs of the same workload) of this job.
935Default: 1.
936.TP
937.BI do_verify \fR=\fPbool
938Run the verify phase after a write phase. Only valid if \fBverify\fR is set.
939Default: true.
940.TP
941.BI verify \fR=\fPstr
942Method of verifying file contents after each iteration of the job. Allowed
943values are:
944.RS
945.RS
946.TP
b892dc08 947.B md5 crc16 crc32 crc32c crc32c-intel crc64 crc7 sha256 sha512 sha1
0539d758
JA
948Store appropriate checksum in the header of each block. crc32c-intel is
949hardware accelerated SSE4.2 driven, falls back to regular crc32c if
950not supported by the system.
d60e92d1
AC
951.TP
952.B meta
953Write extra information about each I/O (timestamp, block number, etc.). The
996093bb 954block number is verified. See \fBverify_pattern\fR as well.
d60e92d1
AC
955.TP
956.B null
957Pretend to verify. Used for testing internals.
958.RE
b892dc08
JA
959
960This option can be used for repeated burn-in tests of a system to make sure
961that the written data is also correctly read back. If the data direction given
962is a read or random read, fio will assume that it should verify a previously
963written file. If the data direction includes any form of write, the verify will
964be of the newly written data.
d60e92d1
AC
965.RE
966.TP
5c9323fb 967.BI verifysort \fR=\fPbool
d60e92d1
AC
968If true, written verify blocks are sorted if \fBfio\fR deems it to be faster to
969read them back in a sorted manner. Default: true.
970.TP
fa769d44
SW
971.BI verifysort_nr \fR=\fPint
972Pre-load and sort verify blocks for a read workload.
973.TP
f7fa2653 974.BI verify_offset \fR=\fPint
d60e92d1 975Swap the verification header with data somewhere else in the block before
d1429b5c 976writing. It is swapped back before verifying.
d60e92d1 977.TP
f7fa2653 978.BI verify_interval \fR=\fPint
d60e92d1
AC
979Write the verification header for this number of bytes, which should divide
980\fBblocksize\fR. Default: \fBblocksize\fR.
981.TP
996093bb
JA
982.BI verify_pattern \fR=\fPstr
983If set, fio will fill the io buffers with this pattern. Fio defaults to filling
984with totally random bytes, but sometimes it's interesting to fill with a known
985pattern for io verification purposes. Depending on the width of the pattern,
986fio will fill 1/2/3/4 bytes of the buffer at the time(it can be either a
987decimal or a hex number). The verify_pattern if larger than a 32-bit quantity
988has to be a hex number that starts with either "0x" or "0X". Use with
989\fBverify\fP=meta.
990.TP
d60e92d1
AC
991.BI verify_fatal \fR=\fPbool
992If true, exit the job on the first observed verification failure. Default:
993false.
994.TP
b463e936
JA
995.BI verify_dump \fR=\fPbool
996If set, dump the contents of both the original data block and the data block we
997read off disk to files. This allows later analysis to inspect just what kind of
ef71e317 998data corruption occurred. Off by default.
b463e936 999.TP
e8462bd8
JA
1000.BI verify_async \fR=\fPint
1001Fio will normally verify IO inline from the submitting thread. This option
1002takes an integer describing how many async offload threads to create for IO
1003verification instead, causing fio to offload the duty of verifying IO contents
c85c324c
JA
1004to one or more separate threads. If using this offload option, even sync IO
1005engines can benefit from using an \fBiodepth\fR setting higher than 1, as it
1006allows them to have IO in flight while verifies are running.
e8462bd8
JA
1007.TP
1008.BI verify_async_cpus \fR=\fPstr
1009Tell fio to set the given CPU affinity on the async IO verification threads.
1010See \fBcpus_allowed\fP for the format used.
1011.TP
6f87418f
JA
1012.BI verify_backlog \fR=\fPint
1013Fio will normally verify the written contents of a job that utilizes verify
1014once that job has completed. In other words, everything is written then
1015everything is read back and verified. You may want to verify continually
1016instead for a variety of reasons. Fio stores the meta data associated with an
1017IO block in memory, so for large verify workloads, quite a bit of memory would
092f707f
DN
1018be used up holding this meta data. If this option is enabled, fio will write
1019only N blocks before verifying these blocks.
