Commit | Line | Data |
---|---|---|
523bad63 | 1 | .TH fio 1 "August 2017" "User Manual" |
d60e92d1 AC |
2 | .SH NAME |
3 | fio \- flexible I/O tester | |
4 | .SH SYNOPSIS | |
5 | .B fio | |
6 | [\fIoptions\fR] [\fIjobfile\fR]... | |
7 | .SH DESCRIPTION | |
8 | .B fio | |
9 | is a tool that will spawn a number of threads or processes doing a | |
10 | particular type of I/O action as specified by the user. | |
11 | The typical use of fio is to write a job file matching the I/O load | |
12 | one wants to simulate. | |
13 | .SH OPTIONS | |
14 | .TP | |
49da1240 | 15 | .BI \-\-debug \fR=\fPtype |
7db7a5a0 TK |
16 | Enable verbose tracing \fItype\fR of various fio actions. May be `all' for all \fItype\fRs |
17 | or individual types separated by a comma (e.g. `\-\-debug=file,mem' will enable | |
bdd88be3 TK |
18 | file and memory debugging). `help' will list all available tracing options. |
19 | .TP | |
7db7a5a0 | 20 | .BI \-\-parse\-only |
bdd88be3 | 21 | Parse options only, don't start any I/O. |
49da1240 | 22 | .TP |
b9921d1a DZ |
23 | .BI \-\-merge\-blktrace\-only |
24 | Merge blktraces only, don't start any I/O. | |
25 | .TP | |
d60e92d1 AC |
26 | .BI \-\-output \fR=\fPfilename |
27 | Write output to \fIfilename\fR. | |
28 | .TP | |
7db7a5a0 TK |
29 | .BI \-\-output\-format \fR=\fPformat |
30 | Set the reporting \fIformat\fR to `normal', `terse', `json', or | |
31 | `json+'. Multiple formats can be selected, separate by a comma. `terse' | |
32 | is a CSV based format. `json+' is like `json', except it adds a full | |
513e37ee | 33 | dump of the latency buckets. |
e28ee21d | 34 | .TP |
7db7a5a0 | 35 | .BI \-\-bandwidth\-log |
d23ae827 | 36 | Generate aggregate bandwidth logs. |
d60e92d1 | 37 | .TP |
7db7a5a0 TK |
38 | .BI \-\-minimal |
39 | Print statistics in a terse, semicolon\-delimited format. | |
d60e92d1 | 40 | .TP |
7db7a5a0 TK |
41 | .BI \-\-append\-terse |
42 | Print statistics in selected mode AND terse, semicolon\-delimited format. | |
43 | \fBDeprecated\fR, use \fB\-\-output\-format\fR instead to select multiple formats. | |
f6a7df53 | 44 | .TP |
065248bf | 45 | .BI \-\-terse\-version \fR=\fPversion |
7db7a5a0 | 46 | Set terse \fIversion\fR output format (default `3', or `2', `4', `5'). |
49da1240 | 47 | .TP |
7db7a5a0 | 48 | .BI \-\-version |
bdd88be3 TK |
49 | Print version information and exit. |
50 | .TP | |
7db7a5a0 | 51 | .BI \-\-help |
bdd88be3 | 52 | Print a summary of the command line options and exit. |
49da1240 | 53 | .TP |
7db7a5a0 | 54 | .BI \-\-cpuclock\-test |
bdd88be3 | 55 | Perform test and validation of internal CPU clock. |
fec0f21c | 56 | .TP |
bdd88be3 | 57 | .BI \-\-crctest \fR=\fP[test] |
7db7a5a0 | 58 | Test the speed of the built\-in checksumming functions. If no argument is given, |
bdd88be3 | 59 | all of them are tested. Alternatively, a comma separated list can be passed, in which |
fec0f21c JA |
60 | case the given ones are tested. |
61 | .TP | |
49da1240 | 62 | .BI \-\-cmdhelp \fR=\fPcommand |
bdd88be3 | 63 | Print help information for \fIcommand\fR. May be `all' for all commands. |
49da1240 | 64 | .TP |
7db7a5a0 TK |
65 | .BI \-\-enghelp \fR=\fP[ioengine[,command]] |
66 | List all commands defined by \fIioengine\fR, or print help for \fIcommand\fR | |
67 | defined by \fIioengine\fR. If no \fIioengine\fR is given, list all | |
68 | available ioengines. | |
de890a1e | 69 | .TP |
d60e92d1 | 70 | .BI \-\-showcmd \fR=\fPjobfile |
7db7a5a0 | 71 | Convert \fIjobfile\fR to a set of command\-line options. |
d60e92d1 | 72 | .TP |
bdd88be3 | 73 | .BI \-\-readonly |
4027b2a1 | 74 | Turn on safety read\-only checks, preventing writes and trims. The \fB\-\-readonly\fR |
bdd88be3 | 75 | option is an extra safety guard to prevent users from accidentally starting |
4027b2a1 VF |
76 | a write or trim workload when that is not desired. Fio will only modify the |
77 | device under test if `rw=write/randwrite/rw/randrw/trim/randtrim/trimwrite' | |
78 | is given. This safety net can be used as an extra precaution. | |
bdd88be3 | 79 | .TP |
d60e92d1 | 80 | .BI \-\-eta \fR=\fPwhen |
7db7a5a0 | 81 | Specifies when real\-time ETA estimate should be printed. \fIwhen\fR may |
db37d890 JA |
82 | be `always', `never' or `auto'. `auto' is the default, it prints ETA when |
83 | requested if the output is a TTY. `always' disregards the output type, and | |
84 | prints ETA when requested. `never' never prints ETA. | |
85 | .TP | |
86 | .BI \-\-eta\-interval \fR=\fPtime | |
87 | By default, fio requests client ETA status roughly every second. With this | |
88 | option, the interval is configurable. Fio imposes a minimum allowed time to | |
89 | avoid flooding the console, less than 250 msec is not supported. | |
d60e92d1 | 90 | .TP |
30b5d57f | 91 | .BI \-\-eta\-newline \fR=\fPtime |
bdd88be3 TK |
92 | Force a new line for every \fItime\fR period passed. When the unit is omitted, |
93 | the value is interpreted in seconds. | |
30b5d57f JA |
94 | .TP |
95 | .BI \-\-status\-interval \fR=\fPtime | |
aa6cb459 VF |
96 | Force a full status dump of cumulative (from job start) values at \fItime\fR |
97 | intervals. This option does *not* provide per-period measurements. So | |
98 | values such as bandwidth are running averages. When the time unit is omitted, | |
c1f4de8a JA |
99 | \fItime\fR is interpreted in seconds. Note that using this option with |
100 | `\-\-output-format=json' will yield output that technically isn't valid json, | |
101 | since the output will be collated sets of valid json. It will need to be split | |
102 | into valid sets of json after the run. | |
bdd88be3 TK |
103 | .TP |
104 | .BI \-\-section \fR=\fPname | |
105 | Only run specified section \fIname\fR in job file. Multiple sections can be specified. | |
7db7a5a0 | 106 | The \fB\-\-section\fR option allows one to combine related jobs into one file. |
bdd88be3 | 107 | E.g. one job file could define light, moderate, and heavy sections. Tell |
7db7a5a0 | 108 | fio to run only the "heavy" section by giving `\-\-section=heavy' |
bdd88be3 | 109 | command line option. One can also specify the "write" operations in one |
7db7a5a0 | 110 | section and "verify" operation in another section. The \fB\-\-section\fR option |
bdd88be3 TK |
111 | only applies to job sections. The reserved *global* section is always |
112 | parsed and used. | |
c0a5d35e | 113 | .TP |
49da1240 | 114 | .BI \-\-alloc\-size \fR=\fPkb |
4a419903 VF |
115 | Allocate additional internal smalloc pools of size \fIkb\fR in KiB. The |
116 | \fB\-\-alloc\-size\fR option increases shared memory set aside for use by fio. | |
bdd88be3 TK |
117 | If running large jobs with randommap enabled, fio can run out of memory. |
118 | Smalloc is an internal allocator for shared structures from a fixed size | |
119 | memory pool and can grow to 16 pools. The pool size defaults to 16MiB. | |
7db7a5a0 TK |
120 | NOTE: While running `.fio_smalloc.*' backing store files are visible |
121 | in `/tmp'. | |
d60e92d1 | 122 | .TP |
49da1240 JA |
123 | .BI \-\-warnings\-fatal |
124 | All fio parser warnings are fatal, causing fio to exit with an error. | |
9183788d | 125 | .TP |
49da1240 | 126 | .BI \-\-max\-jobs \fR=\fPnr |
7db7a5a0 | 127 | Set the maximum number of threads/processes to support to \fInr\fR. |
7f4811bb RNS |
128 | NOTE: On Linux, it may be necessary to increase the shared-memory limit |
129 | (`/proc/sys/kernel/shmmax') if fio runs into errors while creating jobs. | |
d60e92d1 | 130 | .TP |
49da1240 | 131 | .BI \-\-server \fR=\fPargs |
7db7a5a0 TK |
132 | Start a backend server, with \fIargs\fR specifying what to listen to. |
133 | See \fBCLIENT/SERVER\fR section. | |
f57a9c59 | 134 | .TP |
49da1240 | 135 | .BI \-\-daemonize \fR=\fPpidfile |
7db7a5a0 | 136 | Background a fio server, writing the pid to the given \fIpidfile\fR file. |
49da1240 | 137 | .TP |
bdd88be3 | 138 | .BI \-\-client \fR=\fPhostname |
7db7a5a0 TK |
139 | Instead of running the jobs locally, send and run them on the given \fIhostname\fR |
140 | or set of \fIhostname\fRs. See \fBCLIENT/SERVER\fR section. | |
bdd88be3 | 141 | .TP |
7db7a5a0 TK |
142 | .BI \-\-remote\-config \fR=\fPfile |
143 | Tell fio server to load this local \fIfile\fR. | |
f2a2ce0e HL |
144 | .TP |
145 | .BI \-\-idle\-prof \fR=\fPoption | |
7db7a5a0 | 146 | Report CPU idleness. \fIoption\fR is one of the following: |
bdd88be3 TK |
147 | .RS |
148 | .RS | |
149 | .TP | |
150 | .B calibrate | |
151 | Run unit work calibration only and exit. | |
152 | .TP | |
153 | .B system | |
154 | Show aggregate system idleness and unit work. | |
155 | .TP | |
156 | .B percpu | |
7db7a5a0 | 157 | As \fBsystem\fR but also show per CPU idleness. |
bdd88be3 TK |
158 | .RE |
159 | .RE | |
160 | .TP | |
7db7a5a0 TK |
161 | .BI \-\-inflate\-log \fR=\fPlog |
162 | Inflate and output compressed \fIlog\fR. | |
bdd88be3 | 163 | .TP |
7db7a5a0 TK |
164 | .BI \-\-trigger\-file \fR=\fPfile |
165 | Execute trigger command when \fIfile\fR exists. | |
bdd88be3 | 166 | .TP |
7db7a5a0 TK |
167 | .BI \-\-trigger\-timeout \fR=\fPtime |
168 | Execute trigger at this \fItime\fR. | |
bdd88be3 | 169 | .TP |
7db7a5a0 TK |
170 | .BI \-\-trigger \fR=\fPcommand |
171 | Set this \fIcommand\fR as local trigger. | |
bdd88be3 | 172 | .TP |
7db7a5a0 TK |
173 | .BI \-\-trigger\-remote \fR=\fPcommand |
174 | Set this \fIcommand\fR as remote trigger. | |
bdd88be3 | 175 | .TP |
7db7a5a0 | 176 | .BI \-\-aux\-path \fR=\fPpath |
f4401bf8 SW |
177 | Use the directory specified by \fIpath\fP for generated state files instead |
178 | of the current working directory. | |
d60e92d1 | 179 | .SH "JOB FILE FORMAT" |
7a14cf18 TK |
180 | Any parameters following the options will be assumed to be job files, unless |
181 | they match a job file parameter. Multiple job files can be listed and each job | |
7db7a5a0 | 182 | file will be regarded as a separate group. Fio will \fBstonewall\fR execution |
7a14cf18 TK |
183 | between each group. |
184 | ||
185 | Fio accepts one or more job files describing what it is | |
186 | supposed to do. The job file format is the classic ini file, where the names | |
187 | enclosed in [] brackets define the job name. You are free to use any ASCII name | |
188 | you want, except *global* which has special meaning. Following the job name is | |
189 | a sequence of zero or more parameters, one per line, that define the behavior of | |
190 | the job. If the first character in a line is a ';' or a '#', the entire line is | |
191 | discarded as a comment. | |
192 | ||
193 | A *global* section sets defaults for the jobs described in that file. A job may | |
194 | override a *global* section parameter, and a job file may even have several | |
195 | *global* sections if so desired. A job is only affected by a *global* section | |
196 | residing above it. | |
197 | ||
7db7a5a0 TK |
198 | The \fB\-\-cmdhelp\fR option also lists all options. If used with an \fIcommand\fR |
199 | argument, \fB\-\-cmdhelp\fR will detail the given \fIcommand\fR. | |
7a14cf18 | 200 | |
7db7a5a0 TK |
201 | See the `examples/' directory for inspiration on how to write job files. Note |
202 | the copyright and license requirements currently apply to | |
203 | `examples/' files. | |
07a2919d NP |
204 | |
205 | Note that the maximum length of a line in the job file is 8192 bytes. | |
54eb4569 TK |
206 | .SH "JOB FILE PARAMETERS" |
207 | Some parameters take an option of a given type, such as an integer or a | |
208 | string. Anywhere a numeric value is required, an arithmetic expression may be | |
209 | used, provided it is surrounded by parentheses. Supported operators are: | |
d59aa780 | 210 | .RS |
7db7a5a0 | 211 | .P |
d59aa780 | 212 | .B addition (+) |
7db7a5a0 TK |
213 | .P |
214 | .B subtraction (\-) | |
215 | .P | |
d59aa780 | 216 | .B multiplication (*) |
7db7a5a0 | 217 | .P |
d59aa780 | 218 | .B division (/) |
7db7a5a0 | 219 | .P |
d59aa780 | 220 | .B modulus (%) |
7db7a5a0 | 221 | .P |
d59aa780 JA |
222 | .B exponentiation (^) |
223 | .RE | |
d59aa780 JA |
224 | .P |
225 | For time values in expressions, units are microseconds by default. This is | |
226 | different than for time values not in expressions (not enclosed in | |
54eb4569 TK |
227 | parentheses). |
228 | .SH "PARAMETER TYPES" | |
229 | The following parameter types are used. | |
d60e92d1 AC |
230 | .TP |
231 | .I str | |
6b86fc18 TK |
232 | String. A sequence of alphanumeric characters. |
233 | .TP | |
234 | .I time | |
235 | Integer with possible time suffix. Without a unit value is interpreted as | |
236 | seconds unless otherwise specified. Accepts a suffix of 'd' for days, 'h' for | |
237 | hours, 'm' for minutes, 's' for seconds, 'ms' (or 'msec') for milliseconds and 'us' | |
238 | (or 'usec') for microseconds. For example, use 10m for 10 minutes. | |
d60e92d1 AC |
239 | .TP |
240 | .I int | |
6d500c2e RE |
241 | Integer. A whole number value, which may contain an integer prefix |
242 | and an integer suffix. | |
0b43a833 TK |
243 | .RS |
244 | .RS | |
245 | .P | |
6b86fc18 | 246 | [*integer prefix*] **number** [*integer suffix*] |
0b43a833 TK |
247 | .RE |
248 | .P | |
6b86fc18 TK |
249 | The optional *integer prefix* specifies the number's base. The default |
250 | is decimal. *0x* specifies hexadecimal. | |
0b43a833 | 251 | .P |
6b86fc18 TK |
252 | The optional *integer suffix* specifies the number's units, and includes an |
253 | optional unit prefix and an optional unit. For quantities of data, the | |
254 | default unit is bytes. For quantities of time, the default unit is seconds | |
255 | unless otherwise specified. | |
0b43a833 TK |
256 | .P |
257 | With `kb_base=1000', fio follows international standards for unit | |
338f2db5 | 258 | prefixes. To specify power-of-10 decimal values defined in the |
6b86fc18 | 259 | International System of Units (SI): |
0b43a833 TK |
260 | .RS |
261 | .P | |
7db7a5a0 | 262 | .PD 0 |
eccce61a | 263 | K means kilo (K) or 1000 |
7db7a5a0 | 264 | .P |
eccce61a | 265 | M means mega (M) or 1000**2 |
7db7a5a0 | 266 | .P |
eccce61a | 267 | G means giga (G) or 1000**3 |
7db7a5a0 | 268 | .P |
eccce61a | 269 | T means tera (T) or 1000**4 |
7db7a5a0 | 270 | .P |
eccce61a | 271 | P means peta (P) or 1000**5 |
7db7a5a0 | 272 | .PD |
0b43a833 TK |
273 | .RE |
274 | .P | |
338f2db5 | 275 | To specify power-of-2 binary values defined in IEC 80000-13: |
0b43a833 TK |
276 | .RS |
277 | .P | |
7db7a5a0 | 278 | .PD 0 |
eccce61a | 279 | Ki means kibi (Ki) or 1024 |
7db7a5a0 | 280 | .P |
eccce61a | 281 | Mi means mebi (Mi) or 1024**2 |
7db7a5a0 | 282 | .P |
eccce61a | 283 | Gi means gibi (Gi) or 1024**3 |
7db7a5a0 | 284 | .P |
eccce61a | 285 | Ti means tebi (Ti) or 1024**4 |
7db7a5a0 | 286 | .P |
eccce61a | 287 | Pi means pebi (Pi) or 1024**5 |
7db7a5a0 | 288 | .PD |
0b43a833 TK |
289 | .RE |
290 | .P | |
291 | With `kb_base=1024' (the default), the unit prefixes are opposite | |
338f2db5 | 292 | from those specified in the SI and IEC 80000-13 standards to provide |
6b86fc18 | 293 | compatibility with old scripts. For example, 4k means 4096. |
0b43a833 | 294 | .P |
6b86fc18 TK |
295 | For quantities of data, an optional unit of 'B' may be included |
296 | (e.g., 'kB' is the same as 'k'). | |
0b43a833 | 297 | .P |
6b86fc18 TK |
298 | The *integer suffix* is not case sensitive (e.g., m/mi mean mebi/mega, |
299 | not milli). 'b' and 'B' both mean byte, not bit. | |
0b43a833 TK |
300 | .P |
301 | Examples with `kb_base=1000': | |
302 | .RS | |
303 | .P | |
7db7a5a0 | 304 | .PD 0 |
6d500c2e | 305 | 4 KiB: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB |
7db7a5a0 | 306 | .P |
6d500c2e | 307 | 1 MiB: 1048576, 1m, 1024k |
7db7a5a0 | 308 | .P |
6d500c2e | 309 | 1 MB: 1000000, 1mi, 1000ki |
7db7a5a0 | 310 | .P |
6d500c2e | 311 | 1 TiB: 1073741824, 1t, 1024m, 1048576k |
7db7a5a0 | 312 | .P |
6d500c2e | 313 | 1 TB: 1000000000, 1ti, 1000mi, 1000000ki |
7db7a5a0 | 314 | .PD |
0b43a833 TK |
315 | .RE |
316 | .P | |
317 | Examples with `kb_base=1024' (default): | |
318 | .RS | |
319 | .P | |
7db7a5a0 | 320 | .PD 0 |
6d500c2e | 321 | 4 KiB: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB |
7db7a5a0 | 322 | .P |
6d500c2e | 323 | 1 MiB: 1048576, 1m, 1024k |
7db7a5a0 | 324 | .P |
6d500c2e | 325 | 1 MB: 1000000, 1mi, 1000ki |
7db7a5a0 | 326 | .P |
6d500c2e | 327 | 1 TiB: 1073741824, 1t, 1024m, 1048576k |
7db7a5a0 | 328 | .P |
6d500c2e | 329 | 1 TB: 1000000000, 1ti, 1000mi, 1000000ki |
7db7a5a0 | 330 | .PD |
0b43a833 TK |
331 | .RE |
332 | .P | |
6d500c2e | 333 | To specify times (units are not case sensitive): |
0b43a833 TK |
334 | .RS |
335 | .P | |
7db7a5a0 | 336 | .PD 0 |
6d500c2e | 337 | D means days |
7db7a5a0 | 338 | .P |
6d500c2e | 339 | H means hours |
7db7a5a0 | 340 | .P |
6d500c2e | 341 | M mean minutes |
7db7a5a0 | 342 | .P |
6d500c2e | 343 | s or sec means seconds (default) |
7db7a5a0 | 344 | .P |
6d500c2e | 345 | ms or msec means milliseconds |
7db7a5a0 | 346 | .P |
6d500c2e | 347 | us or usec means microseconds |
7db7a5a0 | 348 | .PD |
0b43a833 TK |
349 | .RE |
350 | .P | |
8f39afa7 AD |
351 | `z' suffix specifies that the value is measured in zones. |
352 | Value is recalculated once block device's zone size becomes known. | |
353 | .P | |
6b86fc18 | 354 | If the option accepts an upper and lower range, use a colon ':' or |
7db7a5a0 | 355 | minus '\-' to separate such values. See \fIirange\fR parameter type. |
6b86fc18 TK |
356 | If the lower value specified happens to be larger than the upper value |
357 | the two values are swapped. | |
0b43a833 | 358 | .RE |
d60e92d1 AC |
359 | .TP |
360 | .I bool | |
6b86fc18 TK |
361 | Boolean. Usually parsed as an integer, however only defined for |
362 | true and false (1 and 0). | |
d60e92d1 AC |
363 | .TP |
364 | .I irange | |
6b86fc18 | 365 | Integer range with suffix. Allows value range to be given, such as |
7db7a5a0 | 366 | 1024\-4096. A colon may also be used as the separator, e.g. 1k:4k. If the |
6b86fc18 | 367 | option allows two sets of ranges, they can be specified with a ',' or '/' |
7db7a5a0 | 368 | delimiter: 1k\-4k/8k\-32k. Also see \fIint\fR parameter type. |
83349190 YH |
369 | .TP |
370 | .I float_list | |
6b86fc18 | 371 | A list of floating point numbers, separated by a ':' character. |
523bad63 | 372 | .SH "JOB PARAMETERS" |
54eb4569 | 373 | With the above in mind, here follows the complete list of fio job parameters. |
523bad63 | 374 | .SS "Units" |
d60e92d1 | 375 | .TP |
523bad63 TK |
376 | .BI kb_base \fR=\fPint |
377 | Select the interpretation of unit prefixes in input parameters. | |
378 | .RS | |
379 | .RS | |
d60e92d1 | 380 | .TP |
523bad63 | 381 | .B 1000 |
338f2db5 | 382 | Inputs comply with IEC 80000-13 and the International |
523bad63 TK |
383 | System of Units (SI). Use: |
384 | .RS | |
385 | .P | |
386 | .PD 0 | |
338f2db5 | 387 | \- power-of-2 values with IEC prefixes (e.g., KiB) |
523bad63 | 388 | .P |
338f2db5 | 389 | \- power-of-10 values with SI prefixes (e.g., kB) |
523bad63 TK |
390 | .PD |
391 | .RE | |
392 | .TP | |
393 | .B 1024 | |
394 | Compatibility mode (default). To avoid breaking old scripts: | |
395 | .P | |
396 | .RS | |
397 | .PD 0 | |
338f2db5 | 398 | \- power-of-2 values with SI prefixes |
523bad63 | 399 | .P |
338f2db5 | 400 | \- power-of-10 values with IEC prefixes |
523bad63 TK |
401 | .PD |
402 | .RE | |
403 | .RE | |
404 | .P | |
405 | See \fBbs\fR for more details on input parameters. | |
406 | .P | |
407 | Outputs always use correct prefixes. Most outputs include both | |
338f2db5 | 408 | side-by-side, like: |
523bad63 TK |
409 | .P |
410 | .RS | |
411 | bw=2383.3kB/s (2327.4KiB/s) | |
412 | .RE | |
413 | .P | |
414 | If only one value is reported, then kb_base selects the one to use: | |
415 | .P | |
416 | .RS | |
417 | .PD 0 | |
418 | 1000 \-\- SI prefixes | |
419 | .P | |
420 | 1024 \-\- IEC prefixes | |
421 | .PD | |
422 | .RE | |
423 | .RE | |
424 | .TP | |
425 | .BI unit_base \fR=\fPint | |
426 | Base unit for reporting. Allowed values are: | |
427 | .RS | |
428 | .RS | |
429 | .TP | |
430 | .B 0 | |
338f2db5 | 431 | Use auto-detection (default). |
523bad63 TK |
432 | .TP |
433 | .B 8 | |
434 | Byte based. | |
435 | .TP | |
436 | .B 1 | |
437 | Bit based. | |
438 | .RE | |
439 | .RE | |
440 | .SS "Job description" | |
441 | .TP | |
442 | .BI name \fR=\fPstr | |
443 | ASCII name of the job. This may be used to override the name printed by fio | |
444 | for this job. Otherwise the job name is used. On the command line this | |
445 | parameter has the special purpose of also signaling the start of a new job. | |
9cc8cb91 | 446 | .TP |
d60e92d1 | 447 | .BI description \fR=\fPstr |
523bad63 TK |
448 | Text description of the job. Doesn't do anything except dump this text |
449 | description when this job is run. It's not parsed. | |
450 | .TP | |
451 | .BI loops \fR=\fPint | |
452 | Run the specified number of iterations of this job. Used to repeat the same | |
453 | workload a given number of times. Defaults to 1. | |
454 | .TP | |
455 | .BI numjobs \fR=\fPint | |
456 | Create the specified number of clones of this job. Each clone of job | |
457 | is spawned as an independent thread or process. May be used to setup a | |
458 | larger number of threads/processes doing the same thing. Each thread is | |
459 | reported separately; to see statistics for all clones as a whole, use | |
460 | \fBgroup_reporting\fR in conjunction with \fBnew_group\fR. | |
461 | See \fB\-\-max\-jobs\fR. Default: 1. | |
462 | .SS "Time related parameters" | |
463 | .TP | |
464 | .BI runtime \fR=\fPtime | |
465 | Tell fio to terminate processing after the specified period of time. It | |
466 | can be quite hard to determine for how long a specified job will run, so | |
467 | this parameter is handy to cap the total runtime to a given time. When | |
f1dd3fb1 | 468 | the unit is omitted, the value is interpreted in seconds. |
523bad63 TK |
469 | .TP |
470 | .BI time_based | |
471 | If set, fio will run for the duration of the \fBruntime\fR specified | |
472 | even if the file(s) are completely read or written. It will simply loop over | |
473 | the same workload as many times as the \fBruntime\fR allows. | |
474 | .TP | |
475 | .BI startdelay \fR=\fPirange(int) | |
476 | Delay the start of job for the specified amount of time. Can be a single | |
477 | value or a range. When given as a range, each thread will choose a value | |
478 | randomly from within the range. Value is in seconds if a unit is omitted. | |
479 | .TP | |
480 | .BI ramp_time \fR=\fPtime | |
481 | If set, fio will run the specified workload for this amount of time before | |
482 | logging any performance numbers. Useful for letting performance settle | |