6f87418f
JA
1020.TP
1021.BI verify_backlog_batch \fR=\fPint
1022Control how many blocks fio will verify if verify_backlog is set. If not set,
1023will default to the value of \fBverify_backlog\fR (meaning the entire queue is
092f707f
DN
1024read back and verified). If \fBverify_backlog_batch\fR is less than
1025\fBverify_backlog\fR then not all blocks will be verified, if
1026\fBverify_backlog_batch\fR is larger than \fBverify_backlog\fR, some blocks
1027will be verified more than once.
6f87418f 1028.TP
fa769d44
SW
1029.BI trim_percentage \fR=\fPint
1030Number of verify blocks to discard/trim.
1031.TP
1032.BI trim_verify_zero \fR=\fPbool
1033Verify that trim/discarded blocks are returned as zeroes.
1034.TP
1035.BI trim_backlog \fR=\fPint
1036Trim after this number of blocks are written.
1037.TP
1038.BI trim_backlog_batch \fR=\fPint
1039Trim this number of IO blocks.
1040.TP
1041.BI experimental_verify \fR=\fPbool
1042Enable experimental verification.
1043.TP
d392365e 1044.B stonewall "\fR,\fP wait_for_previous"
5982a925 1045Wait for preceding jobs in the job file to exit before starting this one.
d60e92d1
AC
1046\fBstonewall\fR implies \fBnew_group\fR.
1047.TP
1048.B new_group
1049Start a new reporting group. If not given, all jobs in a file will be part
1050of the same reporting group, unless separated by a stonewall.
1051.TP
1052.BI numjobs \fR=\fPint
1053Number of clones (processes/threads performing the same workload) of this job.
1054Default: 1.
1055.TP
1056.B group_reporting
1057If set, display per-group reports instead of per-job when \fBnumjobs\fR is
1058specified.
1059.TP
1060.B thread
1061Use threads created with \fBpthread_create\fR\|(3) instead of processes created
1062with \fBfork\fR\|(2).
1063.TP
f7fa2653 1064.BI zonesize \fR=\fPint
d60e92d1
AC
1065Divide file into zones of the specified size in bytes. See \fBzoneskip\fR.
1066.TP
fa769d44
SW
1067.BI zonerange \fR=\fPint
1068Give size of an IO zone. See \fBzoneskip\fR.
1069.TP
f7fa2653 1070.BI zoneskip \fR=\fPint
d1429b5c 1071Skip the specified number of bytes when \fBzonesize\fR bytes of data have been
d60e92d1
AC
1072read.
1073.TP
1074.BI write_iolog \fR=\fPstr
5b42a488
SH
1075Write the issued I/O patterns to the specified file. Specify a separate file
1076for each job, otherwise the iologs will be interspersed and the file may be
1077corrupt.
d60e92d1
AC
1078.TP
1079.BI read_iolog \fR=\fPstr
1080Replay the I/O patterns contained in the specified file generated by
1081\fBwrite_iolog\fR, or may be a \fBblktrace\fR binary file.
1082.TP
64bbb865
DN
1083.BI replay_no_stall \fR=\fPint
1084While replaying I/O patterns using \fBread_iolog\fR the default behavior
1085attempts to respect timing information between I/Os. Enabling
1086\fBreplay_no_stall\fR causes I/Os to be replayed as fast as possible while
1087still respecting ordering.
1088.TP
d1c46c04
DN
1089.BI replay_redirect \fR=\fPstr
1090While replaying I/O patterns using \fBread_iolog\fR the default behavior
1091is to replay the IOPS onto the major/minor device that each IOP was recorded
1092from. Setting \fBreplay_redirect\fR causes all IOPS to be replayed onto the
1093single specified device regardless of the device it was recorded from.
1094.TP
836bad52 1095.BI write_bw_log \fR=\fPstr
901bb994
JA
1096If given, write a bandwidth log of the jobs in this job file. Can be used to
1097store data of the bandwidth of the jobs in their lifetime. The included
1098fio_generate_plots script uses gnuplot to turn these text files into nice
26b26fca 1099graphs. See \fBwrite_lat_log\fR for behaviour of given filename. For this
901bb994 1100option, the postfix is _bw.log.
d60e92d1 1101.TP
836bad52 1102.BI write_lat_log \fR=\fPstr
901bb994
JA
1103Same as \fBwrite_bw_log\fR, but writes I/O completion latencies. If no
1104filename is given with this option, the default filename of "jobname_type.log"
1105is used. Even if the filename is given, fio will still append the type of log.
1106.TP
c8eeb9df
JA
1107.BI write_iops_log \fR=\fPstr
1108Same as \fBwrite_bw_log\fR, but writes IOPS. If no filename is given with this
1109option, the default filename of "jobname_type.log" is used. Even if the
1110filename is given, fio will still append the type of log.
1111.TP
b8bc8cba
JA
1112.BI log_avg_msec \fR=\fPint
1113By default, fio will log an entry in the iops, latency, or bw log for every
1114IO that completes. When writing to the disk log, that can quickly grow to a
1115very large size. Setting this option makes fio average the each log entry
1116over the specified period of time, reducing the resolution of the log.
1117Defaults to 0.
1118.TP
836bad52 1119.BI disable_lat \fR=\fPbool
02af0988 1120Disable measurements of total latency numbers. Useful only for cutting
ccc2b328 1121back the number of calls to \fBgettimeofday\fR\|(2), as that does impact performance at
901bb994
JA
1122really high IOPS rates. Note that to really get rid of a large amount of these
1123calls, this option must be used with disable_slat and disable_bw as well.
1124.TP
836bad52 1125.BI disable_clat \fR=\fPbool
c95f9daf 1126Disable measurements of completion latency numbers. See \fBdisable_lat\fR.
02af0988 1127.TP
836bad52 1128.BI disable_slat \fR=\fPbool
02af0988 1129Disable measurements of submission latency numbers. See \fBdisable_lat\fR.
901bb994 1130.TP
836bad52 1131.BI disable_bw_measurement \fR=\fPbool
02af0988 1132Disable measurements of throughput/bandwidth numbers. See \fBdisable_lat\fR.
d60e92d1 1133.TP
f7fa2653 1134.BI lockmem \fR=\fPint
d60e92d1 1135Pin the specified amount of memory with \fBmlock\fR\|(2). Can be used to
81c6b6cd 1136simulate a smaller amount of memory. The amount specified is per worker.
d60e92d1
AC
1137.TP
1138.BI exec_prerun \fR=\fPstr
1139Before running the job, execute the specified command with \fBsystem\fR\|(3).
ce486495
EV
1140.RS
1141Output is redirected in a file called \fBjobname.prerun.txt\fR
1142.RE
d60e92d1
AC
1143.TP
1144.BI exec_postrun \fR=\fPstr
1145Same as \fBexec_prerun\fR, but the command is executed after the job completes.
ce486495
EV
1146.RS
1147Output is redirected in a file called \fBjobname.postrun.txt\fR
1148.RE
d60e92d1
AC
1149.TP
1150.BI ioscheduler \fR=\fPstr
1151Attempt to switch the device hosting the file to the specified I/O scheduler.
1152.TP
1153.BI cpuload \fR=\fPint
1154If the job is a CPU cycle-eater, attempt to use the specified percentage of
1155CPU cycles.
1156.TP
1157.BI cpuchunks \fR=\fPint
1158If the job is a CPU cycle-eater, split the load into cycles of the
1159given time in milliseconds.
1160.TP
1161.BI disk_util \fR=\fPbool
d1429b5c 1162Generate disk utilization statistics if the platform supports it. Default: true.
901bb994 1163.TP
23893646
JA
1164.BI clocksource \fR=\fPstr
1165Use the given clocksource as the base of timing. The supported options are:
1166.RS
1167.TP
1168.B gettimeofday
ccc2b328 1169\fBgettimeofday\fR\|(2)
23893646
JA
1170.TP
1171.B clock_gettime
ccc2b328 1172\fBclock_gettime\fR\|(2)
23893646
JA
1173.TP
1174.B cpu
1175Internal CPU clock source
1176.TP
1177.RE
1178.P
1179\fBcpu\fR is the preferred clocksource if it is reliable, as it is very fast
1180(and fio is heavy on time calls). Fio will automatically use this clocksource
1181if it's supported and considered reliable on the system it is running on,
1182unless another clocksource is specifically set. For x86/x86-64 CPUs, this
1183means supporting TSC Invariant.