483 | before logging results, thus minimizing the runtime required for stable | |
484 | results. Note that the \fBramp_time\fR is considered lead in time for a job, | |
485 | thus it will increase the total runtime if a special timeout or | |
486 | \fBruntime\fR is specified. When the unit is omitted, the value is | |
487 | given in seconds. | |
488 | .TP | |
489 | .BI clocksource \fR=\fPstr | |
490 | Use the given clocksource as the base of timing. The supported options are: | |
491 | .RS | |
492 | .RS | |
493 | .TP | |
494 | .B gettimeofday | |
495 | \fBgettimeofday\fR\|(2) | |
496 | .TP | |
497 | .B clock_gettime | |
498 | \fBclock_gettime\fR\|(2) | |
499 | .TP | |
500 | .B cpu | |
501 | Internal CPU clock source | |
502 | .RE | |
503 | .P | |
504 | \fBcpu\fR is the preferred clocksource if it is reliable, as it is very fast (and | |
505 | fio is heavy on time calls). Fio will automatically use this clocksource if | |
506 | it's supported and considered reliable on the system it is running on, | |
507 | unless another clocksource is specifically set. For x86/x86\-64 CPUs, this | |
508 | means supporting TSC Invariant. | |
509 | .RE | |
510 | .TP | |
511 | .BI gtod_reduce \fR=\fPbool | |
512 | Enable all of the \fBgettimeofday\fR\|(2) reducing options | |
513 | (\fBdisable_clat\fR, \fBdisable_slat\fR, \fBdisable_bw_measurement\fR) plus | |
514 | reduce precision of the timeout somewhat to really shrink the | |
515 | \fBgettimeofday\fR\|(2) call count. With this option enabled, we only do | |
516 | about 0.4% of the \fBgettimeofday\fR\|(2) calls we would have done if all | |
517 | time keeping was enabled. | |
518 | .TP | |
519 | .BI gtod_cpu \fR=\fPint | |
520 | Sometimes it's cheaper to dedicate a single thread of execution to just | |
521 | getting the current time. Fio (and databases, for instance) are very | |
522 | intensive on \fBgettimeofday\fR\|(2) calls. With this option, you can set | |
523 | one CPU aside for doing nothing but logging current time to a shared memory | |
524 | location. Then the other threads/processes that run I/O workloads need only | |
525 | copy that segment, instead of entering the kernel with a | |
526 | \fBgettimeofday\fR\|(2) call. The CPU set aside for doing these time | |
527 | calls will be excluded from other uses. Fio will manually clear it from the | |
528 | CPU mask of other jobs. | |
529 | .SS "Target file/device" | |
d60e92d1 AC |
530 | .TP |
531 | .BI directory \fR=\fPstr | |
523bad63 TK |
532 | Prefix \fBfilename\fRs with this directory. Used to place files in a different |
533 | location than `./'. You can specify a number of directories by | |
534 | separating the names with a ':' character. These directories will be | |
535 | assigned equally distributed to job clones created by \fBnumjobs\fR as | |
536 | long as they are using generated filenames. If specific \fBfilename\fR(s) are | |
537 | set fio will use the first listed directory, and thereby matching the | |
f4401bf8 SW |
538 | \fBfilename\fR semantic (which generates a file for each clone if not |
539 | specified, but lets all clones use the same file if set). | |
523bad63 TK |
540 | .RS |
541 | .P | |
3b803fe1 | 542 | See the \fBfilename\fR option for information on how to escape ':' |
523bad63 | 543 | characters within the directory path itself. |
f4401bf8 SW |
544 | .P |
545 | Note: To control the directory fio will use for internal state files | |
546 | use \fB\-\-aux\-path\fR. | |
523bad63 | 547 | .RE |
d60e92d1 AC |
548 | .TP |
549 | .BI filename \fR=\fPstr | |
523bad63 TK |
550 | Fio normally makes up a \fBfilename\fR based on the job name, thread number, and |
551 | file number (see \fBfilename_format\fR). If you want to share files | |
552 | between threads in a job or several | |
553 | jobs with fixed file paths, specify a \fBfilename\fR for each of them to override | |
554 | the default. If the ioengine is file based, you can specify a number of files | |
555 | by separating the names with a ':' colon. So if you wanted a job to open | |
556 | `/dev/sda' and `/dev/sdb' as the two working files, you would use | |
557 | `filename=/dev/sda:/dev/sdb'. This also means that whenever this option is | |
558 | specified, \fBnrfiles\fR is ignored. The size of regular files specified | |
559 | by this option will be \fBsize\fR divided by number of files unless an | |
560 | explicit size is specified by \fBfilesize\fR. | |
561 | .RS | |
562 | .P | |
3b803fe1 | 563 | Each colon in the wanted path must be escaped with a '\\' |
523bad63 TK |
564 | character. For instance, if the path is `/dev/dsk/foo@3,0:c' then you |
565 | would use `filename=/dev/dsk/foo@3,0\\:c' and if the path is | |
3b803fe1 | 566 | `F:\\filename' then you would use `filename=F\\:\\filename'. |
523bad63 | 567 | .P |
ffc90a44 SW |
568 | On Windows, disk devices are accessed as `\\\\.\\PhysicalDrive0' for |
569 | the first device, `\\\\.\\PhysicalDrive1' for the second etc. | |
523bad63 | 570 | Note: Windows and FreeBSD prevent write access to areas |
338f2db5 | 571 | of the disk containing in-use data (e.g. filesystems). |
523bad63 TK |
572 | .P |
573 | The filename `\-' is a reserved name, meaning *stdin* or *stdout*. Which | |
574 | of the two depends on the read/write direction set. | |
575 | .RE | |
d60e92d1 | 576 | .TP |
de98bd30 | 577 | .BI filename_format \fR=\fPstr |
523bad63 TK |
578 | If sharing multiple files between jobs, it is usually necessary to have fio |
579 | generate the exact names that you want. By default, fio will name a file | |
de98bd30 | 580 | based on the default file format specification of |
523bad63 | 581 | `jobname.jobnumber.filenumber'. With this option, that can be |
de98bd30 JA |
582 | customized. Fio will recognize and replace the following keywords in this |
583 | string: | |
584 | .RS | |
585 | .RS | |
586 | .TP | |
587 | .B $jobname | |
588 | The name of the worker thread or process. | |
589 | .TP | |
8d53c5f8 TG |
590 | .B $clientuid |
591 | IP of the fio process when using client/server mode. | |
592 | .TP | |
de98bd30 JA |
593 | .B $jobnum |
594 | The incremental number of the worker thread or process. | |
595 | .TP | |
596 | .B $filenum | |
597 | The incremental number of the file for that worker thread or process. | |
598 | .RE | |
599 | .P | |
523bad63 TK |
600 | To have dependent jobs share a set of files, this option can be set to have |
601 | fio generate filenames that are shared between the two. For instance, if | |
602 | `testfiles.$filenum' is specified, file number 4 for any job will be | |
603 | named `testfiles.4'. The default of `$jobname.$jobnum.$filenum' | |
de98bd30 | 604 | will be used if no other format specifier is given. |
645943c0 JB |
605 | .P |
606 | If you specify a path then the directories will be created up to the main | |
607 | directory for the file. So for example if you specify `a/b/c/$jobnum` then the | |
608 | directories a/b/c will be created before the file setup part of the job. If you | |
609 | specify \fBdirectory\fR then the path will be relative that directory, otherwise | |
610 | it is treated as the absolute path. | |
de98bd30 | 611 | .RE |
de98bd30 | 612 | .TP |
922a5be8 | 613 | .BI unique_filename \fR=\fPbool |
523bad63 TK |
614 | To avoid collisions between networked clients, fio defaults to prefixing any |
615 | generated filenames (with a directory specified) with the source of the | |
616 | client connecting. To disable this behavior, set this option to 0. | |
617 | .TP | |
618 | .BI opendir \fR=\fPstr | |
619 | Recursively open any files below directory \fIstr\fR. | |
922a5be8 | 620 | .TP |
3ce9dcaf | 621 | .BI lockfile \fR=\fPstr |
523bad63 TK |
622 | Fio defaults to not locking any files before it does I/O to them. If a file |
623 | or file descriptor is shared, fio can serialize I/O to that file to make the | |
624 | end result consistent. This is usual for emulating real workloads that share | |
625 | files. The lock modes are: | |
3ce9dcaf JA |
626 | .RS |
627 | .RS | |
628 | .TP | |
629 | .B none | |
523bad63 | 630 | No locking. The default. |
3ce9dcaf JA |
631 | .TP |
632 | .B exclusive | |
523bad63 | 633 | Only one thread or process may do I/O at a time, excluding all others. |
3ce9dcaf JA |
634 | .TP |
635 | .B readwrite | |
523bad63 TK |
636 | Read\-write locking on the file. Many readers may |
637 | access the file at the same time, but writes get exclusive access. | |
3ce9dcaf | 638 | .RE |
ce594fbe | 639 | .RE |
523bad63 TK |
640 | .TP |
641 | .BI nrfiles \fR=\fPint | |
642 | Number of files to use for this job. Defaults to 1. The size of files | |
643 | will be \fBsize\fR divided by this unless explicit size is specified by | |
644 | \fBfilesize\fR. Files are created for each thread separately, and each | |
645 | file will have a file number within its name by default, as explained in | |
646 | \fBfilename\fR section. | |
647 | .TP | |
648 | .BI openfiles \fR=\fPint | |
649 | Number of files to keep open at the same time. Defaults to the same as | |
650 | \fBnrfiles\fR, can be set smaller to limit the number simultaneous | |
651 | opens. | |
652 | .TP | |
653 | .BI file_service_type \fR=\fPstr | |
654 | Defines how fio decides which file from a job to service next. The following | |
655 | types are defined: | |
656 | .RS | |
657 | .RS | |
658 | .TP | |
659 | .B random | |
660 | Choose a file at random. | |
661 | .TP | |
662 | .B roundrobin | |
663 | Round robin over opened files. This is the default. | |
664 | .TP | |
665 | .B sequential | |
666 | Finish one file before moving on to the next. Multiple files can | |
667 | still be open depending on \fBopenfiles\fR. | |
668 | .TP | |
669 | .B zipf | |
670 | Use a Zipf distribution to decide what file to access. | |
671 | .TP | |
672 | .B pareto | |
673 | Use a Pareto distribution to decide what file to access. | |
674 | .TP | |
675 | .B normal | |
676 | Use a Gaussian (normal) distribution to decide what file to access. | |
677 | .TP | |
678 | .B gauss | |
679 | Alias for normal. | |
680 | .RE | |
3ce9dcaf | 681 | .P |
523bad63 TK |
682 | For \fBrandom\fR, \fBroundrobin\fR, and \fBsequential\fR, a postfix can be appended to |
683 | tell fio how many I/Os to issue before switching to a new file. For example, | |
684 | specifying `file_service_type=random:8' would cause fio to issue | |
338f2db5 | 685 | 8 I/Os before selecting a new file at random. For the non-uniform |
523bad63 TK |
686 | distributions, a floating point postfix can be given to influence how the |
687 | distribution is skewed. See \fBrandom_distribution\fR for a description | |
688 | of how that would work. | |
689 | .RE | |
690 | .TP | |
691 | .BI ioscheduler \fR=\fPstr | |
692 | Attempt to switch the device hosting the file to the specified I/O scheduler | |
5592e992 DLM |
693 | before running. If the file is a pipe, a character device file or if device |
694 | hosting the file could not be determined, this option is ignored. | |
523bad63 TK |
695 | .TP |
696 | .BI create_serialize \fR=\fPbool | |
697 | If true, serialize the file creation for the jobs. This may be handy to | |
698 | avoid interleaving of data files, which may greatly depend on the filesystem | |
699 | used and even the number of processors in the system. Default: true. | |
700 | .TP | |
701 | .BI create_fsync \fR=\fPbool | |
702 | \fBfsync\fR\|(2) the data file after creation. This is the default. | |
703 | .TP | |
704 | .BI create_on_open \fR=\fPbool | |
338f2db5 SW |
705 | If true, don't pre-create files but allow the job's open() to create a file |
706 | when it's time to do I/O. Default: false \-\- pre-create all necessary files | |
523bad63 TK |
707 | when the job starts. |
708 | .TP | |
709 | .BI create_only \fR=\fPbool | |
710 | If true, fio will only run the setup phase of the job. If files need to be | |
711 | laid out or updated on disk, only that will be done \-\- the actual job contents | |
712 | are not executed. Default: false. | |
713 | .TP | |
714 | .BI allow_file_create \fR=\fPbool | |
715 | If true, fio is permitted to create files as part of its workload. If this | |
716 | option is false, then fio will error out if | |
717 | the files it needs to use don't already exist. Default: true. | |
718 | .TP | |
719 | .BI allow_mounted_write \fR=\fPbool | |
720 | If this isn't set, fio will abort jobs that are destructive (e.g. that write) | |
721 | to what appears to be a mounted device or partition. This should help catch | |
722 | creating inadvertently destructive tests, not realizing that the test will | |
723 | destroy data on the mounted file system. Note that some platforms don't allow | |
724 | writing against a mounted device regardless of this option. Default: false. | |
725 | .TP | |
726 | .BI pre_read \fR=\fPbool | |
338f2db5 | 727 | If this is given, files will be pre-read into memory before starting the |
523bad63 | 728 | given I/O operation. This will also clear the \fBinvalidate\fR flag, |
338f2db5 SW |
729 | since it is pointless to pre-read and then drop the cache. This will only |
730 | work for I/O engines that are seek-able, since they allow you to read the | |
731 | same data multiple times. Thus it will not work on non-seekable I/O engines | |
523bad63 TK |
732 | (e.g. network, splice). Default: false. |
733 | .TP | |
734 | .BI unlink \fR=\fPbool | |
735 | Unlink the job files when done. Not the default, as repeated runs of that | |
736 | job would then waste time recreating the file set again and again. Default: | |
737 | false. | |
738 | .TP | |
739 | .BI unlink_each_loop \fR=\fPbool | |
740 | Unlink job files after each iteration or loop. Default: false. | |
741 | .TP | |
7b865a2f BVA |
742 | .BI zonemode \fR=\fPstr |
743 | Accepted values are: | |
744 | .RS | |
745 | .RS | |
746 | .TP | |
747 | .B none | |
b8dd9750 HH |
748 | The \fBzonerange\fR, \fBzonesize\fR \fBzonecapacity\fR and \fBzoneskip\fR |
749 | parameters are ignored. | |
7b865a2f BVA |
750 | .TP |
751 | .B strided | |
752 | I/O happens in a single zone until \fBzonesize\fR bytes have been transferred. | |
753 | After that number of bytes has been transferred processing of the next zone | |
b8dd9750 | 754 | starts. The \fBzonecapacity\fR parameter is ignored. |
7b865a2f BVA |
755 | .TP |
756 | .B zbd | |
757 | Zoned block device mode. I/O happens sequentially in each zone, even if random | |
758 | I/O has been selected. Random I/O happens across all zones instead of being | |
759 | restricted to a single zone. | |
760 | .RE | |
761 | .RE | |
523bad63 TK |
762 | .TP |
763 | .BI zonerange \fR=\fPint | |
d4e058cd DLM |
764 | For \fBzonemode\fR=strided, this is the size of a single zone. See also |
765 | \fBzonesize\fR and \fBzoneskip\fR. | |
766 | ||
767 | For \fBzonemode\fR=zbd, this parameter is ignored. | |
5faddc64 BVA |
768 | .TP |
769 | .BI zonesize \fR=\fPint | |
7b865a2f BVA |
770 | For \fBzonemode\fR=strided, this is the number of bytes to transfer before |
771 | skipping \fBzoneskip\fR bytes. If this parameter is smaller than | |
772 | \fBzonerange\fR then only a fraction of each zone with \fBzonerange\fR bytes | |
773 | will be accessed. If this parameter is larger than \fBzonerange\fR then each | |
774 | zone will be accessed multiple times before skipping to the next zone. | |
775 | ||
d4e058cd DLM |
776 | For \fBzonemode\fR=zbd, this is the size of a single zone. The |
777 | \fBzonerange\fR parameter is ignored in this mode. For a job accessing a | |
778 | zoned block device, the specified \fBzonesize\fR must be 0 or equal to the | |
779 | device zone size. For a regular block device or file, the specified | |
780 | \fBzonesize\fR must be at least 512B. | |
523bad63 | 781 | .TP |
b8dd9750 HH |
782 | .BI zonecapacity \fR=\fPint |
783 | For \fBzonemode\fR=zbd, this defines the capacity of a single zone, which is | |
784 | the accessible area starting from the zone start address. This parameter only | |
785 | applies when using \fBzonemode\fR=zbd in combination with regular block devices. | |
786 | If not specified it defaults to the zone size. If the target device is a zoned | |
787 | block device, the zone capacity is obtained from the device information and this | |
788 | option is ignored. | |
789 | .TP | |
8f39afa7 | 790 | .BI zoneskip \fR=\fPint[z] |
7b865a2f | 791 | For \fBzonemode\fR=strided, the number of bytes to skip after \fBzonesize\fR |
4d37720a DLM |
792 | bytes of data have been transferred. |
793 | ||
794 | For \fBzonemode\fR=zbd, the \fBzonesize\fR aligned number of bytes to skip | |
795 | once a zone is fully written (write workloads) or all written data in the | |
796 | zone have been read (read workloads). This parameter is valid only for | |
797 | sequential workloads and ignored for random workloads. For read workloads, | |
798 | see also \fBread_beyond_wp\fR. | |
5faddc64 | 799 | |
bfbdd35b BVA |
800 | .TP |
801 | .BI read_beyond_wp \fR=\fPbool | |
802 | This parameter applies to \fBzonemode=zbd\fR only. | |
803 | ||
804 | Zoned block devices are block devices that consist of multiple zones. Each | |
805 | zone has a type, e.g. conventional or sequential. A conventional zone can be | |
806 | written at any offset that is a multiple of the block size. Sequential zones | |
807 | must be written sequentially. The position at which a write must occur is | |
402f0887 DLM |
808 | called the write pointer. A zoned block device can be either host managed or |
809 | host aware. For host managed devices the host must ensure that writes happen | |
810 | sequentially. Fio recognizes host managed devices and serializes writes to | |
811 | sequential zones for these devices. | |
bfbdd35b BVA |
812 | |
813 | If a read occurs in a sequential zone beyond the write pointer then the zoned | |
814 | block device will complete the read without reading any data from the storage | |
815 | medium. Since such reads lead to unrealistically high bandwidth and IOPS | |
816 | numbers fio only reads beyond the write pointer if explicitly told to do | |
817 | so. Default: false. | |
59b07544 BVA |
818 | .TP |
819 | .BI max_open_zones \fR=\fPint | |
820 | When running a random write test across an entire drive many more zones will be | |
821 | open than in a typical application workload. Hence this command line option | |
822 | that allows to limit the number of open zones. The number of open zones is | |
219c662d AD |
823 | defined as the number of zones to which write commands are issued by all |
824 | threads/processes. | |
825 | .TP | |
826 | .BI job_max_open_zones \fR=\fPint | |
827 | Limit on the number of simultaneously opened zones per single thread/process. | |
a7c2b6fc BVA |
828 | .TP |
829 | .BI zone_reset_threshold \fR=\fPfloat | |
830 | A number between zero and one that indicates the ratio of logical blocks with | |
831 | data to the total number of logical blocks in the test above which zones | |
832 | should be reset periodically. | |
833 | .TP | |
834 | .BI zone_reset_frequency \fR=\fPfloat | |
835 | A number between zero and one that indicates how often a zone reset should be | |
836 | issued if the zone reset threshold has been exceeded. A zone reset is | |
837 | submitted after each (1 / zone_reset_frequency) write requests. This and the | |
838 | previous parameter can be used to simulate garbage collection activity. | |
bfbdd35b | 839 | |
523bad63 TK |
840 | .SS "I/O type" |
841 | .TP | |
842 | .BI direct \fR=\fPbool | |
338f2db5 | 843 | If value is true, use non-buffered I/O. This is usually O_DIRECT. Note that |
8e889110 | 844 | OpenBSD and ZFS on Solaris don't support direct I/O. On Windows the synchronous |
523bad63 TK |
845 | ioengines don't support direct I/O. Default: false. |
846 | .TP | |
847 | .BI atomic \fR=\fPbool | |
848 | If value is true, attempt to use atomic direct I/O. Atomic writes are | |
849 | guaranteed to be stable once acknowledged by the operating system. Only | |
850 | Linux supports O_ATOMIC right now. | |
851 | .TP | |
852 | .BI buffered \fR=\fPbool | |
853 | If value is true, use buffered I/O. This is the opposite of the | |
854 | \fBdirect\fR option. Defaults to true. | |
d60e92d1 AC |
855 | .TP |
856 | .BI readwrite \fR=\fPstr "\fR,\fP rw" \fR=\fPstr | |
523bad63 | 857 | Type of I/O pattern. Accepted values are: |
d60e92d1 AC |
858 | .RS |
859 | .RS | |
860 | .TP | |
861 | .B read | |
d1429b5c | 862 | Sequential reads. |
d60e92d1 AC |
863 | .TP |
864 | .B write | |
d1429b5c | 865 | Sequential writes. |
d60e92d1 | 866 | .TP |
fa769d44 | 867 | .B trim |
3740cfc8 | 868 | Sequential trims (Linux block devices and SCSI character devices only). |
fa769d44 | 869 | .TP |
d60e92d1 | 870 | .B randread |
d1429b5c | 871 | Random reads. |
d60e92d1 AC |
872 | .TP |
873 | .B randwrite | |
d1429b5c | 874 | Random writes. |
d60e92d1 | 875 | .TP |
fa769d44 | 876 | .B randtrim |
3740cfc8 | 877 | Random trims (Linux block devices and SCSI character devices only). |
fa769d44 | 878 | .TP |
523bad63 TK |
879 | .B rw,readwrite |
880 | Sequential mixed reads and writes. | |
d60e92d1 | 881 | .TP |
ff6bb260 | 882 | .B randrw |
523bad63 | 883 | Random mixed reads and writes. |
82a90686 JA |
884 | .TP |
885 | .B trimwrite | |
523bad63 TK |
886 | Sequential trim+write sequences. Blocks will be trimmed first, |
887 | then the same blocks will be written to. | |
d60e92d1 AC |
888 | .RE |
889 | .P | |
523bad63 TK |
890 | Fio defaults to read if the option is not specified. For the mixed I/O |
891 | types, the default is to split them 50/50. For certain types of I/O the | |
892 | result may still be skewed a bit, since the speed may be different. | |
893 | .P | |
894 | It is possible to specify the number of I/Os to do before getting a new | |
895 | offset by appending `:<nr>' to the end of the string given. For a | |
896 | random read, it would look like `rw=randread:8' for passing in an offset | |
897 | modifier with a value of 8. If the suffix is used with a sequential I/O | |
898 | pattern, then the `<nr>' value specified will be added to the generated | |
899 | offset for each I/O turning sequential I/O into sequential I/O with holes. | |
900 | For instance, using `rw=write:4k' will skip 4k for every write. Also see | |
901 | the \fBrw_sequencer\fR option. | |
d60e92d1 AC |
902 | .RE |
903 | .TP | |
38dad62d | 904 | .BI rw_sequencer \fR=\fPstr |
523bad63 TK |
905 | If an offset modifier is given by appending a number to the `rw=\fIstr\fR' |
906 | line, then this option controls how that number modifies the I/O offset | |
907 | being generated. Accepted values are: | |
38dad62d JA |
908 | .RS |
909 | .RS | |
910 | .TP | |
911 | .B sequential | |