1184.TP
901bb994 1185.BI gtod_reduce \fR=\fPbool
ccc2b328 1186Enable all of the \fBgettimeofday\fR\|(2) reducing options (disable_clat, disable_slat,
901bb994 1187disable_bw) plus reduce precision of the timeout somewhat to really shrink the
ccc2b328 1188\fBgettimeofday\fR\|(2) call count. With this option enabled, we only do about 0.4% of
901bb994
JA
1189the gtod() calls we would have done if all time keeping was enabled.
1190.TP
1191.BI gtod_cpu \fR=\fPint
1192Sometimes it's cheaper to dedicate a single thread of execution to just getting
1193the current time. Fio (and databases, for instance) are very intensive on
ccc2b328 1194\fBgettimeofday\fR\|(2) calls. With this option, you can set one CPU aside for doing
901bb994
JA
1195nothing but logging current time to a shared memory location. Then the other
1196threads/processes that run IO workloads need only copy that segment, instead of
ccc2b328 1197entering the kernel with a \fBgettimeofday\fR\|(2) call. The CPU set aside for doing
901bb994
JA
1198these time calls will be excluded from other uses. Fio will manually clear it
1199from the CPU mask of other jobs.
f2bba182 1200.TP
8b28bd41
DM
1201.BI ignore_error \fR=\fPstr
1202Sometimes you want to ignore some errors during test in that case you can specify
1203error list for each error type.
1204.br
1205ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST
1206.br
1207errors for given error type is separated with ':'.
1208Error may be symbol ('ENOSPC', 'ENOMEM') or an integer.
1209.br
1210Example: ignore_error=EAGAIN,ENOSPC:122 .
1211.br
1212This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from WRITE.
1213.TP
1214.BI error_dump \fR=\fPbool
1215If set dump every error even if it is non fatal, true by default. If disabled
1216only fatal error will be dumped
1217.TP
fa769d44
SW
1218.BI profile \fR=\fPstr
1219Select a specific builtin performance test.
1220.TP
a696fa2a
JA
1221.BI cgroup \fR=\fPstr
1222Add job to this control group. If it doesn't exist, it will be created.
6adb38a1
JA
1223The system must have a mounted cgroup blkio mount point for this to work. If
1224your system doesn't have it mounted, you can do so with:
1225
5982a925 1226# mount \-t cgroup \-o blkio none /cgroup
a696fa2a
JA
1227.TP
1228.BI cgroup_weight \fR=\fPint
1229Set the weight of the cgroup to this value. See the documentation that comes
1230with the kernel, allowed values are in the range of 100..1000.
e0b0d892 1231.TP
7de87099
VG
1232.BI cgroup_nodelete \fR=\fPbool
1233Normally fio will delete the cgroups it has created after the job completion.
1234To override this behavior and to leave cgroups around after the job completion,
1235set cgroup_nodelete=1. This can be useful if one wants to inspect various
1236cgroup files after job completion. Default: false
1237.TP
e0b0d892
JA
1238.BI uid \fR=\fPint
1239Instead of running as the invoking user, set the user ID to this value before
1240the thread/process does any work.
1241.TP
1242.BI gid \fR=\fPint
1243Set group ID, see \fBuid\fR.
83349190 1244.TP
fa769d44
SW
1245.BI unit_base \fR=\fPint
1246Base unit for reporting. Allowed values are:
1247.RS
1248.TP
1249.B 0
1250Use auto-detection (default).
1251.TP
1252.B 8
1253Byte based.
1254.TP
1255.B 1
1256Bit based.
1257.RE
1258.P
1259.TP
9e684a49
DE
1260.BI flow_id \fR=\fPint
1261The ID of the flow. If not specified, it defaults to being a global flow. See
1262\fBflow\fR.
1263.TP
1264.BI flow \fR=\fPint
1265Weight in token-based flow control. If this value is used, then there is a
1266\fBflow counter\fR which is used to regulate the proportion of activity between
1267two or more jobs. fio attempts to keep this flow counter near zero. The
1268\fBflow\fR parameter stands for how much should be added or subtracted to the
1269flow counter on each iteration of the main I/O loop. That is, if one job has
1270\fBflow=8\fR and another job has \fBflow=-1\fR, then there will be a roughly
12711:8 ratio in how much one runs vs the other.
1272.TP
1273.BI flow_watermark \fR=\fPint
1274The maximum value that the absolute value of the flow counter is allowed to
1275reach before the job must wait for a lower value of the counter.
1276.TP
1277.BI flow_sleep \fR=\fPint
1278The period of time, in microseconds, to wait after the flow watermark has been
1279exceeded before retrying operations
1280.TP
83349190
YH
1281.BI clat_percentiles \fR=\fPbool
1282Enable the reporting of percentiles of completion latencies.
1283.TP
1284.BI percentile_list \fR=\fPfloat_list
1285Overwrite the default list of percentiles for completion
1286latencies. Each number is a floating number in the range (0,100], and
1287the maximum length of the list is 20. Use ':' to separate the
3eb07285 1288numbers. For example, \-\-percentile_list=99.5:99.9 will cause fio to
83349190
YH
1289report the values of completion latency below which 99.5% and 99.9% of
1290the observed latencies fell, respectively.
de890a1e
SL
1291.SS "Ioengine Parameters List"
1292Some parameters are only valid when a specific ioengine is in use. These are
1293used identically to normal parameters, with the caveat that when used on the
1294command line, the must come after the ioengine that defines them is selected.
1295.TP
e4585935
JA
1296.BI (cpu)cpuload \fR=\fPint
1297Attempt to use the specified percentage of CPU cycles.
1298.TP
1299.BI (cpu)cpuchunks \fR=\fPint
1300Split the load into cycles of the given time. In microseconds.
1301.TP
de890a1e
SL
1302.BI (libaio)userspace_reap
1303Normally, with the libaio engine in use, fio will use
1304the io_getevents system call to reap newly returned events.
1305With this flag turned on, the AIO ring will be read directly
1306from user-space to reap events. The reaping mode is only
1307enabled when polling for a minimum of 0 events (eg when
1308iodepth_batch_complete=0).
1309.TP
1310.BI (net,netsplice)hostname \fR=\fPstr
1311The host name or IP address to use for TCP or UDP based IO.
1312If the job is a TCP listener or UDP reader, the hostname is not
b511c9aa 1313used and must be omitted unless it is a valid UDP multicast address.
de890a1e
SL
1314.TP
1315.BI (net,netsplice)port \fR=\fPint
1316The TCP or UDP port to bind to or connect to.
1317.TP
b93b6a2e
SB
1318.BI (net,netsplice)interface \fR=\fPstr
1319The IP address of the network interface used to send or receive UDP multicast
1320packets.
1321.TP
d3a623de
SB
1322.BI (net,netsplice)ttl \fR=\fPint
1323Time-to-live value for outgoing UDP multicast packets. Default: 1
1324.TP
1d360ffb
JA
1325.BI (net,netsplice)nodelay \fR=\fPbool
1326Set TCP_NODELAY on TCP connections.
1327.TP
de890a1e
SL
1328.BI (net,netsplice)protocol \fR=\fPstr "\fR,\fP proto" \fR=\fPstr
1329The network protocol to use. Accepted values are:
1330.RS
1331.RS
1332.TP
1333.B tcp
1334Transmission control protocol
1335.TP
1336.B udp
f5cc3d0e 1337User datagram protocol
de890a1e
SL
1338.TP
1339.B unix
1340UNIX domain socket
1341.RE
1342.P
1343When the protocol is TCP or UDP, the port must also be given,
1344as well as the hostname if the job is a TCP listener or UDP
1345reader. For unix sockets, the normal filename option should be
1346used and the port is invalid.