523bad63 | 912 | Generate sequential offset. |
38dad62d JA |
913 | .TP |
914 | .B identical | |
523bad63 | 915 | Generate the same offset. |
38dad62d JA |
916 | .RE |
917 | .P | |
523bad63 TK |
918 | \fBsequential\fR is only useful for random I/O, where fio would normally |
919 | generate a new random offset for every I/O. If you append e.g. 8 to randread, | |
920 | you would get a new random offset for every 8 I/Os. The result would be a | |
921 | seek for only every 8 I/Os, instead of for every I/O. Use `rw=randread:8' | |
922 | to specify that. As sequential I/O is already sequential, setting | |
923 | \fBsequential\fR for that would not result in any differences. \fBidentical\fR | |
924 | behaves in a similar fashion, except it sends the same offset 8 number of | |
925 | times before generating a new offset. | |
38dad62d | 926 | .RE |
90fef2d1 | 927 | .TP |
5cb8a8cd | 928 | .BI unified_rw_reporting \fR=\fPstr |
771e58be | 929 | Fio normally reports statistics on a per data direction basis, meaning that |
5cb8a8cd BP |
930 | reads, writes, and trims are accounted and reported separately. This option |
931 | determines whether fio reports the results normally, summed together, or as | |
932 | both options. | |
933 | Accepted values are: | |
934 | .RS | |
935 | .TP | |
936 | .B none | |
937 | Normal statistics reporting. | |
938 | .TP | |
939 | .B mixed | |
940 | Statistics are summed per data direction and reported together. | |
941 | .TP | |
942 | .B both | |
943 | Statistics are reported normally, followed by the mixed statistics. | |
944 | .TP | |
945 | .B 0 | |
946 | Backward-compatible alias for \fBnone\fR. | |
947 | .TP | |
948 | .B 1 | |
949 | Backward-compatible alias for \fBmixed\fR. | |
950 | .TP | |
951 | .B 2 | |
952 | Alias for \fBboth\fR. | |
953 | .RE | |
771e58be | 954 | .TP |
d60e92d1 | 955 | .BI randrepeat \fR=\fPbool |
523bad63 TK |
956 | Seed the random number generator used for random I/O patterns in a |
957 | predictable way so the pattern is repeatable across runs. Default: true. | |
56e2a5fc CE |
958 | .TP |
959 | .BI allrandrepeat \fR=\fPbool | |
960 | Seed all random number generators in a predictable way so results are | |
523bad63 | 961 | repeatable across runs. Default: false. |
d60e92d1 | 962 | .TP |
04778baf JA |
963 | .BI randseed \fR=\fPint |
964 | Seed the random number generators based on this seed value, to be able to | |
965 | control what sequence of output is being generated. If not set, the random | |
966 | sequence depends on the \fBrandrepeat\fR setting. | |
967 | .TP | |
a596f047 | 968 | .BI fallocate \fR=\fPstr |
338f2db5 | 969 | Whether pre-allocation is performed when laying down files. |
523bad63 | 970 | Accepted values are: |
a596f047 EG |
971 | .RS |
972 | .RS | |
973 | .TP | |
974 | .B none | |
338f2db5 | 975 | Do not pre-allocate space. |
a596f047 | 976 | .TP |
2c3e17be | 977 | .B native |
338f2db5 | 978 | Use a platform's native pre-allocation call but fall back to |
523bad63 | 979 | \fBnone\fR behavior if it fails/is not implemented. |
2c3e17be | 980 | .TP |
a596f047 | 981 | .B posix |
338f2db5 | 982 | Pre-allocate via \fBposix_fallocate\fR\|(3). |
a596f047 EG |
983 | .TP |
984 | .B keep | |
338f2db5 | 985 | Pre-allocate via \fBfallocate\fR\|(2) with |
523bad63 | 986 | FALLOC_FL_KEEP_SIZE set. |
a596f047 | 987 | .TP |
38ca5f03 TV |
988 | .B truncate |
989 | Extend file to final size using \fBftruncate\fR|(2) | |
990 | instead of allocating. | |
991 | .TP | |
a596f047 | 992 | .B 0 |
338f2db5 | 993 | Backward-compatible alias for \fBnone\fR. |
a596f047 EG |
994 | .TP |
995 | .B 1 | |
338f2db5 | 996 | Backward-compatible alias for \fBposix\fR. |
a596f047 EG |
997 | .RE |
998 | .P | |
523bad63 TK |
999 | May not be available on all supported platforms. \fBkeep\fR is only available |
1000 | on Linux. If using ZFS on Solaris this cannot be set to \fBposix\fR | |
338f2db5 | 1001 | because ZFS doesn't support pre-allocation. Default: \fBnative\fR if any |
38ca5f03 TV |
1002 | pre-allocation methods except \fBtruncate\fR are available, \fBnone\fR if not. |
1003 | .P | |
1004 | Note that using \fBtruncate\fR on Windows will interact surprisingly | |
1005 | with non-sequential write patterns. When writing to a file that has | |
1006 | been extended by setting the end-of-file information, Windows will | |
1007 | backfill the unwritten portion of the file up to that offset with | |
1008 | zeroes before issuing the new write. This means that a single small | |
1009 | write to the end of an extended file will stall until the entire | |
1010 | file has been filled with zeroes. | |
a596f047 | 1011 | .RE |
7bc8c2cf | 1012 | .TP |
ecb2083d | 1013 | .BI fadvise_hint \fR=\fPstr |
c712c97a JA |
1014 | Use \fBposix_fadvise\fR\|(2) or \fBposix_madvise\fR\|(2) to advise the kernel |
1015 | what I/O patterns are likely to be issued. Accepted values are: | |
ecb2083d JA |
1016 | .RS |
1017 | .RS | |
1018 | .TP | |
1019 | .B 0 | |
1020 | Backwards compatible hint for "no hint". | |
1021 | .TP | |
1022 | .B 1 | |
1023 | Backwards compatible hint for "advise with fio workload type". This | |
523bad63 | 1024 | uses FADV_RANDOM for a random workload, and FADV_SEQUENTIAL |
ecb2083d JA |
1025 | for a sequential workload. |
1026 | .TP | |
1027 | .B sequential | |
523bad63 | 1028 | Advise using FADV_SEQUENTIAL. |
ecb2083d JA |
1029 | .TP |
1030 | .B random | |
523bad63 | 1031 | Advise using FADV_RANDOM. |
ecb2083d JA |
1032 | .RE |
1033 | .RE | |
d60e92d1 | 1034 | .TP |
8f4b9f24 | 1035 | .BI write_hint \fR=\fPstr |
523bad63 TK |
1036 | Use \fBfcntl\fR\|(2) to advise the kernel what life time to expect |
1037 | from a write. Only supported on Linux, as of version 4.13. Accepted | |
8f4b9f24 JA |
1038 | values are: |
1039 | .RS | |
1040 | .RS | |
1041 | .TP | |
1042 | .B none | |
1043 | No particular life time associated with this file. | |
1044 | .TP | |
1045 | .B short | |
1046 | Data written to this file has a short life time. | |
1047 | .TP | |
1048 | .B medium | |
1049 | Data written to this file has a medium life time. | |
1050 | .TP | |
1051 | .B long | |
1052 | Data written to this file has a long life time. | |
1053 | .TP | |
1054 | .B extreme | |
1055 | Data written to this file has a very long life time. | |
1056 | .RE | |
523bad63 TK |
1057 | .P |
1058 | The values are all relative to each other, and no absolute meaning | |
1059 | should be associated with them. | |
8f4b9f24 | 1060 | .RE |
37659335 | 1061 | .TP |
8f39afa7 | 1062 | .BI offset \fR=\fPint[%|z] |
523bad63 | 1063 | Start I/O at the provided offset in the file, given as either a fixed size in |
83c8b093 JF |
1064 | bytes or a percentage. If a percentage is given, the generated offset will be |
1065 | aligned to the minimum \fBblocksize\fR or to the value of \fBoffset_align\fR if | |
1066 | provided. Data before the given offset will not be touched. This | |
523bad63 TK |
1067 | effectively caps the file size at `real_size \- offset'. Can be combined with |
1068 | \fBsize\fR to constrain the start and end range of the I/O workload. | |
1069 | A percentage can be specified by a number between 1 and 100 followed by '%', | |
1070 | for example, `offset=20%' to specify 20%. | |
6d500c2e | 1071 | .TP |
83c8b093 JF |
1072 | .BI offset_align \fR=\fPint |
1073 | If set to non-zero value, the byte offset generated by a percentage \fBoffset\fR | |
1074 | is aligned upwards to this value. Defaults to 0 meaning that a percentage | |
1075 | offset is aligned to the minimum block size. | |
1076 | .TP | |
8f39afa7 | 1077 | .BI offset_increment \fR=\fPint[%|z] |
523bad63 TK |
1078 | If this is provided, then the real offset becomes `\fBoffset\fR + \fBoffset_increment\fR |
1079 | * thread_number', where the thread number is a counter that starts at 0 and | |
338f2db5 | 1080 | is incremented for each sub-job (i.e. when \fBnumjobs\fR option is |
523bad63 TK |
1081 | specified). This option is useful if there are several jobs which are |
1082 | intended to operate on a file in parallel disjoint segments, with even | |
0b288ba1 VF |
1083 | spacing between the starting points. Percentages can be used for this option. |
1084 | If a percentage is given, the generated offset will be aligned to the minimum | |
1085 | \fBblocksize\fR or to the value of \fBoffset_align\fR if provided. | |
6d500c2e | 1086 | .TP |
523bad63 TK |
1087 | .BI number_ios \fR=\fPint |
1088 | Fio will normally perform I/Os until it has exhausted the size of the region | |
1089 | set by \fBsize\fR, or if it exhaust the allocated time (or hits an error | |
1090 | condition). With this setting, the range/size can be set independently of | |
1091 | the number of I/Os to perform. When fio reaches this number, it will exit | |
1092 | normally and report status. Note that this does not extend the amount of I/O | |
1093 | that will be done, it will only stop fio if this condition is met before | |
338f2db5 | 1094 | other end-of-job criteria. |
d60e92d1 | 1095 | .TP |
523bad63 TK |
1096 | .BI fsync \fR=\fPint |
1097 | If writing to a file, issue an \fBfsync\fR\|(2) (or its equivalent) of | |
1098 | the dirty data for every number of blocks given. For example, if you give 32 | |
1099 | as a parameter, fio will sync the file after every 32 writes issued. If fio is | |
338f2db5 | 1100 | using non-buffered I/O, we may not sync the file. The exception is the sg |
523bad63 TK |
1101 | I/O engine, which synchronizes the disk cache anyway. Defaults to 0, which |
1102 | means fio does not periodically issue and wait for a sync to complete. Also | |
1103 | see \fBend_fsync\fR and \fBfsync_on_close\fR. | |
6d500c2e | 1104 | .TP |
523bad63 TK |
1105 | .BI fdatasync \fR=\fPint |
1106 | Like \fBfsync\fR but uses \fBfdatasync\fR\|(2) to only sync data and | |
44f668d7 | 1107 | not metadata blocks. In Windows, FreeBSD, DragonFlyBSD or OSX there is no |
523bad63 TK |
1108 | \fBfdatasync\fR\|(2) so this falls back to using \fBfsync\fR\|(2). |
1109 | Defaults to 0, which means fio does not periodically issue and wait for a | |
338f2db5 | 1110 | data-only sync to complete. |
d60e92d1 | 1111 | .TP |
523bad63 TK |
1112 | .BI write_barrier \fR=\fPint |
1113 | Make every N\-th write a barrier write. | |
901bb994 | 1114 | .TP |
523bad63 TK |
1115 | .BI sync_file_range \fR=\fPstr:int |
1116 | Use \fBsync_file_range\fR\|(2) for every \fIint\fR number of write | |
1117 | operations. Fio will track range of writes that have happened since the last | |
1118 | \fBsync_file_range\fR\|(2) call. \fIstr\fR can currently be one or more of: | |
1119 | .RS | |
1120 | .RS | |
fd68418e | 1121 | .TP |
523bad63 TK |
1122 | .B wait_before |
1123 | SYNC_FILE_RANGE_WAIT_BEFORE | |
c5751c62 | 1124 | .TP |
523bad63 TK |
1125 | .B write |
1126 | SYNC_FILE_RANGE_WRITE | |
c5751c62 | 1127 | .TP |
523bad63 TK |
1128 | .B wait_after |
1129 | SYNC_FILE_RANGE_WRITE_AFTER | |
2fa5a241 | 1130 | .RE |
523bad63 TK |
1131 | .P |
1132 | So if you do `sync_file_range=wait_before,write:8', fio would use | |
1133 | `SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE' for every 8 | |
1134 | writes. Also see the \fBsync_file_range\fR\|(2) man page. This option is | |
1135 | Linux specific. | |
2fa5a241 | 1136 | .RE |
ce35b1ec | 1137 | .TP |
523bad63 TK |
1138 | .BI overwrite \fR=\fPbool |
1139 | If true, writes to a file will always overwrite existing data. If the file | |
1140 | doesn't already exist, it will be created before the write phase begins. If | |
1141 | the file exists and is large enough for the specified write phase, nothing | |
1142 | will be done. Default: false. | |
5c94b008 | 1143 | .TP |
523bad63 TK |
1144 | .BI end_fsync \fR=\fPbool |
1145 | If true, \fBfsync\fR\|(2) file contents when a write stage has completed. | |
1146 | Default: false. | |
d60e92d1 | 1147 | .TP |
523bad63 TK |
1148 | .BI fsync_on_close \fR=\fPbool |
1149 | If true, fio will \fBfsync\fR\|(2) a dirty file on close. This differs | |
1150 | from \fBend_fsync\fR in that it will happen on every file close, not | |
1151 | just at the end of the job. Default: false. | |
d60e92d1 | 1152 | .TP |
523bad63 TK |
1153 | .BI rwmixread \fR=\fPint |
1154 | Percentage of a mixed workload that should be reads. Default: 50. | |
1155 | .TP | |
1156 | .BI rwmixwrite \fR=\fPint | |
1157 | Percentage of a mixed workload that should be writes. If both | |
1158 | \fBrwmixread\fR and \fBrwmixwrite\fR is given and the values do not | |
1159 | add up to 100%, the latter of the two will be used to override the | |
1160 | first. This may interfere with a given rate setting, if fio is asked to | |
1161 | limit reads or writes to a certain rate. If that is the case, then the | |
1162 | distribution may be skewed. Default: 50. | |
1163 | .TP | |
a87c90fd | 1164 | .BI random_distribution \fR=\fPstr:float[:float][,str:float][,str:float] |
523bad63 TK |
1165 | By default, fio will use a completely uniform random distribution when asked |
1166 | to perform random I/O. Sometimes it is useful to skew the distribution in | |
1167 | specific ways, ensuring that some parts of the data is more hot than others. | |
1168 | fio includes the following distribution models: | |
d60e92d1 AC |
1169 | .RS |
1170 | .RS | |
1171 | .TP | |
1172 | .B random | |
523bad63 | 1173 | Uniform random distribution |
8c07860d JA |
1174 | .TP |
1175 | .B zipf | |
523bad63 | 1176 | Zipf distribution |
8c07860d JA |
1177 | .TP |
1178 | .B pareto | |
523bad63 | 1179 | Pareto distribution |
8c07860d | 1180 | .TP |
dd3503d3 | 1181 | .B normal |
523bad63 | 1182 | Normal (Gaussian) distribution |
dd3503d3 | 1183 | .TP |
523bad63 TK |
1184 | .B zoned |
1185 | Zoned random distribution | |
59466396 JA |
1186 | .B zoned_abs |
1187 | Zoned absolute random distribution | |
d60e92d1 AC |
1188 | .RE |
1189 | .P | |
523bad63 TK |
1190 | When using a \fBzipf\fR or \fBpareto\fR distribution, an input value is also |
1191 | needed to define the access pattern. For \fBzipf\fR, this is the `Zipf theta'. | |
1192 | For \fBpareto\fR, it's the `Pareto power'. Fio includes a test | |
1193 | program, \fBfio\-genzipf\fR, that can be used visualize what the given input | |
1194 | values will yield in terms of hit rates. If you wanted to use \fBzipf\fR with | |
1195 | a `theta' of 1.2, you would use `random_distribution=zipf:1.2' as the | |
1196 | option. If a non\-uniform model is used, fio will disable use of the random | |
1197 | map. For the \fBnormal\fR distribution, a normal (Gaussian) deviation is | |
1198 | supplied as a value between 0 and 100. | |
1199 | .P | |
a87c90fd AK |
1200 | The second, optional float is allowed for \fBpareto\fR, \fBzipf\fR and \fBnormal\fR |
1201 | distributions. It allows to set base of distribution in non-default place, giving | |
1202 | more control over most probable outcome. This value is in range [0-1] which maps linearly to | |
1203 | range of possible random values. | |
1204 | Defaults are: random for \fBpareto\fR and \fBzipf\fR, and 0.5 for \fBnormal\fR. | |
1205 | If you wanted to use \fBzipf\fR with a `theta` of 1.2 centered on 1/4 of allowed value range, | |
1206 | you would use `random_distibution=zipf:1.2:0.25`. | |
1207 | .P | |
523bad63 TK |
1208 | For a \fBzoned\fR distribution, fio supports specifying percentages of I/O |
1209 | access that should fall within what range of the file or device. For | |
1210 | example, given a criteria of: | |
d60e92d1 | 1211 | .RS |
523bad63 TK |
1212 | .P |
1213 | .PD 0 | |
1214 | 60% of accesses should be to the first 10% | |
1215 | .P | |
1216 | 30% of accesses should be to the next 20% | |
1217 | .P | |
1218 | 8% of accesses should be to the next 30% | |
1219 | .P | |
1220 | 2% of accesses should be to the next 40% | |
1221 | .PD | |
1222 | .RE | |
1223 | .P | |
1224 | we can define that through zoning of the random accesses. For the above | |
1225 | example, the user would do: | |
1226 | .RS | |
1227 | .P | |
1228 | random_distribution=zoned:60/10:30/20:8/30:2/40 | |
1229 | .RE | |
1230 | .P | |
59466396 JA |
1231 | A \fBzoned_abs\fR distribution works exactly like the\fBzoned\fR, except that |
1232 | it takes absolute sizes. For example, let's say you wanted to define access | |
1233 | according to the following criteria: | |
1234 | .RS | |
1235 | .P | |
1236 | .PD 0 | |
1237 | 60% of accesses should be to the first 20G | |
1238 | .P | |
1239 | 30% of accesses should be to the next 100G | |
1240 | .P | |
1241 | 10% of accesses should be to the next 500G | |
1242 | .PD | |
1243 | .RE | |
1244 | .P | |
1245 | we can define an absolute zoning distribution with: | |
1246 | .RS | |
1247 | .P | |
1248 | random_distribution=zoned:60/10:30/20:8/30:2/40 | |
1249 | .RE | |
1250 | .P | |
6a16ece8 JA |
1251 | For both \fBzoned\fR and \fBzoned_abs\fR, fio supports defining up to 256 |
1252 | separate zones. | |
1253 | .P | |
59466396 | 1254 | Similarly to how \fBbssplit\fR works for setting ranges and percentages |
523bad63 TK |
1255 | of block sizes. Like \fBbssplit\fR, it's possible to specify separate |
1256 | zones for reads, writes, and trims. If just one set is given, it'll apply to | |
1257 | all of them. | |
1258 | .RE | |
1259 | .TP | |
1260 | .BI percentage_random \fR=\fPint[,int][,int] | |
1261 | For a random workload, set how big a percentage should be random. This | |
1262 | defaults to 100%, in which case the workload is fully random. It can be set | |
1263 | from anywhere from 0 to 100. Setting it to 0 would make the workload fully | |
1264 | sequential. Any setting in between will result in a random mix of sequential | |
338f2db5 | 1265 | and random I/O, at the given percentages. Comma-separated values may be |
523bad63 TK |
1266 | specified for reads, writes, and trims as described in \fBblocksize\fR. |
1267 | .TP | |
1268 | .BI norandommap | |
1269 | Normally fio will cover every block of the file when doing random I/O. If | |
1270 | this option is given, fio will just get a new random offset without looking | |
1271 | at past I/O history. This means that some blocks may not be read or written, | |
1272 | and that some blocks may be read/written more than once. If this option is | |
1273 | used with \fBverify\fR and multiple blocksizes (via \fBbsrange\fR), | |
338f2db5 | 1274 | only intact blocks are verified, i.e., partially-overwritten blocks are |
47e6a6e5 SW |
1275 | ignored. With an async I/O engine and an I/O depth > 1, it is possible for |
1276 | the same block to be overwritten, which can cause verification errors. Either | |
1277 | do not use norandommap in this case, or also use the lfsr random generator. | |
523bad63 TK |
1278 | .TP |
1279 | .BI softrandommap \fR=\fPbool | |
1280 | See \fBnorandommap\fR. If fio runs with the random block map enabled and | |
1281 | it fails to allocate the map, if this option is set it will continue without | |
1282 | a random block map. As coverage will not be as complete as with random maps, | |
1283 | this option is disabled by default. | |
1284 | .TP | |
1285 | .BI random_generator \fR=\fPstr | |
1286 | Fio supports the following engines for generating I/O offsets for random I/O: | |
1287 | .RS | |
1288 | .RS | |
1289 | .TP | |
1290 | .B tausworthe | |
1291 | Strong 2^88 cycle random number generator. | |
1292 | .TP | |
1293 | .B lfsr | |
1294 | Linear feedback shift register generator. | |
1295 | .TP | |
1296 | .B tausworthe64 | |
1297 | Strong 64\-bit 2^258 cycle random number generator. | |
1298 | .RE | |
1299 | .P | |
1300 | \fBtausworthe\fR is a strong random number generator, but it requires tracking | |
1301 | on the side if we want to ensure that blocks are only read or written | |
1302 | once. \fBlfsr\fR guarantees that we never generate the same offset twice, and | |
1303 | it's also less computationally expensive. It's not a true random generator, | |
1304 | however, though for I/O purposes it's typically good enough. \fBlfsr\fR only | |
1305 | works with single block sizes, not with workloads that use multiple block | |
1306 | sizes. If used with such a workload, fio may read or write some blocks | |
1307 | multiple times. The default value is \fBtausworthe\fR, unless the required | |
1308 | space exceeds 2^32 blocks. If it does, then \fBtausworthe64\fR is | |
1309 | selected automatically. | |
1310 | .RE | |
1311 | .SS "Block size" | |
1312 | .TP | |
1313 | .BI blocksize \fR=\fPint[,int][,int] "\fR,\fB bs" \fR=\fPint[,int][,int] | |
1314 | The block size in bytes used for I/O units. Default: 4096. A single value | |
338f2db5 | 1315 | applies to reads, writes, and trims. Comma-separated values may be |
523bad63 TK |
1316 | specified for reads, writes, and trims. A value not terminated in a comma |
1317 | applies to subsequent types. Examples: | |
1318 | .RS | |
1319 | .RS | |
1320 | .P | |
1321 | .PD 0 | |
1322 | bs=256k means 256k for reads, writes and trims. | |
1323 | .P | |
1324 | bs=8k,32k means 8k for reads, 32k for writes and trims. | |
1325 | .P | |
1326 | bs=8k,32k, means 8k for reads, 32k for writes, and default for trims. | |
1327 | .P | |
1328 | bs=,8k means default for reads, 8k for writes and trims. | |
1329 | .P | |
1330 | bs=,8k, means default for reads, 8k for writes, and default for trims. | |
1331 | .PD | |
1332 | .RE | |
1333 | .RE | |
1334 | .TP | |
1335 | .BI blocksize_range \fR=\fPirange[,irange][,irange] "\fR,\fB bsrange" \fR=\fPirange[,irange][,irange] | |
1336 | A range of block sizes in bytes for I/O units. The issued I/O unit will | |
1337 | always be a multiple of the minimum size, unless | |
1338 | \fBblocksize_unaligned\fR is set. | |
338f2db5 | 1339 | Comma-separated ranges may be specified for reads, writes, and trims as |
523bad63 TK |
1340 | described in \fBblocksize\fR. Example: |
1341 | .RS | |
1342 | .RS | |
1343 | .P | |
1344 | bsrange=1k\-4k,2k\-8k | |
1345 | .RE | |
1346 | .RE | |
1347 | .TP | |
1348 | .BI bssplit \fR=\fPstr[,str][,str] | |
1349 | Sometimes you want even finer grained control of the block sizes issued, not | |
1350 | just an even split between them. This option allows you to weight various | |
1351 | block sizes, so that you are able to define a specific amount of block sizes | |
1352 | issued. The format for this option is: | |
1353 | .RS | |
1354 | .RS | |
1355 | .P | |
1356 | bssplit=blocksize/percentage:blocksize/percentage | |
1357 | .RE | |
1358 | .P | |
1359 | for as many block sizes as needed. So if you want to define a workload that | |
1360 | has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would write: | |
1361 | .RS | |
1362 | .P | |
1363 | bssplit=4k/10:64k/50:32k/40 | |
1364 | .RE | |
1365 | .P | |
1366 | Ordering does not matter. If the percentage is left blank, fio will fill in | |
1367 | the remaining values evenly. So a bssplit option like this one: | |
1368 | .RS | |
1369 | .P | |
1370 | bssplit=4k/50:1k/:32k/ | |
1371 | .RE | |
1372 | .P | |
1373 | would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always add up | |
1374 | to 100, if bssplit is given a range that adds up to more, it will error out. | |
1375 | .P | |
338f2db5 | 1376 | Comma-separated values may be specified for reads, writes, and trims as |
523bad63 TK |
1377 | described in \fBblocksize\fR. |
1378 | .P | |
1379 | If you want a workload that has 50% 2k reads and 50% 4k reads, while having | |
1380 | 90% 4k writes and 10% 8k writes, you would specify: | |
1381 | .RS | |
1382 | .P | |
cf04b906 | 1383 | bssplit=2k/50:4k/50,4k/90:8k/10 |
523bad63 | 1384 | .RE |
6a16ece8 JA |
1385 | .P |
1386 | Fio supports defining up to 64 different weights for each data direction. | |
523bad63 TK |
1387 | .RE |
1388 | .TP | |
1389 | .BI blocksize_unaligned "\fR,\fB bs_unaligned" | |
1390 | If set, fio will issue I/O units with any size within | |
1391 | \fBblocksize_range\fR, not just multiples of the minimum size. This | |
1392 | typically won't work with direct I/O, as that normally requires sector | |
1393 | alignment. | |
1394 | .TP | |
1395 | .BI bs_is_seq_rand \fR=\fPbool | |
1396 | If this option is set, fio will use the normal read,write blocksize settings | |
1397 | as sequential,random blocksize settings instead. Any random read or write | |
1398 | will use the WRITE blocksize settings, and any sequential read or write will | |
1399 | use the READ blocksize settings. | |
1400 | .TP | |
1401 | .BI blockalign \fR=\fPint[,int][,int] "\fR,\fB ba" \fR=\fPint[,int][,int] | |
1402 | Boundary to which fio will align random I/O units. Default: | |
1403 | \fBblocksize\fR. Minimum alignment is typically 512b for using direct | |
1404 | I/O, though it usually depends on the hardware block size. This option is | |
1405 | mutually exclusive with using a random map for files, so it will turn off | |
338f2db5 | 1406 | that option. Comma-separated values may be specified for reads, writes, and |
523bad63 TK |
1407 | trims as described in \fBblocksize\fR. |
1408 | .SS "Buffers and memory" | |
1409 | .TP | |
1410 | .BI zero_buffers | |
1411 | Initialize buffers with all zeros. Default: fill buffers with random data. | |
1412 | .TP | |
1413 | .BI refill_buffers | |
1414 | If this option is given, fio will refill the I/O buffers on every | |
1415 | submit. The default is to only fill it at init time and reuse that | |
1416 | data. Only makes sense if zero_buffers isn't specified, naturally. If data | |
1417 | verification is enabled, \fBrefill_buffers\fR is also automatically enabled. | |
1418 | .TP | |
1419 | .BI scramble_buffers \fR=\fPbool | |
1420 | If \fBrefill_buffers\fR is too costly and the target is using data | |
1421 | deduplication, then setting this option will slightly modify the I/O buffer | |
338f2db5 | 1422 | contents to defeat normal de-dupe attempts. This is not enough to defeat |
523bad63 TK |
1423 | more clever block compression attempts, but it will stop naive dedupe of |
1424 | blocks. Default: true. | |
1425 | .TP | |
1426 | .BI buffer_compress_percentage \fR=\fPint | |
72592780 SW |
1427 | If this is set, then fio will attempt to provide I/O buffer content |
1428 | (on WRITEs) that compresses to the specified level. Fio does this by | |
1429 | providing a mix of random data followed by fixed pattern data. The | |
1430 | fixed pattern is either zeros, or the pattern specified by | |
1431 | \fBbuffer_pattern\fR. If the \fBbuffer_pattern\fR option is used, it | |
1432 | might skew the compression ratio slightly. Setting | |
1433 | \fBbuffer_compress_percentage\fR to a value other than 100 will also | |
1434 | enable \fBrefill_buffers\fR in order to reduce the likelihood that | |
1435 | adjacent blocks are so similar that they over compress when seen | |
1436 | together. See \fBbuffer_compress_chunk\fR for how to set a finer or | |
1437 | coarser granularity of the random/fixed data regions. Defaults to unset | |
1438 | i.e., buffer data will not adhere to any compression level. | |
523bad63 TK |
1439 | .TP |
1440 | .BI buffer_compress_chunk \fR=\fPint | |
72592780 SW |
1441 | This setting allows fio to manage how big the random/fixed data region |
1442 | is when using \fBbuffer_compress_percentage\fR. When | |
1443 | \fBbuffer_compress_chunk\fR is set to some non-zero value smaller than the | |
1444 | block size, fio can repeat the random/fixed region throughout the I/O | |
1445 | buffer at the specified interval (which particularly useful when | |
1446 | bigger block sizes are used for a job). When set to 0, fio will use a | |
1447 | chunk size that matches the block size resulting in a single | |
1448 | random/fixed region within the I/O buffer. Defaults to 512. When the | |
1449 | unit is omitted, the value is interpreted in bytes. | |
523bad63 TK |
1450 | .TP |
1451 | .BI buffer_pattern \fR=\fPstr | |
1452 | If set, fio will fill the I/O buffers with this pattern or with the contents | |
1453 | of a file. If not set, the contents of I/O buffers are defined by the other | |
1454 | options related to buffer contents. The setting can be any pattern of bytes, | |
1455 | and can be prefixed with 0x for hex values. It may also be a string, where | |
1456 | the string must then be wrapped with "". Or it may also be a filename, | |
1457 | where the filename must be wrapped with '' in which case the file is | |
1458 | opened and read. Note that not all the file contents will be read if that | |
1459 | would cause the buffers to overflow. So, for example: | |
1460 | .RS | |
1461 | .RS | |
1462 | .P | |
1463 | .PD 0 | |
1464 | buffer_pattern='filename' | |
1465 | .P | |
1466 | or: | |
1467 | .P | |
1468 | buffer_pattern="abcd" | |
1469 | .P | |
1470 | or: | |
1471 | .P | |
1472 | buffer_pattern=\-12 | |
1473 | .P | |
1474 | or: | |
1475 | .P | |
1476 | buffer_pattern=0xdeadface | |
1477 | .PD | |
1478 | .RE | |
1479 | .P | |
1480 | Also you can combine everything together in any order: | |
1481 | .RS | |
1482 | .P | |
1483 | buffer_pattern=0xdeadface"abcd"\-12'filename' | |
1484 | .RE | |
1485 | .RE | |
1486 | .TP | |
1487 | .BI dedupe_percentage \fR=\fPint | |
1488 | If set, fio will generate this percentage of identical buffers when | |
1489 | writing. These buffers will be naturally dedupable. The contents of the | |
1490 | buffers depend on what other buffer compression settings have been set. It's | |
1491 | possible to have the individual buffers either fully compressible, or not at | |
72592780 SW |
1492 | all \-\- this option only controls the distribution of unique buffers. Setting |
1493 | this option will also enable \fBrefill_buffers\fR to prevent every buffer | |
1494 | being identical. | |
523bad63 TK |
1495 | .TP |
1496 | .BI invalidate \fR=\fPbool | |
1497 | Invalidate the buffer/page cache parts of the files to be used prior to | |
1498 | starting I/O if the platform and file type support it. Defaults to true. | |
1499 | This will be ignored if \fBpre_read\fR is also specified for the | |
1500 | same job. | |
1501 | .TP | |
eb9f8d7f AF |
1502 | .BI sync \fR=\fPstr |
1503 | Whether, and what type, of synchronous I/O to use for writes. The allowed | |
1504 | values are: | |
1505 | .RS | |
1506 | .RS | |
1507 | .TP | |
1508 | .B none | |
1509 | Do not use synchronous IO, the default. | |
1510 | .TP | |
1511 | .B 0 | |
1512 | Same as \fBnone\fR. | |
1513 | .TP | |
1514 | .B sync | |
1515 | Use synchronous file IO. For the majority of I/O engines, | |
1516 | this means using O_SYNC. | |
1517 | .TP | |
1518 | .B 1 | |
1519 | Same as \fBsync\fR. | |
1520 | .TP | |
1521 | .B dsync | |
1522 | Use synchronous data IO. For the majority of I/O engines, | |
1523 | this means using O_DSYNC. | |
1524 | .PD | |
1525 | .RE | |
1526 | .RE | |
523bad63 TK |
1527 | .TP |
1528 | .BI iomem \fR=\fPstr "\fR,\fP mem" \fR=\fPstr | |
1529 | Fio can use various types of memory as the I/O unit buffer. The allowed | |
1530 | values are: | |
1531 | .RS | |
1532 | .RS | |
1533 | .TP | |
1534 | .B malloc | |
1535 | Use memory from \fBmalloc\fR\|(3) as the buffers. Default memory type. | |
1536 | .TP | |
1537 | .B shm | |
1538 | Use shared memory as the buffers. Allocated through \fBshmget\fR\|(2). | |
1539 | .TP | |
1540 | .B shmhuge | |
1541 | Same as \fBshm\fR, but use huge pages as backing. | |
1542 | .TP | |
1543 | .B mmap | |
1544 | Use \fBmmap\fR\|(2) to allocate buffers. May either be anonymous memory, or can | |
1545 | be file backed if a filename is given after the option. The format | |
1546 | is `mem=mmap:/path/to/file'. | |
1547 | .TP | |
1548 | .B mmaphuge | |
1549 | Use a memory mapped huge file as the buffer backing. Append filename | |
1550 | after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file'. | |
1551 | .TP | |
1552 | .B mmapshared | |
1553 | Same as \fBmmap\fR, but use a MMAP_SHARED mapping. | |
1554 | .TP | |
1555 | .B cudamalloc | |
1556 | Use GPU memory as the buffers for GPUDirect RDMA benchmark. | |
1557 | The \fBioengine\fR must be \fBrdma\fR. | |
1558 | .RE | |
1559 | .P | |
1560 | The area allocated is a function of the maximum allowed bs size for the job, | |
1561 | multiplied by the I/O depth given. Note that for \fBshmhuge\fR and | |
1562 | \fBmmaphuge\fR to work, the system must have free huge pages allocated. This | |
1563 | can normally be checked and set by reading/writing | |
1564 | `/proc/sys/vm/nr_hugepages' on a Linux system. Fio assumes a huge page | |
1565 | is 4MiB in size. So to calculate the number of huge pages you need for a | |
1566 | given job file, add up the I/O depth of all jobs (normally one unless | |
1567 | \fBiodepth\fR is used) and multiply by the maximum bs set. Then divide | |
1568 | that number by the huge page size. You can see the size of the huge pages in | |
338f2db5 | 1569 | `/proc/meminfo'. If no huge pages are allocated by having a non-zero |
523bad63 TK |
1570 | number in `nr_hugepages', using \fBmmaphuge\fR or \fBshmhuge\fR will fail. Also |
1571 | see \fBhugepage\-size\fR. | |
1572 | .P | |
1573 | \fBmmaphuge\fR also needs to have hugetlbfs mounted and the file location | |
1574 | should point there. So if it's mounted in `/huge', you would use | |
1575 | `mem=mmaphuge:/huge/somefile'. | |
1576 | .RE | |
1577 | .TP | |
1578 | .BI iomem_align \fR=\fPint "\fR,\fP mem_align" \fR=\fPint | |
1579 | This indicates the memory alignment of the I/O memory buffers. Note that | |
1580 | the given alignment is applied to the first I/O unit buffer, if using | |
1581 | \fBiodepth\fR the alignment of the following buffers are given by the | |
1582 | \fBbs\fR used. In other words, if using a \fBbs\fR that is a | |
1583 | multiple of the page sized in the system, all buffers will be aligned to | |
1584 | this value. If using a \fBbs\fR that is not page aligned, the alignment | |
1585 | of subsequent I/O memory buffers is the sum of the \fBiomem_align\fR and | |
1586 | \fBbs\fR used. | |
1587 | .TP | |
1588 | .BI hugepage\-size \fR=\fPint | |
1589 | Defines the size of a huge page. Must at least be equal to the system | |
1590 | setting, see `/proc/meminfo'. Defaults to 4MiB. Should probably | |
1591 | always be a multiple of megabytes, so using `hugepage\-size=Xm' is the | |
338f2db5 | 1592 | preferred way to set this to avoid setting a non-pow-2 bad value. |
523bad63 TK |
1593 | .TP |
1594 | .BI lockmem \fR=\fPint | |
1595 | Pin the specified amount of memory with \fBmlock\fR\|(2). Can be used to | |
1596 | simulate a smaller amount of memory. The amount specified is per worker. | |
1597 | .SS "I/O size" | |
1598 | .TP | |
8f39afa7 | 1599 | .BI size \fR=\fPint[%|z] |
523bad63 TK |
1600 | The total size of file I/O for each thread of this job. Fio will run until |
1601 | this many bytes has been transferred, unless runtime is limited by other options | |
1602 | (such as \fBruntime\fR, for instance, or increased/decreased by \fBio_size\fR). | |
1603 | Fio will divide this size between the available files determined by options | |
1604 | such as \fBnrfiles\fR, \fBfilename\fR, unless \fBfilesize\fR is | |
1605 | specified by the job. If the result of division happens to be 0, the size is | |
1606 | set to the physical size of the given files or devices if they exist. | |
1607 | If this option is not specified, fio will use the full size of the given | |
1608 | files or devices. If the files do not exist, size must be given. It is also | |
1609 | possible to give size as a percentage between 1 and 100. If `size=20%' is | |
1610 | given, fio will use 20% of the full size of the given files or devices. | |
1611 | Can be combined with \fBoffset\fR to constrain the start and end range | |
1612 | that I/O will be done within. | |
1613 | .TP | |
8f39afa7 | 1614 | .BI io_size \fR=\fPint[%|z] "\fR,\fB io_limit" \fR=\fPint[%|z] |
523bad63 TK |
1615 | Normally fio operates within the region set by \fBsize\fR, which means |
1616 | that the \fBsize\fR option sets both the region and size of I/O to be | |
1617 | performed. Sometimes that is not what you want. With this option, it is | |
1618 | possible to define just the amount of I/O that fio should do. For instance, | |
1619 | if \fBsize\fR is set to 20GiB and \fBio_size\fR is set to 5GiB, fio | |
1620 | will perform I/O within the first 20GiB but exit when 5GiB have been | |
1621 | done. The opposite is also possible \-\- if \fBsize\fR is set to 20GiB, | |
1622 | and \fBio_size\fR is set to 40GiB, then fio will do 40GiB of I/O within | |
f248a525 AD |
1623 | the 0..20GiB region. Value can be set as percentage: \fBio_size\fR=N%. |
1624 | In this case \fBio_size\fR multiplies \fBsize\fR= value. | |
523bad63 TK |
1625 | .TP |
1626 | .BI filesize \fR=\fPirange(int) | |
1627 | Individual file sizes. May be a range, in which case fio will select sizes | |
1628 | for files at random within the given range and limited to \fBsize\fR in | |
1629 | total (if that is given). If not given, each created file is the same size. | |
1630 | This option overrides \fBsize\fR in terms of file size, which means | |
1631 | this value is used as a fixed size or possible range of each file. | |
1632 | .TP | |
1633 | .BI file_append \fR=\fPbool | |
1634 | Perform I/O after the end of the file. Normally fio will operate within the | |
1635 | size of a file. If this option is set, then fio will append to the file | |
1636 | instead. This has identical behavior to setting \fBoffset\fR to the size | |
338f2db5 | 1637 | of a file. This option is ignored on non-regular files. |
523bad63 TK |
1638 | .TP |
1639 | .BI fill_device \fR=\fPbool "\fR,\fB fill_fs" \fR=\fPbool | |
1640 | Sets size to something really large and waits for ENOSPC (no space left on | |
1641 | device) as the terminating condition. Only makes sense with sequential | |
1642 | write. For a read workload, the mount point will be filled first then I/O | |
1643 | started on the result. This option doesn't make sense if operating on a raw | |
1644 | device node, since the size of that is already known by the file system. | |
338f2db5 | 1645 | Additionally, writing beyond end-of-device will not return ENOSPC there. |
523bad63 TK |
1646 | .SS "I/O engine" |
1647 | .TP | |
1648 | .BI ioengine \fR=\fPstr | |
1649 | Defines how the job issues I/O to the file. The following types are defined: | |
1650 | .RS | |
1651 | .RS | |
1652 | .TP | |
1653 | .B sync | |
1654 | Basic \fBread\fR\|(2) or \fBwrite\fR\|(2) | |
1655 | I/O. \fBlseek\fR\|(2) is used to position the I/O location. | |
1656 | See \fBfsync\fR and \fBfdatasync\fR for syncing write I/Os. | |
1657 | .TP | |
1658 | .B psync | |
1659 | Basic \fBpread\fR\|(2) or \fBpwrite\fR\|(2) I/O. Default on | |
1660 | all supported operating systems except for Windows. | |
1661 | .TP | |
1662 | .B vsync | |
1663 | Basic \fBreadv\fR\|(2) or \fBwritev\fR\|(2) I/O. Will emulate | |
1664 | queuing by coalescing adjacent I/Os into a single submission. | |
1665 | .TP | |
1666 | .B pvsync | |
1667 | Basic \fBpreadv\fR\|(2) or \fBpwritev\fR\|(2) I/O. | |
a46c5e01 | 1668 | .TP |
2cafffbe JA |
1669 | .B pvsync2 |
1670 | Basic \fBpreadv2\fR\|(2) or \fBpwritev2\fR\|(2) I/O. | |
1671 | .TP | |
d60e92d1 | 1672 | .B libaio |
523bad63 | 1673 | Linux native asynchronous I/O. Note that Linux may only support |
338f2db5 | 1674 | queued behavior with non-buffered I/O (set `direct=1' or |
523bad63 TK |
1675 | `buffered=0'). |
1676 | This engine defines engine specific options. | |
d60e92d1 AC |
1677 | .TP |
1678 | .B posixaio | |
523bad63 TK |
1679 | POSIX asynchronous I/O using \fBaio_read\fR\|(3) and |
1680 | \fBaio_write\fR\|(3). | |
03e20d68 BC |
1681 | .TP |
1682 | .B solarisaio | |
1683 | Solaris native asynchronous I/O. | |
1684 | .TP | |
1685 | .B windowsaio | |
38f8c318 | 1686 | Windows native asynchronous I/O. Default on Windows. |
d60e92d1 AC |
1687 | .TP |
1688 | .B mmap | |
523bad63 TK |
1689 | File is memory mapped with \fBmmap\fR\|(2) and data copied |
1690 | to/from using \fBmemcpy\fR\|(3). | |
d60e92d1 AC |
1691 | .TP |
1692 | .B splice | |
523bad63 TK |
1693 | \fBsplice\fR\|(2) is used to transfer the data and |
1694 | \fBvmsplice\fR\|(2) to transfer data from user space to the | |
1695 | kernel. | |
d60e92d1 | 1696 | .TP |
d60e92d1 | 1697 | .B sg |
523bad63 TK |
1698 | SCSI generic sg v3 I/O. May either be synchronous using the SG_IO |
1699 | ioctl, or if the target is an sg character device we use | |
1700 | \fBread\fR\|(2) and \fBwrite\fR\|(2) for asynchronous | |
1701 | I/O. Requires \fBfilename\fR option to specify either block or | |
3740cfc8 VF |
1702 | character devices. This engine supports trim operations. The |
1703 | sg engine includes engine specific options. | |
d60e92d1 | 1704 | .TP |
56a19325 DF |
1705 | .B libzbc |
1706 | Synchronous I/O engine for SMR hard-disks using the \fBlibzbc\fR | |
1707 | library. The target can be either an sg character device or | |
1708 | a block device file. This engine supports the zonemode=zbd zone | |
1709 | operations. | |
1710 | .TP | |
d60e92d1 | 1711 | .B null |
523bad63 TK |
1712 | Doesn't transfer any data, just pretends to. This is mainly used to |
1713 | exercise fio itself and for debugging/testing purposes. | |
d60e92d1 AC |
1714 | .TP |
1715 | .B net | |
523bad63 TK |
1716 | Transfer over the network to given `host:port'. Depending on the |
1717 | \fBprotocol\fR used, the \fBhostname\fR, \fBport\fR, | |
1718 | \fBlisten\fR and \fBfilename\fR options are used to specify | |
1719 | what sort of connection to make, while the \fBprotocol\fR option | |
1720 | determines which protocol will be used. This engine defines engine | |
1721 | specific options. | |
d60e92d1 AC |
1722 | .TP |
1723 | .B netsplice | |
523bad63 TK |
1724 | Like \fBnet\fR, but uses \fBsplice\fR\|(2) and |
1725 | \fBvmsplice\fR\|(2) to map data and send/receive. | |
1726 | This engine defines engine specific options. | |
d60e92d1 | 1727 | .TP |
53aec0a4 | 1728 | .B cpuio |
523bad63 | 1729 | Doesn't transfer any data, but burns CPU cycles according to the |
9de473a8 EV |
1730 | \fBcpuload\fR, \fBcpuchunks\fR and \fBcpumode\fR options. |
1731 | A job never finishes unless there is at least one non-cpuio job. | |
1732 | .RS | |
1733 | .P | |
1734 | .PD 0 | |
1735 | \fBcpuload\fR\=85 will cause that job to do nothing but burn 85% of the CPU. | |
1736 | In case of SMP machines, use \fBnumjobs=<nr_of_cpu>\fR\ to get desired CPU usage, | |
1737 | as the cpuload only loads a single CPU at the desired rate. | |
1738 | ||
1739 | .P | |
1740 | \fBcpumode\fR\=qsort replace the default noop instructions loop | |
1741 | by a qsort algorithm to consume more energy. | |
1742 | ||
1743 | .P | |
1744 | .RE | |
d60e92d1 | 1745 | .TP |
21b8aee8 | 1746 | .B rdma |
523bad63 TK |
1747 | The RDMA I/O engine supports both RDMA memory semantics |
1748 | (RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the | |
609ac152 SB |
1749 | InfiniBand, RoCE and iWARP protocols. This engine defines engine |
1750 | specific options. | |
d54fce84 DM |
1751 | .TP |
1752 | .B falloc | |
523bad63 TK |
1753 | I/O engine that does regular fallocate to simulate data transfer as |
1754 | fio ioengine. | |
1755 | .RS | |
1756 | .P | |
1757 | .PD 0 | |
1758 | DDIR_READ does fallocate(,mode = FALLOC_FL_KEEP_SIZE,). | |
1759 | .P | |
1760 | DIR_WRITE does fallocate(,mode = 0). | |
1761 | .P | |
1762 | DDIR_TRIM does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE). | |
1763 | .PD | |
1764 | .RE | |
1765 | .TP | |
1766 | .B ftruncate | |
1767 | I/O engine that sends \fBftruncate\fR\|(2) operations in response | |
1768 | to write (DDIR_WRITE) events. Each ftruncate issued sets the file's | |
1769 | size to the current block offset. \fBblocksize\fR is ignored. | |
d54fce84 DM |
1770 | .TP |
1771 | .B e4defrag | |
523bad63 TK |
1772 | I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate |
1773 | defragment activity in request to DDIR_WRITE event. | |
0d978694 | 1774 | .TP |
d5f9b0ea IF |
1775 | .B rados |
1776 | I/O engine supporting direct access to Ceph Reliable Autonomic Distributed | |
1777 | Object Store (RADOS) via librados. This ioengine defines engine specific | |
1778 | options. | |
1779 | .TP | |
0d978694 | 1780 | .B rbd |
523bad63 TK |
1781 | I/O engine supporting direct access to Ceph Rados Block Devices |
1782 | (RBD) via librbd without the need to use the kernel rbd driver. This | |
1783 | ioengine defines engine specific options. | |
a7c386f4 | 1784 | .TP |
c2f6a13d LMB |
1785 | .B http |
1786 | I/O engine supporting GET/PUT requests over HTTP(S) with libcurl to | |
1787 | a WebDAV or S3 endpoint. This ioengine defines engine specific options. | |
1788 | ||
1789 | This engine only supports direct IO of iodepth=1; you need to scale this | |
1790 | via numjobs. blocksize defines the size of the objects to be created. | |
1791 | ||
1792 | TRIM is translated to object deletion. | |
1793 | .TP | |
a7c386f4 | 1794 | .B gfapi |
523bad63 TK |
1795 | Using GlusterFS libgfapi sync interface to direct access to |
1796 | GlusterFS volumes without having to go through FUSE. This ioengine | |
1797 | defines engine specific options. | |
cc47f094 | 1798 | .TP |
1799 | .B gfapi_async | |
523bad63 TK |
1800 | Using GlusterFS libgfapi async interface to direct access to |
1801 | GlusterFS volumes without having to go through FUSE. This ioengine | |
1802 | defines engine specific options. | |
1b10477b | 1803 | .TP |
b74e419e | 1804 | .B libhdfs |
523bad63 TK |
1805 | Read and write through Hadoop (HDFS). The \fBfilename\fR option |