1347.RE
1348.TP
1349.BI (net,netsplice)listen
1350For TCP network connections, tell fio to listen for incoming
1351connections rather than initiating an outgoing connection. The
1352hostname must be omitted if this option is used.
d54fce84 1353.TP
7aeb1e94 1354.BI (net, pingpong) \fR=\fPbool
cecbfd47 1355Normally a network writer will just continue writing data, and a network reader
7aeb1e94
JA
1356will just consume packages. If pingpong=1 is set, a writer will send its normal
1357payload to the reader, then wait for the reader to send the same payload back.
1358This allows fio to measure network latencies. The submission and completion
1359latencies then measure local time spent sending or receiving, and the
1360completion latency measures how long it took for the other end to receive and
b511c9aa
SB
1361send back. For UDP multicast traffic pingpong=1 should only be set for a single
1362reader when multiple readers are listening to the same address.
7aeb1e94 1363.TP
d54fce84
DM
1364.BI (e4defrag,donorname) \fR=\fPstr
1365File will be used as a block donor (swap extents between files)
1366.TP
1367.BI (e4defrag,inplace) \fR=\fPint
1368Configure donor file block allocation strategy
1369.RS
1370.BI 0(default) :
1371Preallocate donor's file on init
1372.TP
1373.BI 1:
cecbfd47 1374allocate space immediately inside defragment event, and free right after event
d54fce84 1375.RE
d60e92d1 1376.SH OUTPUT
d1429b5c
AC
1377While running, \fBfio\fR will display the status of the created jobs. For
1378example:
d60e92d1 1379.RS
d1429b5c 1380.P
d60e92d1
AC
1381Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
1382.RE
1383.P
d1429b5c
AC
1384The characters in the first set of brackets denote the current status of each
1385threads. The possible values are:
1386.P
1387.PD 0
d60e92d1
AC
1388.RS
1389.TP
1390.B P
1391Setup but not started.
1392.TP
1393.B C
1394Thread created.
1395.TP
1396.B I
1397Initialized, waiting.
1398.TP
1399.B R
1400Running, doing sequential reads.
1401.TP
1402.B r
1403Running, doing random reads.
1404.TP
1405.B W
1406Running, doing sequential writes.
1407.TP
1408.B w
1409Running, doing random writes.
1410.TP
1411.B M
1412Running, doing mixed sequential reads/writes.
1413.TP
1414.B m
1415Running, doing mixed random reads/writes.
1416.TP
1417.B F
1418Running, currently waiting for \fBfsync\fR\|(2).
1419.TP
1420.B V
1421Running, verifying written data.
1422.TP
1423.B E
1424Exited, not reaped by main thread.
1425.TP
1426.B \-
1427Exited, thread reaped.
1428.RE
d1429b5c 1429.PD
d60e92d1
AC
1430.P
1431The second set of brackets shows the estimated completion percentage of
1432the current group. The third set shows the read and write I/O rate,
1433respectively. Finally, the estimated run time of the job is displayed.
1434.P
1435When \fBfio\fR completes (or is interrupted by Ctrl-C), it will show data
1436for each thread, each group of threads, and each disk, in that order.
1437.P
1438Per-thread statistics first show the threads client number, group-id, and
1439error code. The remaining figures are as follows:
1440.RS
d60e92d1
AC
1441.TP
1442.B io
1443Number of megabytes of I/O performed.
1444.TP
1445.B bw
1446Average data rate (bandwidth).
1447.TP
1448.B runt
1449Threads run time.
1450.TP
1451.B slat
1452Submission latency minimum, maximum, average and standard deviation. This is
1453the time it took to submit the I/O.
1454.TP
1455.B clat
1456Completion latency minimum, maximum, average and standard deviation. This
1457is the time between submission and completion.
1458.TP
1459.B bw
1460Bandwidth minimum, maximum, percentage of aggregate bandwidth received, average
1461and standard deviation.
1462.TP
1463.B cpu
1464CPU usage statistics. Includes user and system time, number of context switches
1465this thread went through and number of major and minor page faults.
1466.TP
1467.B IO depths
1468Distribution of I/O depths. Each depth includes everything less than (or equal)
1469to it, but greater than the previous depth.