1806 | is used to specify host,port of the hdfs name\-node to connect. This | |
1807 | engine interprets offsets a little differently. In HDFS, files once | |
1808 | created cannot be modified so random writes are not possible. To | |
1809 | imitate this the libhdfs engine expects a bunch of small files to be | |
1810 | created over HDFS and will randomly pick a file from them | |
1811 | based on the offset generated by fio backend (see the example | |
1812 | job file to create such files, use `rw=write' option). Please | |
1813 | note, it may be necessary to set environment variables to work | |
1814 | with HDFS/libhdfs properly. Each job uses its own connection to | |
1815 | HDFS. | |
65fa28ca DE |
1816 | .TP |
1817 | .B mtd | |
523bad63 TK |
1818 | Read, write and erase an MTD character device (e.g., |
1819 | `/dev/mtd0'). Discards are treated as erases. Depending on the | |
1820 | underlying device type, the I/O may have to go in a certain pattern, | |
1821 | e.g., on NAND, writing sequentially to erase blocks and discarding | |
1822 | before overwriting. The \fBtrimwrite\fR mode works well for this | |
65fa28ca | 1823 | constraint. |
5c4ef02e JA |
1824 | .TP |
1825 | .B pmemblk | |
523bad63 | 1826 | Read and write using filesystem DAX to a file on a filesystem |
363a5f65 | 1827 | mounted with DAX on a persistent memory device through the PMDK |
523bad63 | 1828 | libpmemblk library. |
104ee4de | 1829 | .TP |
523bad63 TK |
1830 | .B dev\-dax |
1831 | Read and write using device DAX to a persistent memory device (e.g., | |
363a5f65 | 1832 | /dev/dax0.0) through the PMDK libpmem library. |
d60e92d1 | 1833 | .TP |
523bad63 TK |
1834 | .B external |
1835 | Prefix to specify loading an external I/O engine object file. Append | |
1836 | the engine filename, e.g. `ioengine=external:/tmp/foo.o' to load | |
d243fd6d TK |
1837 | ioengine `foo.o' in `/tmp'. The path can be either |
1838 | absolute or relative. See `engines/skeleton_external.c' in the fio source for | |
1839 | details of writing an external I/O engine. | |
1216cc5a JB |
1840 | .TP |
1841 | .B filecreate | |
b71968b1 SW |
1842 | Simply create the files and do no I/O to them. You still need to set |
1843 | \fBfilesize\fR so that all the accounting still occurs, but no actual I/O will be | |
1844 | done other than creating the file. | |
ae0db592 | 1845 | .TP |
73ccd14e SF |
1846 | .B filestat |
1847 | Simply do stat() and do no I/O to the file. You need to set 'filesize' | |
1848 | and 'nrfiles', so that files will be created. | |
1849 | This engine is to measure file lookup and meta data access. | |
1850 | .TP | |
5561e9dd FS |
1851 | .B filedelete |
1852 | Simply delete files by unlink() and do no I/O to the file. You need to set 'filesize' | |
1853 | and 'nrfiles', so that files will be created. | |
1854 | This engine is to measure file delete. | |
1855 | .TP | |
ae0db592 TI |
1856 | .B libpmem |
1857 | Read and write using mmap I/O to a file on a filesystem | |
363a5f65 | 1858 | mounted with DAX on a persistent memory device through the PMDK |
ae0db592 | 1859 | libpmem library. |
07751e10 JA |
1860 | .TP |
1861 | .B ime_psync | |
1862 | Synchronous read and write using DDN's Infinite Memory Engine (IME). This | |
1863 | engine is very basic and issues calls to IME whenever an IO is queued. | |
1864 | .TP | |
1865 | .B ime_psyncv | |
1866 | Synchronous read and write using DDN's Infinite Memory Engine (IME). This | |
1867 | engine uses iovecs and will try to stack as much IOs as possible (if the IOs | |
1868 | are "contiguous" and the IO depth is not exceeded) before issuing a call to IME. | |
1869 | .TP | |
1870 | .B ime_aio | |
1871 | Asynchronous read and write using DDN's Infinite Memory Engine (IME). This | |
1872 | engine will try to stack as much IOs as possible by creating requests for IME. | |
1873 | FIO will then decide when to commit these requests. | |
247ef2aa KZ |
1874 | .TP |
1875 | .B libiscsi | |
1876 | Read and write iscsi lun with libiscsi. | |
d643a1e2 RJ |
1877 | .TP |
1878 | .B nbd | |
1879 | Synchronous read and write a Network Block Device (NBD). | |
10756b2c BS |
1880 | .TP |
1881 | .B libcufile | |
1882 | I/O engine supporting libcufile synchronous access to nvidia-fs and a | |
1883 | GPUDirect Storage-supported filesystem. This engine performs | |
1884 | I/O without transferring buffers between user-space and the kernel, | |
1885 | unless \fBverify\fR is set or \fBcuda_io\fR is \fBposix\fR. \fBiomem\fR must | |
1886 | not be \fBcudamalloc\fR. This ioengine defines engine specific options. | |
c363fdd7 JL |
1887 | .TP |
1888 | .B dfs | |
1889 | I/O engine supporting asynchronous read and write operations to the DAOS File | |
1890 | System (DFS) via libdfs. | |
523bad63 TK |
1891 | .SS "I/O engine specific parameters" |
1892 | In addition, there are some parameters which are only valid when a specific | |
1893 | \fBioengine\fR is in use. These are used identically to normal parameters, | |
1894 | with the caveat that when used on the command line, they must come after the | |
1895 | \fBioengine\fR that defines them is selected. | |
d60e92d1 | 1896 | .TP |
b2a432bf PC |
1897 | .BI (io_uring, libaio)cmdprio_percentage \fR=\fPint |
1898 | Set the percentage of I/O that will be issued with higher priority by setting | |
1899 | the priority bit. Non-read I/O is likely unaffected by ``cmdprio_percentage``. | |
1900 | This option cannot be used with the `prio` or `prioclass` options. For this | |
1901 | option to set the priority bit properly, NCQ priority must be supported and | |
7896180a VF |
1902 | enabled and `direct=1' option must be used. fio must also be run as the root |
1903 | user. | |
029b42ac JA |
1904 | .TP |
1905 | .BI (io_uring)fixedbufs | |
1906 | If fio is asked to do direct IO, then Linux will map pages for each IO call, and | |
1907 | release them when IO is done. If this option is set, the pages are pre-mapped | |
1908 | before IO is started. This eliminates the need to map and release for each IO. | |
1909 | This is more efficient, and reduces the IO latency as well. | |
1910 | .TP | |
b2a432bf PC |
1911 | .BI (io_uring)hipri |
1912 | If this option is set, fio will attempt to use polled IO completions. Normal IO | |
1913 | completions generate interrupts to signal the completion of IO, polled | |
1914 | completions do not. Hence they are require active reaping by the application. | |
1915 | The benefits are more efficient IO for high IOPS scenarios, and lower latencies | |
1916 | for low queue depth IO. | |
1917 | .TP | |
5ffd5626 JA |
1918 | .BI (io_uring)registerfiles |
1919 | With this option, fio registers the set of files being used with the kernel. | |
1920 | This avoids the overhead of managing file counts in the kernel, making the | |
1921 | submission and completion part more lightweight. Required for the below | |
1922 | sqthread_poll option. | |
1923 | .TP | |
029b42ac JA |
1924 | .BI (io_uring)sqthread_poll |
1925 | Normally fio will submit IO by issuing a system call to notify the kernel of | |
1926 | available items in the SQ ring. If this option is set, the act of submitting IO | |
1927 | will be done by a polling thread in the kernel. This frees up cycles for fio, at | |
1928 | the cost of using more CPU in the system. | |
1929 | .TP | |
1930 | .BI (io_uring)sqthread_poll_cpu | |
1931 | When `sqthread_poll` is set, this option provides a way to define which CPU | |
1932 | should be used for the polling thread. | |
1933 | .TP | |
523bad63 TK |
1934 | .BI (libaio)userspace_reap |
1935 | Normally, with the libaio engine in use, fio will use the | |
1936 | \fBio_getevents\fR\|(3) system call to reap newly returned events. With | |
338f2db5 | 1937 | this flag turned on, the AIO ring will be read directly from user-space to |
523bad63 TK |
1938 | reap events. The reaping mode is only enabled when polling for a minimum of |
1939 | 0 events (e.g. when `iodepth_batch_complete=0'). | |
3ce9dcaf | 1940 | .TP |
523bad63 TK |
1941 | .BI (pvsync2)hipri |
1942 | Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority | |
1943 | than normal. | |
82407585 | 1944 | .TP |
523bad63 TK |
1945 | .BI (pvsync2)hipri_percentage |
1946 | When hipri is set this determines the probability of a pvsync2 I/O being high | |
1947 | priority. The default is 100%. | |
d60e92d1 | 1948 | .TP |
7d42e66e KK |
1949 | .BI (pvsync2,libaio,io_uring)nowait |
1950 | By default if a request cannot be executed immediately (e.g. resource starvation, | |
1951 | waiting on locks) it is queued and the initiating process will be blocked until | |
1952 | the required resource becomes free. | |
1953 | This option sets the RWF_NOWAIT flag (supported from the 4.14 Linux kernel) and | |
1954 | the call will return instantly with EAGAIN or a partial result rather than waiting. | |
1955 | ||
1956 | It is useful to also use \fBignore_error\fR=EAGAIN when using this option. | |
1957 | Note: glibc 2.27, 2.28 have a bug in syscall wrappers preadv2, pwritev2. | |
1958 | They return EOPNOTSUP instead of EAGAIN. | |
1959 | ||
1960 | For cached I/O, using this option usually means a request operates only with | |
1961 | cached data. Currently the RWF_NOWAIT flag does not supported for cached write. | |
1962 | For direct I/O, requests will only succeed if cache invalidation isn't required, | |
1963 | file blocks are fully allocated and the disk request could be issued immediately. | |
1964 | .TP | |
523bad63 TK |
1965 | .BI (cpuio)cpuload \fR=\fPint |
1966 | Attempt to use the specified percentage of CPU cycles. This is a mandatory | |
1967 | option when using cpuio I/O engine. | |
997b5680 | 1968 | .TP |
523bad63 TK |
1969 | .BI (cpuio)cpuchunks \fR=\fPint |
1970 | Split the load into cycles of the given time. In microseconds. | |
1ad01bd1 | 1971 | .TP |
523bad63 TK |
1972 | .BI (cpuio)exit_on_io_done \fR=\fPbool |
1973 | Detect when I/O threads are done, then exit. | |
d60e92d1 | 1974 | .TP |
523bad63 TK |
1975 | .BI (libhdfs)namenode \fR=\fPstr |
1976 | The hostname or IP address of a HDFS cluster namenode to contact. | |
d01612f3 | 1977 | .TP |
523bad63 TK |
1978 | .BI (libhdfs)port |
1979 | The listening port of the HFDS cluster namenode. | |
d60e92d1 | 1980 | .TP |
523bad63 TK |
1981 | .BI (netsplice,net)port |
1982 | The TCP or UDP port to bind to or connect to. If this is used with | |
1983 | \fBnumjobs\fR to spawn multiple instances of the same job type, then | |
1984 | this will be the starting port number since fio will use a range of | |
1985 | ports. | |
d60e92d1 | 1986 | .TP |
e4c4625f | 1987 | .BI (rdma, librpma_*)port |
609ac152 SB |
1988 | The port to use for RDMA-CM communication. This should be the same |
1989 | value on the client and the server side. | |
1990 | .TP | |
1991 | .BI (netsplice,net, rdma)hostname \fR=\fPstr | |
1992 | The hostname or IP address to use for TCP, UDP or RDMA-CM based I/O. | |
1993 | If the job is a TCP listener or UDP reader, the hostname is not used | |
1994 | and must be omitted unless it is a valid UDP multicast address. | |
591e9e06 | 1995 | .TP |
e4c4625f JM |
1996 | .BI (librpma_*)serverip \fR=\fPstr |
1997 | The IP address to be used for RDMA-CM based I/O. | |
1998 | .TP | |
1999 | .BI (librpma_*_server)direct_write_to_pmem \fR=\fPbool | |
2000 | Set to 1 only when Direct Write to PMem from the remote host is possible. Otherwise, set to 0. | |
2001 | .TP | |
523bad63 TK |
2002 | .BI (netsplice,net)interface \fR=\fPstr |
2003 | The IP address of the network interface used to send or receive UDP | |
2004 | multicast. | |
ddf24e42 | 2005 | .TP |
523bad63 TK |
2006 | .BI (netsplice,net)ttl \fR=\fPint |
2007 | Time\-to\-live value for outgoing UDP multicast packets. Default: 1. | |
d60e92d1 | 2008 | .TP |
523bad63 TK |
2009 | .BI (netsplice,net)nodelay \fR=\fPbool |
2010 | Set TCP_NODELAY on TCP connections. | |
fa769d44 | 2011 | .TP |
523bad63 TK |
2012 | .BI (netsplice,net)protocol \fR=\fPstr "\fR,\fP proto" \fR=\fPstr |
2013 | The network protocol to use. Accepted values are: | |
2014 | .RS | |
e76b1da4 JA |
2015 | .RS |
2016 | .TP | |
523bad63 TK |
2017 | .B tcp |
2018 | Transmission control protocol. | |
e76b1da4 | 2019 | .TP |
523bad63 TK |
2020 | .B tcpv6 |
2021 | Transmission control protocol V6. | |
e76b1da4 | 2022 | .TP |
523bad63 TK |
2023 | .B udp |
2024 | User datagram protocol. | |
2025 | .TP | |
2026 | .B udpv6 | |
2027 | User datagram protocol V6. | |
e76b1da4 | 2028 | .TP |
523bad63 TK |
2029 | .B unix |
2030 | UNIX domain socket. | |
e76b1da4 JA |
2031 | .RE |
2032 | .P | |
523bad63 TK |
2033 | When the protocol is TCP or UDP, the port must also be given, as well as the |
2034 | hostname if the job is a TCP listener or UDP reader. For unix sockets, the | |
2035 | normal \fBfilename\fR option should be used and the port is invalid. | |
2036 | .RE | |
2037 | .TP | |
2038 | .BI (netsplice,net)listen | |
2039 | For TCP network connections, tell fio to listen for incoming connections | |
2040 | rather than initiating an outgoing connection. The \fBhostname\fR must | |
2041 | be omitted if this option is used. | |
2042 | .TP | |
2043 | .BI (netsplice,net)pingpong | |
2044 | Normally a network writer will just continue writing data, and a network | |
2045 | reader will just consume packages. If `pingpong=1' is set, a writer will | |
2046 | send its normal payload to the reader, then wait for the reader to send the | |
2047 | same payload back. This allows fio to measure network latencies. The | |
2048 | submission and completion latencies then measure local time spent sending or | |
2049 | receiving, and the completion latency measures how long it took for the | |
2050 | other end to receive and send back. For UDP multicast traffic | |
2051 | `pingpong=1' should only be set for a single reader when multiple readers | |
2052 | are listening to the same address. | |
2053 | .TP | |
2054 | .BI (netsplice,net)window_size \fR=\fPint | |
2055 | Set the desired socket buffer size for the connection. | |
e76b1da4 | 2056 | .TP |
523bad63 TK |
2057 | .BI (netsplice,net)mss \fR=\fPint |
2058 | Set the TCP maximum segment size (TCP_MAXSEG). | |
d60e92d1 | 2059 | .TP |
523bad63 TK |
2060 | .BI (e4defrag)donorname \fR=\fPstr |
2061 | File will be used as a block donor (swap extents between files). | |
d60e92d1 | 2062 | .TP |
523bad63 TK |
2063 | .BI (e4defrag)inplace \fR=\fPint |
2064 | Configure donor file blocks allocation strategy: | |
2065 | .RS | |
2066 | .RS | |
d60e92d1 | 2067 | .TP |
523bad63 TK |
2068 | .B 0 |
2069 | Default. Preallocate donor's file on init. | |
d60e92d1 | 2070 | .TP |
523bad63 TK |
2071 | .B 1 |
2072 | Allocate space immediately inside defragment event, and free right | |
2073 | after event. | |
2074 | .RE | |
2075 | .RE | |
d60e92d1 | 2076 | .TP |
d5f9b0ea | 2077 | .BI (rbd,rados)clustername \fR=\fPstr |
523bad63 | 2078 | Specifies the name of the Ceph cluster. |
92d42d69 | 2079 | .TP |
523bad63 TK |
2080 | .BI (rbd)rbdname \fR=\fPstr |
2081 | Specifies the name of the RBD. | |
92d42d69 | 2082 | .TP |
d5f9b0ea IF |
2083 | .BI (rbd,rados)pool \fR=\fPstr |
2084 | Specifies the name of the Ceph pool containing RBD or RADOS data. | |
92d42d69 | 2085 | .TP |
d5f9b0ea | 2086 | .BI (rbd,rados)clientname \fR=\fPstr |
523bad63 TK |
2087 | Specifies the username (without the 'client.' prefix) used to access the |
2088 | Ceph cluster. If the \fBclustername\fR is specified, the \fBclientname\fR shall be | |
2089 | the full *type.id* string. If no type. prefix is given, fio will add 'client.' | |
2090 | by default. | |
92d42d69 | 2091 | .TP |
d5f9b0ea IF |
2092 | .BI (rbd,rados)busy_poll \fR=\fPbool |
2093 | Poll store instead of waiting for completion. Usually this provides better | |
2094 | throughput at cost of higher(up to 100%) CPU utilization. | |
2095 | .TP | |
2b728756 AK |
2096 | .BI (rados)touch_objects \fR=\fPbool |
2097 | During initialization, touch (create if do not exist) all objects (files). | |
2098 | Touching all objects affects ceph caches and likely impacts test results. | |
2099 | Enabled by default. | |
2100 | .TP | |
c2f6a13d LMB |
2101 | .BI (http)http_host \fR=\fPstr |
2102 | Hostname to connect to. For S3, this could be the bucket name. Default | |
2103 | is \fBlocalhost\fR | |
2104 | .TP | |
2105 | .BI (http)http_user \fR=\fPstr | |
2106 | Username for HTTP authentication. | |
2107 | .TP | |
2108 | .BI (http)http_pass \fR=\fPstr | |
2109 | Password for HTTP authentication. | |
2110 | .TP | |
09fd2966 LMB |
2111 | .BI (http)https \fR=\fPstr |
2112 | Whether to use HTTPS instead of plain HTTP. \fRon\fP enables HTTPS; | |
2113 | \fRinsecure\fP will enable HTTPS, but disable SSL peer verification (use | |
2114 | with caution!). Default is \fBoff\fR. | |
c2f6a13d | 2115 | .TP |
09fd2966 LMB |
2116 | .BI (http)http_mode \fR=\fPstr |
2117 | Which HTTP access mode to use: webdav, swift, or s3. Default is | |
2118 | \fBwebdav\fR. | |
c2f6a13d LMB |
2119 | .TP |
2120 | .BI (http)http_s3_region \fR=\fPstr | |
2121 | The S3 region/zone to include in the request. Default is \fBus-east-1\fR. | |
2122 | .TP | |
2123 | .BI (http)http_s3_key \fR=\fPstr | |
2124 | The S3 secret key. | |
2125 | .TP | |
2126 | .BI (http)http_s3_keyid \fR=\fPstr | |
2127 | The S3 key/access id. | |
2128 | .TP | |
09fd2966 LMB |
2129 | .BI (http)http_swift_auth_token \fR=\fPstr |
2130 | The Swift auth token. See the example configuration file on how to | |
2131 | retrieve this. | |
2132 | .TP | |
c2f6a13d LMB |
2133 | .BI (http)http_verbose \fR=\fPint |
2134 | Enable verbose requests from libcurl. Useful for debugging. 1 turns on | |
2135 | verbose logging from libcurl, 2 additionally enables HTTP IO tracing. | |
2136 | Default is \fB0\fR | |
2137 | .TP | |
523bad63 TK |
2138 | .BI (mtd)skip_bad \fR=\fPbool |
2139 | Skip operations against known bad blocks. | |
8116fd24 | 2140 | .TP |
523bad63 TK |
2141 | .BI (libhdfs)hdfsdirectory |
2142 | libhdfs will create chunk in this HDFS directory. | |
e0a04ac1 | 2143 | .TP |
523bad63 TK |
2144 | .BI (libhdfs)chunk_size |
2145 | The size of the chunk to use for each file. | |
609ac152 SB |
2146 | .TP |
2147 | .BI (rdma)verb \fR=\fPstr | |
2148 | The RDMA verb to use on this side of the RDMA ioengine | |
2149 | connection. Valid values are write, read, send and recv. These | |
2150 | correspond to the equivalent RDMA verbs (e.g. write = rdma_write | |
2151 | etc.). Note that this only needs to be specified on the client side of | |
2152 | the connection. See the examples folder. | |
2153 | .TP | |
2154 | .BI (rdma)bindname \fR=\fPstr | |
2155 | The name to use to bind the local RDMA-CM connection to a local RDMA | |
2156 | device. This could be a hostname or an IPv4 or IPv6 address. On the | |
2157 | server side this will be passed into the rdma_bind_addr() function and | |
2158 | on the client site it will be used in the rdma_resolve_add() | |
2159 | function. This can be useful when multiple paths exist between the | |
2160 | client and the server or in certain loopback configurations. | |
52b81b7c | 2161 | .TP |
93a13ba5 TK |
2162 | .BI (filestat)stat_type \fR=\fPstr |
2163 | Specify stat system call type to measure lookup/getattr performance. | |
2164 | Default is \fBstat\fR for \fBstat\fR\|(2). | |
c446eff0 | 2165 | .TP |
b0dc148e DG |
2166 | .BI (sg)hipri |
2167 | If this option is set, fio will attempt to use polled IO completions. This | |
2168 | will have a similar effect as (io_uring)hipri. Only SCSI READ and WRITE | |
2169 | commands will have the SGV4_FLAG_HIPRI set (not UNMAP (trim) nor VERIFY). | |
2170 | Older versions of the Linux sg driver that do not support hipri will simply | |
2171 | ignore this flag and do normal IO. The Linux SCSI Low Level Driver (LLD) | |
2172 | that "owns" the device also needs to support hipri (also known as iopoll | |
2173 | and mq_poll). The MegaRAID driver is an example of a SCSI LLD. | |
2174 | Default: clear (0) which does normal (interrupted based) IO. | |
2175 | .TP | |
52b81b7c KD |
2176 | .BI (sg)readfua \fR=\fPbool |
2177 | With readfua option set to 1, read operations include the force | |
2178 | unit access (fua) flag. Default: 0. | |
2179 | .TP | |
2180 | .BI (sg)writefua \fR=\fPbool | |
2181 | With writefua option set to 1, write operations include the force | |
2182 | unit access (fua) flag. Default: 0. | |
2c3a9150 VF |
2183 | .TP |
2184 | .BI (sg)sg_write_mode \fR=\fPstr | |
2185 | Specify the type of write commands to issue. This option can take three | |
2186 | values: | |
2187 | .RS | |
2188 | .RS | |
2189 | .TP | |
2190 | .B write (default) | |
2191 | Write opcodes are issued as usual | |
2192 | .TP | |
2193 | .B verify | |
2194 | Issue WRITE AND VERIFY commands. The BYTCHK bit is set to 0. This | |
2195 | directs the device to carry out a medium verification with no data | |
2196 | comparison. The writefua option is ignored with this selection. | |
2197 | .TP | |
2198 | .B same | |
2199 | Issue WRITE SAME commands. This transfers a single block to the device | |
2200 | and writes this same block of data to a contiguous sequence of LBAs | |
2201 | beginning at the specified offset. fio's block size parameter | |
2202 | specifies the amount of data written with each command. However, the | |
2203 | amount of data actually transferred to the device is equal to the | |
2204 | device's block (sector) size. For a device with 512 byte sectors, | |
2205 | blocksize=8k will write 16 sectors with each command. fio will still | |
2206 | generate 8k of data for each command butonly the first 512 bytes will | |
2207 | be used and transferred to the device. The writefua option is ignored | |
2208 | with this selection. | |
f2d6de5d RJ |
2209 | .RE |
2210 | .RE | |
2211 | .TP | |
2212 | .BI (nbd)uri \fR=\fPstr | |
2213 | Specify the NBD URI of the server to test. | |
2214 | The string is a standard NBD URI (see | |
2215 | \fIhttps://github.com/NetworkBlockDevice/nbd/tree/master/doc\fR). | |
2216 | Example URIs: | |
2217 | .RS | |
2218 | .RS | |
2219 | .TP | |
2220 | \fInbd://localhost:10809\fR | |
2221 | .TP | |
2222 | \fInbd+unix:///?socket=/tmp/socket\fR | |
2223 | .TP | |
2224 | \fInbds://tlshost/exportname\fR | |
10756b2c BS |
2225 | .RE |
2226 | .RE | |
2227 | .TP | |
2228 | .BI (libcufile)gpu_dev_ids\fR=\fPstr | |
2229 | Specify the GPU IDs to use with CUDA. This is a colon-separated list of int. | |
2230 | GPUs are assigned to workers roundrobin. Default is 0. | |
2231 | .TP | |
2232 | .BI (libcufile)cuda_io\fR=\fPstr | |
2233 | Specify the type of I/O to use with CUDA. This option | |
2234 | takes the following values: | |
2235 | .RS | |
2236 | .RS | |
2237 | .TP | |
2238 | .B cufile (default) | |
2239 | Use libcufile and nvidia-fs. This option performs I/O directly | |
2240 | between a GPUDirect Storage filesystem and GPU buffers, | |
2241 | avoiding use of a bounce buffer. If \fBverify\fR is set, | |
2242 | cudaMemcpy is used to copy verification data between RAM and GPU(s). | |
2243 | Verification data is copied from RAM to GPU before a write | |
2244 | and from GPU to RAM after a read. | |
2245 | \fBdirect\fR must be 1. | |
2246 | .TP | |
2247 | .BI posix | |
2248 | Use POSIX to perform I/O with a RAM buffer, and use | |
2249 | cudaMemcpy to transfer data between RAM and the GPU(s). | |
2250 | Data is copied from GPU to RAM before a write and copied | |
2251 | from RAM to GPU after a read. \fBverify\fR does not affect | |
2252 | the use of cudaMemcpy. | |
2253 | .RE | |
2254 | .RE | |
c363fdd7 JL |
2255 | .TP |
2256 | .BI (dfs)pool | |
2257 | Specify the UUID of the DAOS pool to connect to. | |
2258 | .TP | |
2259 | .BI (dfs)cont | |
2260 | Specify the UUID of the DAOS DAOS container to open. | |
2261 | .TP | |
2262 | .BI (dfs)chunk_size | |
2263 | Specificy a different chunk size (in bytes) for the dfs file. | |
2264 | Use DAOS container's chunk size by default. | |
2265 | .TP | |
2266 | .BI (dfs)object_class | |
2267 | Specificy a different object class for the dfs file. | |
2268 | Use DAOS container's object class by default. | |
523bad63 TK |
2269 | .SS "I/O depth" |
2270 | .TP | |
2271 | .BI iodepth \fR=\fPint | |
2272 | Number of I/O units to keep in flight against the file. Note that | |
2273 | increasing \fBiodepth\fR beyond 1 will not affect synchronous ioengines (except | |
2274 | for small degrees when \fBverify_async\fR is in use). Even async | |
2275 | engines may impose OS restrictions causing the desired depth not to be | |
2276 | achieved. This may happen on Linux when using libaio and not setting | |
2277 | `direct=1', since buffered I/O is not async on that OS. Keep an | |
2278 | eye on the I/O depth distribution in the fio output to verify that the | |
2279 | achieved depth is as expected. Default: 1. | |
2280 | .TP | |
2281 | .BI iodepth_batch_submit \fR=\fPint "\fR,\fP iodepth_batch" \fR=\fPint | |
2282 | This defines how many pieces of I/O to submit at once. It defaults to 1 | |
2283 | which means that we submit each I/O as soon as it is available, but can be | |
2284 | raised to submit bigger batches of I/O at the time. If it is set to 0 the | |
2285 | \fBiodepth\fR value will be used. | |
2286 | .TP | |
2287 | .BI iodepth_batch_complete_min \fR=\fPint "\fR,\fP iodepth_batch_complete" \fR=\fPint | |
2288 | This defines how many pieces of I/O to retrieve at once. It defaults to 1 | |
2289 | which means that we'll ask for a minimum of 1 I/O in the retrieval process | |
2290 | from the kernel. The I/O retrieval will go on until we hit the limit set by | |
2291 | \fBiodepth_low\fR. If this variable is set to 0, then fio will always | |
2292 | check for completed events before queuing more I/O. This helps reduce I/O | |
2293 | latency, at the cost of more retrieval system calls. | |
2294 | .TP | |
2295 | .BI iodepth_batch_complete_max \fR=\fPint | |
2296 | This defines maximum pieces of I/O to retrieve at once. This variable should | |
2297 | be used along with \fBiodepth_batch_complete_min\fR=\fIint\fR variable, | |
2298 | specifying the range of min and max amount of I/O which should be | |
2299 | retrieved. By default it is equal to \fBiodepth_batch_complete_min\fR | |
2300 | value. Example #1: | |
e0a04ac1 | 2301 | .RS |
e0a04ac1 | 2302 | .RS |
e0a04ac1 | 2303 | .P |
523bad63 TK |
2304 | .PD 0 |
2305 | iodepth_batch_complete_min=1 | |
e0a04ac1 | 2306 | .P |
523bad63 TK |
2307 | iodepth_batch_complete_max=<iodepth> |
2308 | .PD | |
e0a04ac1 JA |
2309 | .RE |
2310 | .P | |
523bad63 TK |
2311 | which means that we will retrieve at least 1 I/O and up to the whole |
2312 | submitted queue depth. If none of I/O has been completed yet, we will wait. | |
2313 | Example #2: | |
e8b1961d | 2314 | .RS |
523bad63 TK |
2315 | .P |
2316 | .PD 0 | |
2317 | iodepth_batch_complete_min=0 | |
2318 | .P | |
2319 | iodepth_batch_complete_max=<iodepth> | |
2320 | .PD | |
e8b1961d JA |
2321 | .RE |
2322 | .P | |
523bad63 TK |
2323 | which means that we can retrieve up to the whole submitted queue depth, but |
2324 | if none of I/O has been completed yet, we will NOT wait and immediately exit | |
2325 | the system call. In this example we simply do polling. | |
2326 | .RE | |
e8b1961d | 2327 | .TP |
523bad63 TK |
2328 | .BI iodepth_low \fR=\fPint |
2329 | The low water mark indicating when to start filling the queue | |
2330 | again. Defaults to the same as \fBiodepth\fR, meaning that fio will | |
2331 | attempt to keep the queue full at all times. If \fBiodepth\fR is set to | |
2332 | e.g. 16 and \fBiodepth_low\fR is set to 4, then after fio has filled the queue of | |
2333 | 16 requests, it will let the depth drain down to 4 before starting to fill | |
2334 | it again. | |
d60e92d1 | 2335 | .TP |
523bad63 TK |
2336 | .BI serialize_overlap \fR=\fPbool |
2337 | Serialize in-flight I/Os that might otherwise cause or suffer from data races. | |
2338 | When two or more I/Os are submitted simultaneously, there is no guarantee that | |
2339 | the I/Os will be processed or completed in the submitted order. Further, if | |
2340 | two or more of those I/Os are writes, any overlapping region between them can | |
2341 | become indeterminate/undefined on certain storage. These issues can cause | |
2342 | verification to fail erratically when at least one of the racing I/Os is | |
2343 | changing data and the overlapping region has a non-zero size. Setting | |
2344 | \fBserialize_overlap\fR tells fio to avoid provoking this behavior by explicitly | |
2345 | serializing in-flight I/Os that have a non-zero overlap. Note that setting | |
2346 | this option can reduce both performance and the \fBiodepth\fR achieved. | |
3d6a6f04 VF |
2347 | .RS |
2348 | .P | |
2349 | This option only applies to I/Os issued for a single job except when it is | |
2350 | enabled along with \fBio_submit_mode\fR=offload. In offload mode, fio | |
2351 | will check for overlap among all I/Os submitted by offload jobs with \fBserialize_overlap\fR | |
307f2246 | 2352 | enabled. |
3d6a6f04 VF |
2353 | .P |
2354 | Default: false. | |
2355 | .RE | |
d60e92d1 | 2356 | .TP |
523bad63 TK |
2357 | .BI io_submit_mode \fR=\fPstr |
2358 | This option controls how fio submits the I/O to the I/O engine. The default | |
2359 | is `inline', which means that the fio job threads submit and reap I/O | |
2360 | directly. If set to `offload', the job threads will offload I/O submission | |
2361 | to a dedicated pool of I/O threads. This requires some coordination and thus | |
2362 | has a bit of extra overhead, especially for lower queue depth I/O where it | |
2363 | can increase latencies. The benefit is that fio can manage submission rates | |
2364 | independently of the device completion rates. This avoids skewed latency | |
2365 | reporting if I/O gets backed up on the device side (the coordinated omission | |
abfd235a | 2366 | problem). Note that this option cannot reliably be used with async IO engines. |
523bad63 | 2367 | .SS "I/O rate" |
d60e92d1 | 2368 | .TP |
523bad63 TK |
2369 | .BI thinktime \fR=\fPtime |
2370 | Stall the job for the specified period of time after an I/O has completed before issuing the | |
2371 | next. May be used to simulate processing being done by an application. | |
2372 | When the unit is omitted, the value is interpreted in microseconds. See | |
2373 | \fBthinktime_blocks\fR and \fBthinktime_spin\fR. | |
d60e92d1 | 2374 | .TP |
523bad63 | 2375 | .BI thinktime_spin \fR=\fPtime |
338f2db5 | 2376 | Only valid if \fBthinktime\fR is set - pretend to spend CPU time doing |
523bad63 TK |
2377 | something with the data received, before falling back to sleeping for the |
2378 | rest of the period specified by \fBthinktime\fR. When the unit is | |
2379 | omitted, the value is interpreted in microseconds. | |
d60e92d1 AC |
2380 | .TP |
2381 | .BI thinktime_blocks \fR=\fPint | |
338f2db5 | 2382 | Only valid if \fBthinktime\fR is set - control how many blocks to issue, |
523bad63 TK |
2383 | before waiting \fBthinktime\fR usecs. If not set, defaults to 1 which will make |
2384 | fio wait \fBthinktime\fR usecs after every block. This effectively makes any | |
2385 | queue depth setting redundant, since no more than 1 I/O will be queued | |
2386 | before we have to complete it and do our \fBthinktime\fR. In other words, this | |
2387 | setting effectively caps the queue depth if the latter is larger. | |
d60e92d1 | 2388 | .TP |
33f42c20 HQ |
2389 | .BI thinktime_blocks_type \fR=\fPstr |
2390 | Only valid if \fBthinktime\fR is set - control how \fBthinktime_blocks\fR triggers. | |
2391 | The default is `complete', which triggers \fBthinktime\fR when fio completes | |
2392 | \fBthinktime_blocks\fR blocks. If this is set to `issue', then the trigger happens | |
2393 | at the issue side. | |
2394 | .TP | |
6d500c2e | 2395 | .BI rate \fR=\fPint[,int][,int] |
523bad63 | 2396 | Cap the bandwidth used by this job. The number is in bytes/sec, the normal |
338f2db5 | 2397 | suffix rules apply. Comma-separated values may be specified for reads, |
523bad63 TK |
2398 | writes, and trims as described in \fBblocksize\fR. |
2399 | .RS | |
2400 | .P | |
2401 | For example, using `rate=1m,500k' would limit reads to 1MiB/sec and writes to | |
2402 | 500KiB/sec. Capping only reads or writes can be done with `rate=,500k' or | |
2403 | `rate=500k,' where the former will only limit writes (to 500KiB/sec) and the | |
2404 | latter will only limit reads. | |
2405 | .RE | |
d60e92d1 | 2406 | .TP |
6d500c2e | 2407 | .BI rate_min \fR=\fPint[,int][,int] |
523bad63 | 2408 | Tell fio to do whatever it can to maintain at least this bandwidth. Failing |
338f2db5 | 2409 | to meet this requirement will cause the job to exit. Comma-separated values |
523bad63 TK |
2410 | may be specified for reads, writes, and trims as described in |
2411 | \fBblocksize\fR. | |
d60e92d1 | 2412 | .TP |
6d500c2e | 2413 | .BI rate_iops \fR=\fPint[,int][,int] |
523bad63 TK |
2414 | Cap the bandwidth to this number of IOPS. Basically the same as |
2415 | \fBrate\fR, just specified independently of bandwidth. If the job is | |
2416 | given a block size range instead of a fixed value, the smallest block size | |
338f2db5 | 2417 | is used as the metric. Comma-separated values may be specified for reads, |
523bad63 | 2418 | writes, and trims as described in \fBblocksize\fR. |
d60e92d1 | 2419 | .TP |
6d500c2e | 2420 | .BI rate_iops_min \fR=\fPint[,int][,int] |
523bad63 | 2421 | If fio doesn't meet this rate of I/O, it will cause the job to exit. |
338f2db5 | 2422 | Comma-separated values may be specified for reads, writes, and trims as |
523bad63 | 2423 | described in \fBblocksize\fR. |
d60e92d1 | 2424 | .TP |
6de65959 | 2425 | .BI rate_process \fR=\fPstr |
523bad63 TK |
2426 | This option controls how fio manages rated I/O submissions. The default is |
2427 | `linear', which submits I/O in a linear fashion with fixed delays between | |
2428 | I/Os that gets adjusted based on I/O completion rates. If this is set to | |
2429 | `poisson', fio will submit I/O based on a more real world random request | |
6de65959 | 2430 | flow, known as the Poisson process |
523bad63 | 2431 | (\fIhttps://en.wikipedia.org/wiki/Poisson_point_process\fR). The lambda will be |
5d02b083 | 2432 | 10^6 / IOPS for the given workload. |
1a9bf814 JA |
2433 | .TP |
2434 | .BI rate_ignore_thinktime \fR=\fPbool | |
2435 | By default, fio will attempt to catch up to the specified rate setting, if any | |
2436 | kind of thinktime setting was used. If this option is set, then fio will | |
2437 | ignore the thinktime and continue doing IO at the specified rate, instead of | |
2438 | entering a catch-up mode after thinktime is done. | |
523bad63 | 2439 | .SS "I/O latency" |
ff6bb260 | 2440 | .TP |
523bad63 | 2441 | .BI latency_target \fR=\fPtime |
3e260a46 | 2442 | If set, fio will attempt to find the max performance point that the given |
523bad63 TK |
2443 | workload will run at while maintaining a latency below this target. When |
2444 | the unit is omitted, the value is interpreted in microseconds. See | |
2445 | \fBlatency_window\fR and \fBlatency_percentile\fR. | |
3e260a46 | 2446 | .TP |
523bad63 | 2447 | .BI latency_window \fR=\fPtime |
3e260a46 | 2448 | Used with \fBlatency_target\fR to specify the sample window that the job |
523bad63 TK |
2449 | is run at varying queue depths to test the performance. When the unit is |
2450 | omitted, the value is interpreted in microseconds. | |
3e260a46 JA |
2451 | .TP |
2452 | .BI latency_percentile \fR=\fPfloat | |
523bad63 TK |
2453 | The percentage of I/Os that must fall within the criteria specified by |
2454 | \fBlatency_target\fR and \fBlatency_window\fR. If not set, this | |
2455 | defaults to 100.0, meaning that all I/Os must be equal or below to the value | |
2456 | set by \fBlatency_target\fR. | |
2457 | .TP | |
e1bcd541 SL |
2458 | .BI latency_run \fR=\fPbool |
2459 | Used with \fBlatency_target\fR. If false (default), fio will find the highest | |
2460 | queue depth that meets \fBlatency_target\fR and exit. If true, fio will continue | |
2461 | running and try to meet \fBlatency_target\fR by adjusting queue depth. | |
2462 | .TP | |
f7cf63bf | 2463 | .BI max_latency \fR=\fPtime[,time][,time] |
523bad63 TK |
2464 | If set, fio will exit the job with an ETIMEDOUT error if it exceeds this |
2465 | maximum latency. When the unit is omitted, the value is interpreted in | |
f7cf63bf VR |
2466 | microseconds. Comma-separated values may be specified for reads, writes, |
2467 | and trims as described in \fBblocksize\fR. | |
523bad63 TK |
2468 | .TP |
2469 | .BI rate_cycle \fR=\fPint | |
2470 | Average bandwidth for \fBrate\fR and \fBrate_min\fR over this number | |
2471 | of milliseconds. Defaults to 1000. | |
2472 | .SS "I/O replay" | |
2473 | .TP | |
2474 | .BI write_iolog \fR=\fPstr | |
2475 | Write the issued I/O patterns to the specified file. See | |
2476 | \fBread_iolog\fR. Specify a separate file for each job, otherwise the | |
2477 | iologs will be interspersed and the file may be corrupt. | |
2478 | .TP | |
2479 | .BI read_iolog \fR=\fPstr | |
2480 | Open an iolog with the specified filename and replay the I/O patterns it | |
2481 | contains. This can be used to store a workload and replay it sometime | |
2482 | later. The iolog given may also be a blktrace binary file, which allows fio | |
2483 | to replay a workload captured by blktrace. See | |
2484 | \fBblktrace\fR\|(8) for how to capture such logging data. For blktrace | |
2485 | replay, the file needs to be turned into a blkparse binary data file first | |
2486 | (`blkparse <device> \-o /dev/null \-d file_for_fio.bin'). | |
c70c7f58 | 2487 | You can specify a number of files by separating the names with a ':' character. |
3b803fe1 | 2488 | See the \fBfilename\fR option for information on how to escape ':' |
c70c7f58 | 2489 | characters within the file names. These files will be sequentially assigned to |
d19c04d1 | 2490 | job clones created by \fBnumjobs\fR. '-' is a reserved name, meaning read from |
2491 | stdin, notably if \fBfilename\fR is set to '-' which means stdin as well, | |
2492 | then this flag can't be set to '-'. | |
3e260a46 | 2493 | .TP |
98e7161c AK |
2494 | .BI read_iolog_chunked \fR=\fPbool |
2495 | Determines how iolog is read. If false (default) entire \fBread_iolog\fR will | |
2496 | be read at once. If selected true, input from iolog will be read gradually. | |
2497 | Useful when iolog is very large, or it is generated. | |
2498 | .TP | |
b9921d1a DZ |
2499 | .BI merge_blktrace_file \fR=\fPstr |
2500 | When specified, rather than replaying the logs passed to \fBread_iolog\fR, | |
2501 | the logs go through a merge phase which aggregates them into a single blktrace. | |
2502 | The resulting file is then passed on as the \fBread_iolog\fR parameter. The | |
2503 | intention here is to make the order of events consistent. This limits the | |
2504 | influence of the scheduler compared to replaying multiple blktraces via | |
2505 | concurrent jobs. | |
2506 | .TP | |
87a48ada DZ |
2507 | .BI merge_blktrace_scalars \fR=\fPfloat_list |
2508 | This is a percentage based option that is index paired with the list of files | |
2509 | passed to \fBread_iolog\fR. When merging is performed, scale the time of each | |
2510 | event by the corresponding amount. For example, | |
2511 | `\-\-merge_blktrace_scalars="50:100"' runs the first trace in halftime and the | |
2512 | second trace in realtime. This knob is separately tunable from | |
2513 | \fBreplay_time_scale\fR which scales the trace during runtime and will not | |
2514 | change the output of the merge unlike this option. | |
2515 | .TP | |
55bfd8c8 DZ |
2516 | .BI merge_blktrace_iters \fR=\fPfloat_list |
2517 | This is a whole number option that is index paired with the list of files | |
2518 | passed to \fBread_iolog\fR. When merging is performed, run each trace for | |
2519 | the specified number of iterations. For example, | |
2520 | `\-\-merge_blktrace_iters="2:1"' runs the first trace for two iterations | |
2521 | and the second trace for one iteration. | |
2522 | .TP | |
523bad63 TK |
2523 | .BI replay_no_stall \fR=\fPbool |
2524 | When replaying I/O with \fBread_iolog\fR the default behavior is to | |
2525 | attempt to respect the timestamps within the log and replay them with the | |
2526 | appropriate delay between IOPS. By setting this variable fio will not | |
2527 | respect the timestamps and attempt to replay them as fast as possible while | |
2528 | still respecting ordering. The result is the same I/O pattern to a given | |
2529 | device, but different timings. | |
2530 | .TP | |
6dd7fa77 JA |
2531 | .BI replay_time_scale \fR=\fPint |
2532 | When replaying I/O with \fBread_iolog\fR, fio will honor the original timing | |
2533 | in the trace. With this option, it's possible to scale the time. It's a | |
2534 | percentage option, if set to 50 it means run at 50% the original IO rate in | |
2535 | the trace. If set to 200, run at twice the original IO rate. Defaults to 100. | |
2536 | .TP | |
523bad63 TK |
2537 | .BI replay_redirect \fR=\fPstr |
2538 | While replaying I/O patterns using \fBread_iolog\fR the default behavior | |
2539 | is to replay the IOPS onto the major/minor device that each IOP was recorded | |
2540 | from. This is sometimes undesirable because on a different machine those | |
2541 | major/minor numbers can map to a different device. Changing hardware on the | |
2542 | same system can also result in a different major/minor mapping. | |
2543 | \fBreplay_redirect\fR causes all I/Os to be replayed onto the single specified | |
2544 | device regardless of the device it was recorded | |
2545 | from. i.e. `replay_redirect=/dev/sdc' would cause all I/O | |
2546 | in the blktrace or iolog to be replayed onto `/dev/sdc'. This means | |
2547 | multiple devices will be replayed onto a single device, if the trace | |
2548 | contains multiple devices. If you want multiple devices to be replayed | |
2549 | concurrently to multiple redirected devices you must blkparse your trace | |
2550 | into separate traces and replay them with independent fio invocations. | |
2551 | Unfortunately this also breaks the strict time ordering between multiple | |
2552 | device accesses. | |
2553 | .TP | |
2554 | .BI replay_align \fR=\fPint | |
350a535d DZ |
2555 | Force alignment of the byte offsets in a trace to this value. The value |
2556 | must be a power of 2. | |
523bad63 TK |
2557 | .TP |
2558 | .BI replay_scale \fR=\fPint | |
350a535d DZ |
2559 | Scale bye offsets down by this factor when replaying traces. Should most |
2560 | likely use \fBreplay_align\fR as well. | |
523bad63 TK |
2561 | .SS "Threads, processes and job synchronization" |
2562 | .TP | |
38f68906 JA |
2563 | .BI replay_skip \fR=\fPstr |
2564 | Sometimes it's useful to skip certain IO types in a replay trace. This could | |
2565 | be, for instance, eliminating the writes in the trace. Or not replaying the | |
2566 | trims/discards, if you are redirecting to a device that doesn't support them. | |
2567 | This option takes a comma separated list of read, write, trim, sync. | |
2568 | .TP | |
523bad63 TK |
2569 | .BI thread |
2570 | Fio defaults to creating jobs by using fork, however if this option is | |
2571 | given, fio will create jobs by using POSIX Threads' function | |
2572 | \fBpthread_create\fR\|(3) to create threads instead. | |
2573 | .TP | |
2574 | .BI wait_for \fR=\fPstr | |
2575 | If set, the current job won't be started until all workers of the specified | |
2576 | waitee job are done. | |
2577 | .\" ignore blank line here from HOWTO as it looks normal without it | |
2578 | \fBwait_for\fR operates on the job name basis, so there are a few | |
2579 | limitations. First, the waitee must be defined prior to the waiter job | |
2580 | (meaning no forward references). Second, if a job is being referenced as a | |
2581 | waitee, it must have a unique name (no duplicate waitees). | |
2582 | .TP | |
2583 | .BI nice \fR=\fPint | |
2584 | Run the job with the given nice value. See man \fBnice\fR\|(2). | |
2585 | .\" ignore blank line here from HOWTO as it looks normal without it | |
2586 | On Windows, values less than \-15 set the process class to "High"; \-1 through | |
2587 | \-15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle" | |
2588 | priority class. | |
2589 | .TP | |
2590 | .BI prio \fR=\fPint | |
2591 | Set the I/O priority value of this job. Linux limits us to a positive value | |
2592 | between 0 and 7, with 0 being the highest. See man | |
2593 | \fBionice\fR\|(1). Refer to an appropriate manpage for other operating | |
b2a432bf PC |
2594 | systems since meaning of priority may differ. For per-command priority |
2595 | setting, see I/O engine specific `cmdprio_percentage` and `hipri_percentage` | |
2596 | options. | |
523bad63 TK |
2597 | .TP |
2598 | .BI prioclass \fR=\fPint | |
b2a432bf PC |
2599 | Set the I/O priority class. See man \fBionice\fR\|(1). For per-command |
2600 | priority setting, see I/O engine specific `cmdprio_percentage` and `hipri_percent` | |
2601 | options. | |
15501535 | 2602 | .TP |
d60e92d1 | 2603 | .BI cpus_allowed \fR=\fPstr |
523bad63 | 2604 | Controls the same options as \fBcpumask\fR, but accepts a textual |
b570e037 SW |