1470.TP
1471.B IO issued
1472Number of read/write requests issued, and number of short read/write requests.
1473.TP
1474.B IO latencies
1475Distribution of I/O completion latencies. The numbers follow the same pattern
1476as \fBIO depths\fR.
1477.RE
d60e92d1
AC
1478.P
1479The group statistics show:
d1429b5c 1480.PD 0
d60e92d1
AC
1481.RS
1482.TP
1483.B io
1484Number of megabytes I/O performed.
1485.TP
1486.B aggrb
1487Aggregate bandwidth of threads in the group.
1488.TP
1489.B minb
1490Minimum average bandwidth a thread saw.
1491.TP
1492.B maxb
1493Maximum average bandwidth a thread saw.
1494.TP
1495.B mint
d1429b5c 1496Shortest runtime of threads in the group.
d60e92d1
AC
1497.TP
1498.B maxt
1499Longest runtime of threads in the group.
1500.RE
d1429b5c 1501.PD
d60e92d1
AC
1502.P
1503Finally, disk statistics are printed with reads first:
d1429b5c 1504.PD 0
d60e92d1
AC
1505.RS
1506.TP
1507.B ios
1508Number of I/Os performed by all groups.
1509.TP
1510.B merge
1511Number of merges in the I/O scheduler.
1512.TP
1513.B ticks
1514Number of ticks we kept the disk busy.
1515.TP
1516.B io_queue
1517Total time spent in the disk queue.
1518.TP
1519.B util
1520Disk utilization.
1521.RE
d1429b5c 1522.PD
8423bd11
JA
1523.P
1524It is also possible to get fio to dump the current output while it is
1525running, without terminating the job. To do that, send fio the \fBUSR1\fR
1526signal.
d60e92d1
AC
1527.SH TERSE OUTPUT
1528If the \fB\-\-minimal\fR option is given, the results will be printed in a
562c2d2f
DN
1529semicolon-delimited format suitable for scripted use - a job description
1530(if provided) follows on a new line. Note that the first
525c2bfa
JA
1531number in the line is the version number. If the output has to be changed
1532for some reason, this number will be incremented by 1 to signify that
1533change. The fields are:
d60e92d1
AC
1534.P
1535.RS
5e726d0a 1536.B terse version, fio version, jobname, groupid, error
d60e92d1
AC
1537.P
1538Read status:
1539.RS
312b4af2 1540.B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, IOPS, runtime \fR(ms)\fP
d60e92d1
AC
1541.P
1542Submission latency:
1543.RS
1544.B min, max, mean, standard deviation
1545.RE
1546Completion latency:
1547.RS
1548.B min, max, mean, standard deviation
1549.RE
1db92cb6
JA
1550Completion latency percentiles (20 fields):
1551.RS
1552.B Xth percentile=usec
1553.RE
525c2bfa
JA
1554Total latency:
1555.RS
1556.B min, max, mean, standard deviation
1557.RE
d60e92d1
AC
1558Bandwidth:
1559.RS
1560.B min, max, aggregate percentage of total, mean, standard deviation
1561.RE
1562.RE
1563.P
1564Write status:
1565.RS
312b4af2 1566.B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, IOPS, runtime \fR(ms)\fP
d60e92d1
AC
1567.P
1568Submission latency:
1569.RS
1570.B min, max, mean, standard deviation
1571.RE
1572Completion latency:
1573.RS
1574.B min, max, mean, standard deviation
1575.RE
1db92cb6
JA
1576Completion latency percentiles (20 fields):
1577.RS
1578.B Xth percentile=usec
1579.RE
525c2bfa
JA
1580Total latency:
1581.RS
1582.B min, max, mean, standard deviation
1583.RE
d60e92d1
AC
1584Bandwidth:
1585.RS
1586.B min, max, aggregate percentage of total, mean, standard deviation
1587.RE
1588.RE
1589.P
d1429b5c 1590CPU usage:
d60e92d1 1591.RS
bd2626f0 1592.B user, system, context switches, major page faults, minor page faults
d60e92d1
AC
1593.RE
1594.P
1595IO depth distribution:
1596.RS
1597.B <=1, 2, 4, 8, 16, 32, >=64
1598.RE
1599.P
562c2d2f 1600IO latency distribution:
d60e92d1 1601.RS
562c2d2f
DN
1602Microseconds:
1603.RS
1604.B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
1605.RE
1606Milliseconds:
1607.RS
1608.B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
1609.RE
1610.RE
1611.P
f2f788dd
JA
1612Disk utilization (1 for each disk used):
1613.RS
1614.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
1615.RE
1616.P
5982a925 1617Error Info (dependent on continue_on_error, default off):
562c2d2f
DN
1618.RS
1619.B total # errors, first error code
d60e92d1
AC
1620.RE
1621.P
562c2d2f 1622.B text description (if provided in config - appears on newline)
d60e92d1 1623.RE
49da1240
JA
1624.SH CLIENT / SERVER
1625Normally you would run fio as a stand-alone application on the machine
1626where the IO workload should be generated. However, it is also possible to
1627run the frontend and backend of fio separately. This makes it possible to
1628have a fio server running on the machine(s) where the IO workload should
1629be running, while controlling it from another machine.