2605 | specification of the permitted CPUs instead and CPUs are indexed from 0. So |
2606 | to use CPUs 0 and 5 you would specify `cpus_allowed=0,5'. This option also | |
2607 | allows a range of CPUs to be specified \-\- say you wanted a binding to CPUs | |
2608 | 0, 5, and 8 to 15, you would set `cpus_allowed=0,5,8\-15'. | |
2609 | .RS | |
2610 | .P | |
2611 | On Windows, when `cpus_allowed' is unset only CPUs from fio's current | |
2612 | processor group will be used and affinity settings are inherited from the | |
2613 | system. An fio build configured to target Windows 7 makes options that set | |
2614 | CPUs processor group aware and values will set both the processor group | |
2615 | and a CPU from within that group. For example, on a system where processor | |
2616 | group 0 has 40 CPUs and processor group 1 has 32 CPUs, `cpus_allowed' | |
2617 | values between 0 and 39 will bind CPUs from processor group 0 and | |
2618 | `cpus_allowed' values between 40 and 71 will bind CPUs from processor | |
2619 | group 1. When using `cpus_allowed_policy=shared' all CPUs specified by a | |
2620 | single `cpus_allowed' option must be from the same processor group. For | |
2621 | Windows fio builds not built for Windows 7, CPUs will only be selected from | |
2622 | (and be relative to) whatever processor group fio happens to be running in | |
2623 | and CPUs from other processor groups cannot be used. | |
2624 | .RE | |
d60e92d1 | 2625 | .TP |
c2acfbac | 2626 | .BI cpus_allowed_policy \fR=\fPstr |
523bad63 TK |
2627 | Set the policy of how fio distributes the CPUs specified by |
2628 | \fBcpus_allowed\fR or \fBcpumask\fR. Two policies are supported: | |
c2acfbac JA |
2629 | .RS |
2630 | .RS | |
2631 | .TP | |
2632 | .B shared | |
2633 | All jobs will share the CPU set specified. | |
2634 | .TP | |
2635 | .B split | |
2636 | Each job will get a unique CPU from the CPU set. | |
2637 | .RE | |
2638 | .P | |
523bad63 | 2639 | \fBshared\fR is the default behavior, if the option isn't specified. If |
b21fc93f | 2640 | \fBsplit\fR is specified, then fio will assign one cpu per job. If not |
523bad63 TK |
2641 | enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs |
2642 | in the set. | |
c2acfbac | 2643 | .RE |
c2acfbac | 2644 | .TP |
b570e037 SW |
2645 | .BI cpumask \fR=\fPint |
2646 | Set the CPU affinity of this job. The parameter given is a bit mask of | |
2647 | allowed CPUs the job may run on. So if you want the allowed CPUs to be 1 | |
2648 | and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man | |
2649 | \fBsched_setaffinity\fR\|(2). This may not work on all supported | |
2650 | operating systems or kernel versions. This option doesn't work well for a | |
2651 | higher CPU count than what you can store in an integer mask, so it can only | |
2652 | control cpus 1\-32. For boxes with larger CPU counts, use | |
2653 | \fBcpus_allowed\fR. | |
2654 | .TP | |
d0b937ed | 2655 | .BI numa_cpu_nodes \fR=\fPstr |
cecbfd47 | 2656 | Set this job running on specified NUMA nodes' CPUs. The arguments allow |
523bad63 TK |
2657 | comma delimited list of cpu numbers, A\-B ranges, or `all'. Note, to enable |
2658 | NUMA options support, fio must be built on a system with libnuma\-dev(el) | |
2659 | installed. | |
d0b937ed YR |
2660 | .TP |
2661 | .BI numa_mem_policy \fR=\fPstr | |
523bad63 TK |
2662 | Set this job's memory policy and corresponding NUMA nodes. Format of the |
2663 | arguments: | |
39c7a2ca VF |
2664 | .RS |
2665 | .RS | |
523bad63 TK |
2666 | .P |
2667 | <mode>[:<nodelist>] | |
39c7a2ca | 2668 | .RE |
523bad63 | 2669 | .P |
f1dd3fb1 | 2670 | `mode' is one of the following memory policies: `default', `prefer', |
523bad63 TK |
2671 | `bind', `interleave' or `local'. For `default' and `local' memory |
2672 | policies, no node needs to be specified. For `prefer', only one node is | |
2673 | allowed. For `bind' and `interleave' the `nodelist' may be as | |
2674 | follows: a comma delimited list of numbers, A\-B ranges, or `all'. | |
39c7a2ca VF |
2675 | .RE |
2676 | .TP | |
523bad63 TK |
2677 | .BI cgroup \fR=\fPstr |
2678 | Add job to this control group. If it doesn't exist, it will be created. The | |
2679 | system must have a mounted cgroup blkio mount point for this to work. If | |
2680 | your system doesn't have it mounted, you can do so with: | |
d60e92d1 AC |
2681 | .RS |
2682 | .RS | |
d60e92d1 | 2683 | .P |
523bad63 TK |
2684 | # mount \-t cgroup \-o blkio none /cgroup |
2685 | .RE | |
d60e92d1 AC |
2686 | .RE |
2687 | .TP | |
523bad63 TK |
2688 | .BI cgroup_weight \fR=\fPint |
2689 | Set the weight of the cgroup to this value. See the documentation that comes | |
2690 | with the kernel, allowed values are in the range of 100..1000. | |
d60e92d1 | 2691 | .TP |
523bad63 TK |
2692 | .BI cgroup_nodelete \fR=\fPbool |
2693 | Normally fio will delete the cgroups it has created after the job | |
2694 | completion. To override this behavior and to leave cgroups around after the | |
2695 | job completion, set `cgroup_nodelete=1'. This can be useful if one wants | |
2696 | to inspect various cgroup files after job completion. Default: false. | |
c8eeb9df | 2697 | .TP |
523bad63 TK |
2698 | .BI flow_id \fR=\fPint |
2699 | The ID of the flow. If not specified, it defaults to being a global | |
2700 | flow. See \fBflow\fR. | |
d60e92d1 | 2701 | .TP |
523bad63 | 2702 | .BI flow \fR=\fPint |
d4e74fda DB |
2703 | Weight in token-based flow control. If this value is used, |
2704 | then fio regulates the activity between two or more jobs | |
2705 | sharing the same flow_id. | |
2706 | Fio attempts to keep each job activity proportional to other jobs' activities | |
2707 | in the same flow_id group, with respect to requested weight per job. | |
2708 | That is, if one job has `flow=3', another job has `flow=2' | |
2709 | and another with `flow=1`, then there will be a roughly 3:2:1 ratio | |
2710 | in how much one runs vs the others. | |
6b7f6851 | 2711 | .TP |
523bad63 | 2712 | .BI flow_sleep \fR=\fPint |
d4e74fda DB |
2713 | The period of time, in microseconds, to wait after the flow counter |
2714 | has exceeded its proportion before retrying operations. | |
25460cf6 | 2715 | .TP |
523bad63 TK |
2716 | .BI stonewall "\fR,\fB wait_for_previous" |
2717 | Wait for preceding jobs in the job file to exit, before starting this | |
2718 | one. Can be used to insert serialization points in the job file. A stone | |
2719 | wall also implies starting a new reporting group, see | |
fd56c235 AW |
2720 | \fBgroup_reporting\fR. Optionally you can use `stonewall=0` to disable or |
2721 | `stonewall=1` to enable it. | |
2378826d | 2722 | .TP |
523bad63 | 2723 | .BI exitall |
64402a8a HW |
2724 | By default, fio will continue running all other jobs when one job finishes. |
2725 | Sometimes this is not the desired action. Setting \fBexitall\fR will instead | |
2726 | make fio terminate all jobs in the same group, as soon as one job of that | |
2727 | group finishes. | |
2728 | .TP | |
fd56c235 | 2729 | .BI exit_what \fR=\fPstr |
64402a8a | 2730 | By default, fio will continue running all other jobs when one job finishes. |
fd56c235 | 2731 | Sometimes this is not the desired action. Setting \fBexitall\fR will instead |
64402a8a | 2732 | make fio terminate all jobs in the same group. The option \fBexit_what\fR |
fd56c235 AW |
2733 | allows you to control which jobs get terminated when \fBexitall\fR is enabled. |
2734 | The default value is \fBgroup\fR. | |
2735 | The allowed values are: | |
2736 | .RS | |
2737 | .RS | |
2738 | .TP | |
2739 | .B all | |
2740 | terminates all jobs. | |
2741 | .TP | |
2742 | .B group | |
2743 | is the default and does not change the behaviour of \fBexitall\fR. | |
2744 | .TP | |
2745 | .B stonewall | |
2746 | terminates all currently running jobs across all groups and continues | |
2747 | execution with the next stonewalled group. | |
2748 | .RE | |
2749 | .RE | |
e81ecca3 | 2750 | .TP |
523bad63 TK |
2751 | .BI exec_prerun \fR=\fPstr |
2752 | Before running this job, issue the command specified through | |
2753 | \fBsystem\fR\|(3). Output is redirected in a file called `jobname.prerun.txt'. | |
e9f48479 | 2754 | .TP |
523bad63 TK |
2755 | .BI exec_postrun \fR=\fPstr |
2756 | After the job completes, issue the command specified though | |
2757 | \fBsystem\fR\|(3). Output is redirected in a file called `jobname.postrun.txt'. | |
d60e92d1 | 2758 | .TP |
523bad63 TK |
2759 | .BI uid \fR=\fPint |
2760 | Instead of running as the invoking user, set the user ID to this value | |
2761 | before the thread/process does any work. | |
39c1c323 | 2762 | .TP |
523bad63 TK |
2763 | .BI gid \fR=\fPint |
2764 | Set group ID, see \fBuid\fR. | |
2765 | .SS "Verification" | |
d60e92d1 | 2766 | .TP |
589e88b7 | 2767 | .BI verify_only |
523bad63 | 2768 | Do not perform specified workload, only verify data still matches previous |
5e4c7118 | 2769 | invocation of this workload. This option allows one to check data multiple |
523bad63 TK |
2770 | times at a later date without overwriting it. This option makes sense only |
2771 | for workloads that write data, and does not support workloads with the | |
5e4c7118 JA |
2772 | \fBtime_based\fR option set. |
2773 | .TP | |
d60e92d1 | 2774 | .BI do_verify \fR=\fPbool |
523bad63 TK |
2775 | Run the verify phase after a write phase. Only valid if \fBverify\fR is |
2776 | set. Default: true. | |
d60e92d1 AC |
2777 | .TP |
2778 | .BI verify \fR=\fPstr | |
523bad63 TK |
2779 | If writing to a file, fio can verify the file contents after each iteration |
2780 | of the job. Each verification method also implies verification of special | |
2781 | header, which is written to the beginning of each block. This header also | |
2782 | includes meta information, like offset of the block, block number, timestamp | |
2783 | when block was written, etc. \fBverify\fR can be combined with | |
2784 | \fBverify_pattern\fR option. The allowed values are: | |
d60e92d1 AC |
2785 | .RS |
2786 | .RS | |
2787 | .TP | |
523bad63 TK |
2788 | .B md5 |
2789 | Use an md5 sum of the data area and store it in the header of | |
2790 | each block. | |
2791 | .TP | |
2792 | .B crc64 | |
2793 | Use an experimental crc64 sum of the data area and store it in the | |
2794 | header of each block. | |
2795 | .TP | |
2796 | .B crc32c | |
2797 | Use a crc32c sum of the data area and store it in the header of | |
2798 | each block. This will automatically use hardware acceleration | |
2799 | (e.g. SSE4.2 on an x86 or CRC crypto extensions on ARM64) but will | |
2800 | fall back to software crc32c if none is found. Generally the | |
f1dd3fb1 | 2801 | fastest checksum fio supports when hardware accelerated. |
523bad63 TK |
2802 | .TP |
2803 | .B crc32c\-intel | |
2804 | Synonym for crc32c. | |
2805 | .TP | |
2806 | .B crc32 | |
2807 | Use a crc32 sum of the data area and store it in the header of each | |
2808 | block. | |
2809 | .TP | |
2810 | .B crc16 | |
2811 | Use a crc16 sum of the data area and store it in the header of each | |
2812 | block. | |
2813 | .TP | |
2814 | .B crc7 | |
2815 | Use a crc7 sum of the data area and store it in the header of each | |
2816 | block. | |
2817 | .TP | |
2818 | .B xxhash | |
2819 | Use xxhash as the checksum function. Generally the fastest software | |
2820 | checksum that fio supports. | |
2821 | .TP | |
2822 | .B sha512 | |
2823 | Use sha512 as the checksum function. | |
2824 | .TP | |
2825 | .B sha256 | |
2826 | Use sha256 as the checksum function. | |
2827 | .TP | |
2828 | .B sha1 | |
2829 | Use optimized sha1 as the checksum function. | |
2830 | .TP | |
2831 | .B sha3\-224 | |
2832 | Use optimized sha3\-224 as the checksum function. | |
2833 | .TP | |
2834 | .B sha3\-256 | |
2835 | Use optimized sha3\-256 as the checksum function. | |
2836 | .TP | |
2837 | .B sha3\-384 | |
2838 | Use optimized sha3\-384 as the checksum function. | |
2839 | .TP | |
2840 | .B sha3\-512 | |
2841 | Use optimized sha3\-512 as the checksum function. | |
d60e92d1 AC |
2842 | .TP |
2843 | .B meta | |
523bad63 TK |
2844 | This option is deprecated, since now meta information is included in |
2845 | generic verification header and meta verification happens by | |
2846 | default. For detailed information see the description of the | |
2847 | \fBverify\fR setting. This option is kept because of | |
2848 | compatibility's sake with old configurations. Do not use it. | |
d60e92d1 | 2849 | .TP |
59245381 | 2850 | .B pattern |
523bad63 TK |
2851 | Verify a strict pattern. Normally fio includes a header with some |
2852 | basic information and checksumming, but if this option is set, only | |
2853 | the specific pattern set with \fBverify_pattern\fR is verified. | |
59245381 | 2854 | .TP |
d60e92d1 | 2855 | .B null |
523bad63 TK |
2856 | Only pretend to verify. Useful for testing internals with |
2857 | `ioengine=null', not for much else. | |
d60e92d1 | 2858 | .RE |
523bad63 TK |
2859 | .P |
2860 | This option can be used for repeated burn\-in tests of a system to make sure | |
2861 | that the written data is also correctly read back. If the data direction | |
2862 | given is a read or random read, fio will assume that it should verify a | |
2863 | previously written file. If the data direction includes any form of write, | |
2864 | the verify will be of the newly written data. | |
47e6a6e5 SW |
2865 | .P |
2866 | To avoid false verification errors, do not use the norandommap option when | |
2867 | verifying data with async I/O engines and I/O depths > 1. Or use the | |
2868 | norandommap and the lfsr random generator together to avoid writing to the | |
2869 | same offset with muliple outstanding I/Os. | |
d60e92d1 AC |
2870 | .RE |
2871 | .TP | |
f7fa2653 | 2872 | .BI verify_offset \fR=\fPint |
d60e92d1 | 2873 | Swap the verification header with data somewhere else in the block before |
523bad63 | 2874 | writing. It is swapped back before verifying. |
d60e92d1 | 2875 | .TP |
f7fa2653 | 2876 | .BI verify_interval \fR=\fPint |
523bad63 TK |
2877 | Write the verification header at a finer granularity than the |
2878 | \fBblocksize\fR. It will be written for chunks the size of | |
2879 | \fBverify_interval\fR. \fBblocksize\fR should divide this evenly. | |
d60e92d1 | 2880 | .TP |
996093bb | 2881 | .BI verify_pattern \fR=\fPstr |
523bad63 TK |
2882 | If set, fio will fill the I/O buffers with this pattern. Fio defaults to |
2883 | filling with totally random bytes, but sometimes it's interesting to fill | |
2884 | with a known pattern for I/O verification purposes. Depending on the width | |
2885 | of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time (it can | |
2886 | be either a decimal or a hex number). The \fBverify_pattern\fR if larger than | |
2887 | a 32\-bit quantity has to be a hex number that starts with either "0x" or | |
2888 | "0X". Use with \fBverify\fR. Also, \fBverify_pattern\fR supports %o | |
2889 | format, which means that for each block offset will be written and then | |
2890 | verified back, e.g.: | |
2fa5a241 RP |
2891 | .RS |
2892 | .RS | |
523bad63 TK |
2893 | .P |
2894 | verify_pattern=%o | |
2fa5a241 | 2895 | .RE |
523bad63 | 2896 | .P |
2fa5a241 | 2897 | Or use combination of everything: |
2fa5a241 | 2898 | .RS |
523bad63 TK |
2899 | .P |
2900 | verify_pattern=0xff%o"abcd"\-12 | |
2fa5a241 RP |
2901 | .RE |
2902 | .RE | |
996093bb | 2903 | .TP |
d60e92d1 | 2904 | .BI verify_fatal \fR=\fPbool |
523bad63 TK |
2905 | Normally fio will keep checking the entire contents before quitting on a |
2906 | block verification failure. If this option is set, fio will exit the job on | |
2907 | the first observed failure. Default: false. | |
d60e92d1 | 2908 | .TP |
b463e936 | 2909 | .BI verify_dump \fR=\fPbool |
523bad63 TK |
2910 | If set, dump the contents of both the original data block and the data block |
2911 | we read off disk to files. This allows later analysis to inspect just what | |
2912 | kind of data corruption occurred. Off by default. | |
b463e936 | 2913 | .TP |
e8462bd8 | 2914 | .BI verify_async \fR=\fPint |
523bad63 TK |
2915 | Fio will normally verify I/O inline from the submitting thread. This option |
2916 | takes an integer describing how many async offload threads to create for I/O | |
2917 | verification instead, causing fio to offload the duty of verifying I/O | |
2918 | contents to one or more separate threads. If using this offload option, even | |
2919 | sync I/O engines can benefit from using an \fBiodepth\fR setting higher | |
2920 | than 1, as it allows them to have I/O in flight while verifies are running. | |
2921 | Defaults to 0 async threads, i.e. verification is not asynchronous. | |
e8462bd8 JA |
2922 | .TP |
2923 | .BI verify_async_cpus \fR=\fPstr | |
523bad63 TK |
2924 | Tell fio to set the given CPU affinity on the async I/O verification |
2925 | threads. See \fBcpus_allowed\fR for the format used. | |
e8462bd8 | 2926 | .TP |
6f87418f JA |
2927 | .BI verify_backlog \fR=\fPint |
2928 | Fio will normally verify the written contents of a job that utilizes verify | |
2929 | once that job has completed. In other words, everything is written then | |
2930 | everything is read back and verified. You may want to verify continually | |
523bad63 TK |
2931 | instead for a variety of reasons. Fio stores the meta data associated with |
2932 | an I/O block in memory, so for large verify workloads, quite a bit of memory | |
2933 | would be used up holding this meta data. If this option is enabled, fio will | |
2934 | write only N blocks before verifying these blocks. | |
6f87418f JA |
2935 | .TP |
2936 | .BI verify_backlog_batch \fR=\fPint | |
523bad63 TK |
2937 | Control how many blocks fio will verify if \fBverify_backlog\fR is |
2938 | set. If not set, will default to the value of \fBverify_backlog\fR | |
2939 | (meaning the entire queue is read back and verified). If | |
2940 | \fBverify_backlog_batch\fR is less than \fBverify_backlog\fR then not all | |
2941 | blocks will be verified, if \fBverify_backlog_batch\fR is larger than | |
2942 | \fBverify_backlog\fR, some blocks will be verified more than once. | |
2943 | .TP | |
2944 | .BI verify_state_save \fR=\fPbool | |
2945 | When a job exits during the write phase of a verify workload, save its | |
2946 | current state. This allows fio to replay up until that point, if the verify | |
2947 | state is loaded for the verify read phase. The format of the filename is, | |
2948 | roughly: | |
2949 | .RS | |
2950 | .RS | |
2951 | .P | |
2952 | <type>\-<jobname>\-<jobindex>\-verify.state. | |
2953 | .RE | |
2954 | .P | |
2955 | <type> is "local" for a local run, "sock" for a client/server socket | |
2956 | connection, and "ip" (192.168.0.1, for instance) for a networked | |
2957 | client/server connection. Defaults to true. | |
2958 | .RE | |
2959 | .TP | |
2960 | .BI verify_state_load \fR=\fPbool | |
2961 | If a verify termination trigger was used, fio stores the current write state | |
2962 | of each thread. This can be used at verification time so that fio knows how | |
2963 | far it should verify. Without this information, fio will run a full | |
2964 | verification pass, according to the settings in the job file used. Default | |
2965 | false. | |
6f87418f | 2966 | .TP |
fa769d44 SW |
2967 | .BI trim_percentage \fR=\fPint |
2968 | Number of verify blocks to discard/trim. | |
2969 | .TP | |
2970 | .BI trim_verify_zero \fR=\fPbool | |
523bad63 | 2971 | Verify that trim/discarded blocks are returned as zeros. |
fa769d44 SW |
2972 | .TP |
2973 | .BI trim_backlog \fR=\fPint | |
523bad63 | 2974 | Verify that trim/discarded blocks are returned as zeros. |
fa769d44 SW |
2975 | .TP |
2976 | .BI trim_backlog_batch \fR=\fPint | |
523bad63 | 2977 | Trim this number of I/O blocks. |
fa769d44 SW |
2978 | .TP |
2979 | .BI experimental_verify \fR=\fPbool | |
2980 | Enable experimental verification. | |
523bad63 | 2981 | .SS "Steady state" |
fa769d44 | 2982 | .TP |
523bad63 TK |
2983 | .BI steadystate \fR=\fPstr:float "\fR,\fP ss" \fR=\fPstr:float |
2984 | Define the criterion and limit for assessing steady state performance. The | |
2985 | first parameter designates the criterion whereas the second parameter sets | |
2986 | the threshold. When the criterion falls below the threshold for the | |
2987 | specified duration, the job will stop. For example, `iops_slope:0.1%' will | |
2988 | direct fio to terminate the job when the least squares regression slope | |
2989 | falls below 0.1% of the mean IOPS. If \fBgroup_reporting\fR is enabled | |
2990 | this will apply to all jobs in the group. Below is the list of available | |
2991 | steady state assessment criteria. All assessments are carried out using only | |
2992 | data from the rolling collection window. Threshold limits can be expressed | |
2993 | as a fixed value or as a percentage of the mean in the collection window. | |
2994 | .RS | |
1cb049d9 VF |
2995 | .P |
2996 | When using this feature, most jobs should include the \fBtime_based\fR | |
2997 | and \fBruntime\fR options or the \fBloops\fR option so that fio does not | |
2998 | stop running after it has covered the full size of the specified file(s) | |
2999 | or device(s). | |
3000 | .RS | |
523bad63 | 3001 | .RS |
d60e92d1 | 3002 | .TP |
523bad63 TK |
3003 | .B iops |
3004 | Collect IOPS data. Stop the job if all individual IOPS measurements | |
3005 | are within the specified limit of the mean IOPS (e.g., `iops:2' | |
3006 | means that all individual IOPS values must be within 2 of the mean, | |
3007 | whereas `iops:0.2%' means that all individual IOPS values must be | |
3008 | within 0.2% of the mean IOPS to terminate the job). | |
d60e92d1 | 3009 | .TP |
523bad63 TK |
3010 | .B iops_slope |
3011 | Collect IOPS data and calculate the least squares regression | |
3012 | slope. Stop the job if the slope falls below the specified limit. | |
d60e92d1 | 3013 | .TP |
523bad63 TK |
3014 | .B bw |
3015 | Collect bandwidth data. Stop the job if all individual bandwidth | |
3016 | measurements are within the specified limit of the mean bandwidth. | |
64bbb865 | 3017 | .TP |
523bad63 TK |
3018 | .B bw_slope |
3019 | Collect bandwidth data and calculate the least squares regression | |
3020 | slope. Stop the job if the slope falls below the specified limit. | |
3021 | .RE | |
3022 | .RE | |
d1c46c04 | 3023 | .TP |
523bad63 TK |
3024 | .BI steadystate_duration \fR=\fPtime "\fR,\fP ss_dur" \fR=\fPtime |
3025 | A rolling window of this duration will be used to judge whether steady state | |
3026 | has been reached. Data will be collected once per second. The default is 0 | |
3027 | which disables steady state detection. When the unit is omitted, the | |