1630
1631To start the server, you would do:
1632
1633\fBfio \-\-server=args\fR
1634
1635on that machine, where args defines what fio listens to. The arguments
811826be 1636are of the form 'type:hostname or IP:port'. 'type' is either 'ip' (or ip4)
20c67f10
MS
1637for TCP/IP v4, 'ip6' for TCP/IP v6, or 'sock' for a local unix domain
1638socket. 'hostname' is either a hostname or IP address, and 'port' is the port to
811826be 1639listen to (only valid for TCP/IP, not a local socket). Some examples:
49da1240 1640
e01e9745 16411) fio \-\-server
49da1240
JA
1642
1643 Start a fio server, listening on all interfaces on the default port (8765).
1644
e01e9745 16452) fio \-\-server=ip:hostname,4444
49da1240
JA
1646
1647 Start a fio server, listening on IP belonging to hostname and on port 4444.
1648
e01e9745 16493) fio \-\-server=ip6:::1,4444
811826be
JA
1650
1651 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
1652
e01e9745 16534) fio \-\-server=,4444
49da1240
JA
1654
1655 Start a fio server, listening on all interfaces on port 4444.
1656
e01e9745 16575) fio \-\-server=1.2.3.4
49da1240
JA
1658
1659 Start a fio server, listening on IP 1.2.3.4 on the default port.
1660
e01e9745 16616) fio \-\-server=sock:/tmp/fio.sock
49da1240
JA
1662
1663 Start a fio server, listening on the local socket /tmp/fio.sock.
1664
1665When a server is running, you can connect to it from a client. The client
1666is run with:
1667
e01e9745 1668fio \-\-local-args \-\-client=server \-\-remote-args <job file(s)>
49da1240 1669
e01e9745
MS
1670where \-\-local-args are arguments that are local to the client where it is
1671running, 'server' is the connect string, and \-\-remote-args and <job file(s)>
49da1240
JA
1672are sent to the server. The 'server' string follows the same format as it
1673does on the server side, to allow IP/hostname/socket and port strings.
1674You can connect to multiple clients as well, to do that you could run:
1675
e01e9745 1676fio \-\-client=server2 \-\-client=server2 <job file(s)>
d60e92d1 1677.SH AUTHORS
49da1240 1678
d60e92d1 1679.B fio
aa58d252
JA
1680was written by Jens Axboe <jens.axboe@oracle.com>,
1681now Jens Axboe <jaxboe@fusionio.com>.
d1429b5c
AC
1682.br
1683This man page was written by Aaron Carroll <aaronc@cse.unsw.edu.au> based
d60e92d1
AC
1684on documentation by Jens Axboe.
1685.SH "REPORTING BUGS"
482900c9 1686Report bugs to the \fBfio\fR mailing list <fio@vger.kernel.org>.
d1429b5c 1687See \fBREADME\fR.
d60e92d1 1688.SH "SEE ALSO"
d1429b5c
AC
1689For further documentation see \fBHOWTO\fR and \fBREADME\fR.
1690.br
1691Sample jobfiles are available in the \fBexamples\fR directory.
d60e92d1 1692