3028 | value is interpreted in seconds. | |
0c63576e | 3029 | .TP |
523bad63 TK |
3030 | .BI steadystate_ramp_time \fR=\fPtime "\fR,\fP ss_ramp" \fR=\fPtime |
3031 | Allow the job to run for the specified duration before beginning data | |
3032 | collection for checking the steady state job termination criterion. The | |
3033 | default is 0. When the unit is omitted, the value is interpreted in seconds. | |
3034 | .SS "Measurements and reporting" | |
0c63576e | 3035 | .TP |
3a5db920 JA |
3036 | .BI per_job_logs \fR=\fPbool |
3037 | If set, this generates bw/clat/iops log with per file private filenames. If | |
523bad63 TK |
3038 | not set, jobs with identical names will share the log filename. Default: |
3039 | true. | |
3040 | .TP | |
3041 | .BI group_reporting | |
3042 | It may sometimes be interesting to display statistics for groups of jobs as | |
3043 | a whole instead of for each individual job. This is especially true if | |
3044 | \fBnumjobs\fR is used; looking at individual thread/process output | |
338f2db5 SW |
3045 | quickly becomes unwieldy. To see the final report per-group instead of |
3046 | per-job, use \fBgroup_reporting\fR. Jobs in a file will be part of the | |
523bad63 TK |
3047 | same reporting group, unless if separated by a \fBstonewall\fR, or by |
3048 | using \fBnew_group\fR. | |
3049 | .TP | |
3050 | .BI new_group | |
3051 | Start a new reporting group. See: \fBgroup_reporting\fR. If not given, | |
3052 | all jobs in a file will be part of the same reporting group, unless | |
3053 | separated by a \fBstonewall\fR. | |
3054 | .TP | |
3055 | .BI stats \fR=\fPbool | |
3056 | By default, fio collects and shows final output results for all jobs | |
3057 | that run. If this option is set to 0, then fio will ignore it in | |
3058 | the final stat output. | |
3a5db920 | 3059 | .TP |
836bad52 | 3060 | .BI write_bw_log \fR=\fPstr |
523bad63 | 3061 | If given, write a bandwidth log for this job. Can be used to store data of |
074f0817 | 3062 | the bandwidth of the jobs in their lifetime. |
523bad63 | 3063 | .RS |
074f0817 SW |
3064 | .P |
3065 | If no str argument is given, the default filename of | |
3066 | `jobname_type.x.log' is used. Even when the argument is given, fio | |
3067 | will still append the type of log. So if one specifies: | |
523bad63 TK |
3068 | .RS |
3069 | .P | |
074f0817 | 3070 | write_bw_log=foo |
523bad63 TK |
3071 | .RE |
3072 | .P | |
074f0817 SW |
3073 | The actual log name will be `foo_bw.x.log' where `x' is the index |
3074 | of the job (1..N, where N is the number of jobs). If | |
3075 | \fBper_job_logs\fR is false, then the filename will not include the | |
3076 | `.x` job index. | |
3077 | .P | |
3078 | The included \fBfio_generate_plots\fR script uses gnuplot to turn these | |
3079 | text files into nice graphs. See the \fBLOG FILE FORMATS\fR section for how data is | |
3080 | structured within the file. | |
523bad63 | 3081 | .RE |
901bb994 | 3082 | .TP |
074f0817 SW |
3083 | .BI write_lat_log \fR=\fPstr |
3084 | Same as \fBwrite_bw_log\fR, except this option creates I/O | |
3085 | submission (e.g., `name_slat.x.log'), completion (e.g., | |
3086 | `name_clat.x.log'), and total (e.g., `name_lat.x.log') latency | |
3087 | files instead. See \fBwrite_bw_log\fR for details about the | |
3088 | filename format and the \fBLOG FILE FORMATS\fR section for how data is structured | |
3089 | within the files. | |
3090 | .TP | |
1e613c9c | 3091 | .BI write_hist_log \fR=\fPstr |
074f0817 SW |
3092 | Same as \fBwrite_bw_log\fR but writes an I/O completion latency |
3093 | histogram file (e.g., `name_hist.x.log') instead. Note that this | |
3094 | file will be empty unless \fBlog_hist_msec\fR has also been set. | |
3095 | See \fBwrite_bw_log\fR for details about the filename format and | |
3096 | the \fBLOG FILE FORMATS\fR section for how data is structured | |
3097 | within the file. | |
1e613c9c | 3098 | .TP |
c8eeb9df | 3099 | .BI write_iops_log \fR=\fPstr |
074f0817 | 3100 | Same as \fBwrite_bw_log\fR, but writes an IOPS file (e.g. |
15417073 SW |
3101 | `name_iops.x.log`) instead. Because fio defaults to individual |
3102 | I/O logging, the value entry in the IOPS log will be 1 unless windowed | |
3103 | logging (see \fBlog_avg_msec\fR) has been enabled. See | |
3104 | \fBwrite_bw_log\fR for details about the filename format and \fBLOG | |
3105 | FILE FORMATS\fR for how data is structured within the file. | |
c8eeb9df | 3106 | .TP |
b8bc8cba JA |
3107 | .BI log_avg_msec \fR=\fPint |
3108 | By default, fio will log an entry in the iops, latency, or bw log for every | |
523bad63 | 3109 | I/O that completes. When writing to the disk log, that can quickly grow to a |
b8bc8cba | 3110 | very large size. Setting this option makes fio average the each log entry |
e6989e10 | 3111 | over the specified period of time, reducing the resolution of the log. See |
523bad63 TK |
3112 | \fBlog_max_value\fR as well. Defaults to 0, logging all entries. |
3113 | Also see \fBLOG FILE FORMATS\fR section. | |
b8bc8cba | 3114 | .TP |
1e613c9c | 3115 | .BI log_hist_msec \fR=\fPint |
523bad63 TK |
3116 | Same as \fBlog_avg_msec\fR, but logs entries for completion latency |
3117 | histograms. Computing latency percentiles from averages of intervals using | |
3118 | \fBlog_avg_msec\fR is inaccurate. Setting this option makes fio log | |
3119 | histogram entries over the specified period of time, reducing log sizes for | |
3120 | high IOPS devices while retaining percentile accuracy. See | |
074f0817 SW |
3121 | \fBlog_hist_coarseness\fR and \fBwrite_hist_log\fR as well. |
3122 | Defaults to 0, meaning histogram logging is disabled. | |
1e613c9c KC |
3123 | .TP |
3124 | .BI log_hist_coarseness \fR=\fPint | |
523bad63 TK |
3125 | Integer ranging from 0 to 6, defining the coarseness of the resolution of |
3126 | the histogram logs enabled with \fBlog_hist_msec\fR. For each increment | |
3127 | in coarseness, fio outputs half as many bins. Defaults to 0, for which | |
3128 | histogram logs contain 1216 latency bins. See \fBLOG FILE FORMATS\fR section. | |
3129 | .TP | |
3130 | .BI log_max_value \fR=\fPbool | |
3131 | If \fBlog_avg_msec\fR is set, fio logs the average over that window. If | |
3132 | you instead want to log the maximum value, set this option to 1. Defaults to | |
3133 | 0, meaning that averaged values are logged. | |
1e613c9c | 3134 | .TP |
ae588852 | 3135 | .BI log_offset \fR=\fPbool |
523bad63 TK |
3136 | If this is set, the iolog options will include the byte offset for the I/O |
3137 | entry as well as the other data values. Defaults to 0 meaning that | |
3138 | offsets are not present in logs. Also see \fBLOG FILE FORMATS\fR section. | |
ae588852 | 3139 | .TP |
aee2ab67 | 3140 | .BI log_compression \fR=\fPint |
523bad63 TK |
3141 | If this is set, fio will compress the I/O logs as it goes, to keep the |
3142 | memory footprint lower. When a log reaches the specified size, that chunk is | |
3143 | removed and compressed in the background. Given that I/O logs are fairly | |
3144 | highly compressible, this yields a nice memory savings for longer runs. The | |
3145 | downside is that the compression will consume some background CPU cycles, so | |
3146 | it may impact the run. This, however, is also true if the logging ends up | |
3147 | consuming most of the system memory. So pick your poison. The I/O logs are | |
3148 | saved normally at the end of a run, by decompressing the chunks and storing | |
3149 | them in the specified log file. This feature depends on the availability of | |
3150 | zlib. | |
aee2ab67 | 3151 | .TP |
c08f9fe2 | 3152 | .BI log_compression_cpus \fR=\fPstr |
523bad63 TK |
3153 | Define the set of CPUs that are allowed to handle online log compression for |
3154 | the I/O jobs. This can provide better isolation between performance | |
0cf90a62 SW |
3155 | sensitive jobs, and background compression work. See \fBcpus_allowed\fR for |
3156 | the format used. | |
c08f9fe2 | 3157 | .TP |
b26317c9 | 3158 | .BI log_store_compressed \fR=\fPbool |
c08f9fe2 | 3159 | If set, fio will store the log files in a compressed format. They can be |
523bad63 TK |
3160 | decompressed with fio, using the \fB\-\-inflate\-log\fR command line |
3161 | parameter. The files will be stored with a `.fz' suffix. | |
b26317c9 | 3162 | .TP |
3aea75b1 KC |
3163 | .BI log_unix_epoch \fR=\fPbool |
3164 | If set, fio will log Unix timestamps to the log files produced by enabling | |
338f2db5 | 3165 | write_type_log for each log type, instead of the default zero-based |
3aea75b1 KC |
3166 | timestamps. |
3167 | .TP | |
66347cfa | 3168 | .BI block_error_percentiles \fR=\fPbool |
338f2db5 | 3169 | If set, record errors in trim block-sized units from writes and trims and |
523bad63 TK |
3170 | output a histogram of how many trims it took to get to errors, and what kind |
3171 | of error was encountered. | |
d60e92d1 | 3172 | .TP |
523bad63 TK |
3173 | .BI bwavgtime \fR=\fPint |
3174 | Average the calculated bandwidth over the given time. Value is specified in | |
3175 | milliseconds. If the job also does bandwidth logging through | |
3176 | \fBwrite_bw_log\fR, then the minimum of this option and | |
3177 | \fBlog_avg_msec\fR will be used. Default: 500ms. | |
d60e92d1 | 3178 | .TP |
523bad63 TK |
3179 | .BI iopsavgtime \fR=\fPint |
3180 | Average the calculated IOPS over the given time. Value is specified in | |
3181 | milliseconds. If the job also does IOPS logging through | |
3182 | \fBwrite_iops_log\fR, then the minimum of this option and | |
3183 | \fBlog_avg_msec\fR will be used. Default: 500ms. | |
d60e92d1 | 3184 | .TP |
d60e92d1 | 3185 | .BI disk_util \fR=\fPbool |
523bad63 TK |
3186 | Generate disk utilization statistics, if the platform supports it. |
3187 | Default: true. | |
fa769d44 | 3188 | .TP |
523bad63 TK |
3189 | .BI disable_lat \fR=\fPbool |
3190 | Disable measurements of total latency numbers. Useful only for cutting back | |
3191 | the number of calls to \fBgettimeofday\fR\|(2), as that does impact | |
3192 | performance at really high IOPS rates. Note that to really get rid of a | |
3193 | large amount of these calls, this option must be used with | |
3194 | \fBdisable_slat\fR and \fBdisable_bw_measurement\fR as well. | |
9e684a49 | 3195 | .TP |
523bad63 TK |
3196 | .BI disable_clat \fR=\fPbool |
3197 | Disable measurements of completion latency numbers. See | |
3198 | \fBdisable_lat\fR. | |
9e684a49 | 3199 | .TP |
523bad63 TK |
3200 | .BI disable_slat \fR=\fPbool |
3201 | Disable measurements of submission latency numbers. See | |
3202 | \fBdisable_lat\fR. | |
9e684a49 | 3203 | .TP |
523bad63 TK |
3204 | .BI disable_bw_measurement \fR=\fPbool "\fR,\fP disable_bw" \fR=\fPbool |
3205 | Disable measurements of throughput/bandwidth numbers. See | |
3206 | \fBdisable_lat\fR. | |
9e684a49 | 3207 | .TP |
dd39b9ce VF |
3208 | .BI slat_percentiles \fR=\fPbool |
3209 | Report submission latency percentiles. Submission latency is not recorded | |
3210 | for synchronous ioengines. | |
3211 | .TP | |
83349190 | 3212 | .BI clat_percentiles \fR=\fPbool |
dd39b9ce | 3213 | Report completion latency percentiles. |
b599759b JA |
3214 | .TP |
3215 | .BI lat_percentiles \fR=\fPbool | |
dd39b9ce VF |
3216 | Report total latency percentiles. Total latency is the sum of submission |
3217 | latency and completion latency. | |
83349190 YH |
3218 | .TP |
3219 | .BI percentile_list \fR=\fPfloat_list | |
dd39b9ce VF |
3220 | Overwrite the default list of percentiles for latencies and the |
3221 | block error histogram. Each number is a floating point number in the range | |
523bad63 | 3222 | (0,100], and the maximum length of the list is 20. Use ':' to separate the |
dd39b9ce VF |
3223 | numbers. For example, `\-\-percentile_list=99.5:99.9' will cause fio to |
3224 | report the latency durations below which 99.5% and 99.9% of the observed | |
3225 | latencies fell, respectively. | |
e883cb35 JF |
3226 | .TP |
3227 | .BI significant_figures \fR=\fPint | |
c32ba107 JA |
3228 | If using \fB\-\-output\-format\fR of `normal', set the significant figures |
3229 | to this value. Higher values will yield more precise IOPS and throughput | |
3230 | units, while lower values will round. Requires a minimum value of 1 and a | |
e883cb35 | 3231 | maximum value of 10. Defaults to 4. |
523bad63 | 3232 | .SS "Error handling" |
e4585935 | 3233 | .TP |
523bad63 TK |
3234 | .BI exitall_on_error |
3235 | When one job finishes in error, terminate the rest. The default is to wait | |
3236 | for each job to finish. | |
e4585935 | 3237 | .TP |
523bad63 TK |
3238 | .BI continue_on_error \fR=\fPstr |
3239 | Normally fio will exit the job on the first observed failure. If this option | |
338f2db5 | 3240 | is set, fio will continue the job when there is a 'non-fatal error' (EIO or |
523bad63 TK |
3241 | EILSEQ) until the runtime is exceeded or the I/O size specified is |
3242 | completed. If this option is used, there are two more stats that are | |
3243 | appended, the total error count and the first error. The error field given | |
3244 | in the stats is the first error that was hit during the run. | |
3245 | The allowed values are: | |
3246 | .RS | |
3247 | .RS | |
046395d7 | 3248 | .TP |
523bad63 TK |
3249 | .B none |
3250 | Exit on any I/O or verify errors. | |
de890a1e | 3251 | .TP |
523bad63 TK |
3252 | .B read |
3253 | Continue on read errors, exit on all others. | |
2cafffbe | 3254 | .TP |
523bad63 TK |
3255 | .B write |
3256 | Continue on write errors, exit on all others. | |
a0679ce5 | 3257 | .TP |
523bad63 TK |
3258 | .B io |
3259 | Continue on any I/O error, exit on all others. | |
de890a1e | 3260 | .TP |
523bad63 TK |
3261 | .B verify |
3262 | Continue on verify errors, exit on all others. | |
de890a1e | 3263 | .TP |
523bad63 TK |
3264 | .B all |
3265 | Continue on all errors. | |
b93b6a2e | 3266 | .TP |
523bad63 | 3267 | .B 0 |
338f2db5 | 3268 | Backward-compatible alias for 'none'. |
d3a623de | 3269 | .TP |
523bad63 | 3270 | .B 1 |
338f2db5 | 3271 | Backward-compatible alias for 'all'. |
523bad63 TK |
3272 | .RE |
3273 | .RE | |
1d360ffb | 3274 | .TP |
523bad63 TK |
3275 | .BI ignore_error \fR=\fPstr |
3276 | Sometimes you want to ignore some errors during test in that case you can | |
3277 | specify error list for each error type, instead of only being able to | |
338f2db5 | 3278 | ignore the default 'non-fatal error' using \fBcontinue_on_error\fR. |
523bad63 TK |
3279 | `ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST' errors for |
3280 | given error type is separated with ':'. Error may be symbol ('ENOSPC', 'ENOMEM') | |
3281 | or integer. Example: | |
de890a1e SL |
3282 | .RS |
3283 | .RS | |
523bad63 TK |
3284 | .P |
3285 | ignore_error=EAGAIN,ENOSPC:122 | |
3286 | .RE | |
3287 | .P | |
3288 | This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from | |
3289 | WRITE. This option works by overriding \fBcontinue_on_error\fR with | |
3290 | the list of errors for each error type if any. | |
3291 | .RE | |
de890a1e | 3292 | .TP |
523bad63 TK |
3293 | .BI error_dump \fR=\fPbool |
3294 | If set dump every error even if it is non fatal, true by default. If | |
3295 | disabled only fatal error will be dumped. | |
3296 | .SS "Running predefined workloads" | |
3297 | Fio includes predefined profiles that mimic the I/O workloads generated by | |
3298 | other tools. | |
49ccb8c1 | 3299 | .TP |
523bad63 TK |
3300 | .BI profile \fR=\fPstr |
3301 | The predefined workload to run. Current profiles are: | |
3302 | .RS | |
3303 | .RS | |
de890a1e | 3304 | .TP |
523bad63 TK |
3305 | .B tiobench |
3306 | Threaded I/O bench (tiotest/tiobench) like workload. | |
49ccb8c1 | 3307 | .TP |
523bad63 TK |
3308 | .B act |
3309 | Aerospike Certification Tool (ACT) like workload. | |
3310 | .RE | |
de890a1e SL |
3311 | .RE |
3312 | .P | |
523bad63 TK |
3313 | To view a profile's additional options use \fB\-\-cmdhelp\fR after specifying |
3314 | the profile. For example: | |
3315 | .RS | |
3316 | .TP | |
3317 | $ fio \-\-profile=act \-\-cmdhelp | |
de890a1e | 3318 | .RE |
523bad63 | 3319 | .SS "Act profile options" |
de890a1e | 3320 | .TP |
523bad63 TK |
3321 | .BI device\-names \fR=\fPstr |
3322 | Devices to use. | |
d54fce84 | 3323 | .TP |
523bad63 TK |
3324 | .BI load \fR=\fPint |
3325 | ACT load multiplier. Default: 1. | |
7aeb1e94 | 3326 | .TP |
523bad63 TK |
3327 | .BI test\-duration\fR=\fPtime |
3328 | How long the entire test takes to run. When the unit is omitted, the value | |
3329 | is given in seconds. Default: 24h. | |
1008602c | 3330 | .TP |
523bad63 TK |
3331 | .BI threads\-per\-queue\fR=\fPint |
3332 | Number of read I/O threads per device. Default: 8. | |
e5f34d95 | 3333 | .TP |
523bad63 TK |
3334 | .BI read\-req\-num\-512\-blocks\fR=\fPint |
3335 | Number of 512B blocks to read at the time. Default: 3. | |
d54fce84 | 3336 | .TP |
523bad63 TK |
3337 | .BI large\-block\-op\-kbytes\fR=\fPint |
3338 | Size of large block ops in KiB (writes). Default: 131072. | |
d54fce84 | 3339 | .TP |
523bad63 TK |
3340 | .BI prep |
3341 | Set to run ACT prep phase. | |
3342 | .SS "Tiobench profile options" | |
6d500c2e | 3343 | .TP |
523bad63 TK |
3344 | .BI size\fR=\fPstr |
3345 | Size in MiB. | |
0d978694 | 3346 | .TP |
523bad63 TK |
3347 | .BI block\fR=\fPint |
3348 | Block size in bytes. Default: 4096. | |
0d978694 | 3349 | .TP |
523bad63 TK |
3350 | .BI numruns\fR=\fPint |
3351 | Number of runs. | |
0d978694 | 3352 | .TP |
523bad63 TK |
3353 | .BI dir\fR=\fPstr |
3354 | Test directory. | |
65fa28ca | 3355 | .TP |
523bad63 TK |
3356 | .BI threads\fR=\fPint |
3357 | Number of threads. | |
d60e92d1 | 3358 | .SH OUTPUT |
40943b9a TK |
3359 | Fio spits out a lot of output. While running, fio will display the status of the |
3360 | jobs created. An example of that would be: | |
d60e92d1 | 3361 | .P |
40943b9a TK |
3362 | .nf |
3363 | Jobs: 1 (f=1): [_(1),M(1)][24.8%][r=20.5MiB/s,w=23.5MiB/s][r=82,w=94 IOPS][eta 01m:31s] | |
3364 | .fi | |
d1429b5c | 3365 | .P |
40943b9a TK |
3366 | The characters inside the first set of square brackets denote the current status of |
3367 | each thread. The first character is the first job defined in the job file, and so | |
3368 | forth. The possible values (in typical life cycle order) are: | |
d60e92d1 AC |
3369 | .RS |
3370 | .TP | |
40943b9a | 3371 | .PD 0 |
d60e92d1 | 3372 | .B P |
40943b9a | 3373 | Thread setup, but not started. |
d60e92d1 AC |
3374 | .TP |
3375 | .B C | |
3376 | Thread created. | |
3377 | .TP | |
3378 | .B I | |
40943b9a TK |
3379 | Thread initialized, waiting or generating necessary data. |
3380 | .TP | |
522c29f6 | 3381 | .B p |
338f2db5 | 3382 | Thread running pre-reading file(s). |
40943b9a TK |
3383 | .TP |
3384 | .B / | |
3385 | Thread is in ramp period. | |
d60e92d1 AC |
3386 | .TP |
3387 | .B R | |
3388 | Running, doing sequential reads. | |
3389 | .TP | |
3390 | .B r | |
3391 | Running, doing random reads. | |
3392 | .TP | |
3393 | .B W | |
3394 | Running, doing sequential writes. | |
3395 | .TP | |
3396 | .B w | |
3397 | Running, doing random writes. | |
3398 | .TP | |
3399 | .B M | |
3400 | Running, doing mixed sequential reads/writes. | |
3401 | .TP | |
3402 | .B m | |
3403 | Running, doing mixed random reads/writes. | |
3404 | .TP | |
40943b9a TK |
3405 | .B D |
3406 | Running, doing sequential trims. | |
3407 | .TP | |
3408 | .B d | |
3409 | Running, doing random trims. | |
3410 | .TP | |
d60e92d1 AC |
3411 | .B F |
3412 | Running, currently waiting for \fBfsync\fR\|(2). | |
3413 | .TP | |
3414 | .B V | |
40943b9a TK |
3415 | Running, doing verification of written data. |
3416 | .TP | |
3417 | .B f | |
3418 | Thread finishing. | |
d60e92d1 AC |
3419 | .TP |
3420 | .B E | |
40943b9a | 3421 | Thread exited, not reaped by main thread yet. |
d60e92d1 AC |
3422 | .TP |
3423 | .B \- | |
40943b9a TK |
3424 | Thread reaped. |
3425 | .TP | |
3426 | .B X | |
3427 | Thread reaped, exited with an error. | |
3428 | .TP | |
3429 | .B K | |
3430 | Thread reaped, exited due to signal. | |
d1429b5c | 3431 | .PD |
40943b9a TK |
3432 | .RE |
3433 | .P | |
3434 | Fio will condense the thread string as not to take up more space on the command | |
3435 | line than needed. For instance, if you have 10 readers and 10 writers running, | |
3436 | the output would look like this: | |
3437 | .P | |
3438 | .nf | |
3439 | Jobs: 20 (f=20): [R(10),W(10)][4.0%][r=20.5MiB/s,w=23.5MiB/s][r=82,w=94 IOPS][eta 57m:36s] | |
3440 | .fi | |
d60e92d1 | 3441 | .P |
40943b9a TK |
3442 | Note that the status string is displayed in order, so it's possible to tell which of |
3443 | the jobs are currently doing what. In the example above this means that jobs 1\-\-10 | |
3444 | are readers and 11\-\-20 are writers. | |
d60e92d1 | 3445 | .P |
40943b9a TK |
3446 | The other values are fairly self explanatory \-\- number of threads currently |
3447 | running and doing I/O, the number of currently open files (f=), the estimated | |
3448 | completion percentage, the rate of I/O since last check (read speed listed first, | |
3449 | then write speed and optionally trim speed) in terms of bandwidth and IOPS, | |
3450 | and time to completion for the current running group. It's impossible to estimate | |
3451 | runtime of the following groups (if any). | |
d60e92d1 | 3452 | .P |
40943b9a TK |
3453 | When fio is done (or interrupted by Ctrl\-C), it will show the data for |
3454 | each thread, group of threads, and disks in that order. For each overall thread (or | |
3455 | group) the output looks like: | |
3456 | .P | |
3457 | .nf | |
3458 | Client1: (groupid=0, jobs=1): err= 0: pid=16109: Sat Jun 24 12:07:54 2017 | |
3459 | write: IOPS=88, BW=623KiB/s (638kB/s)(30.4MiB/50032msec) | |
3460 | slat (nsec): min=500, max=145500, avg=8318.00, stdev=4781.50 | |
3461 | clat (usec): min=170, max=78367, avg=4019.02, stdev=8293.31 | |
3462 | lat (usec): min=174, max=78375, avg=4027.34, stdev=8291.79 | |
3463 | clat percentiles (usec): | |
3464 | | 1.00th=[ 302], 5.00th=[ 326], 10.00th=[ 343], 20.00th=[ 363], | |
3465 | | 30.00th=[ 392], 40.00th=[ 404], 50.00th=[ 416], 60.00th=[ 445], | |
3466 | | 70.00th=[ 816], 80.00th=[ 6718], 90.00th=[12911], 95.00th=[21627], | |
3467 | | 99.00th=[43779], 99.50th=[51643], 99.90th=[68682], 99.95th=[72877], | |
3468 | | 99.99th=[78119] | |
3469 | bw ( KiB/s): min= 532, max= 686, per=0.10%, avg=622.87, stdev=24.82, samples= 100 | |
3470 | iops : min= 76, max= 98, avg=88.98, stdev= 3.54, samples= 100 | |
d3b9694d VF |
3471 | lat (usec) : 250=0.04%, 500=64.11%, 750=4.81%, 1000=2.79% |
3472 | lat (msec) : 2=4.16%, 4=1.84%, 10=4.90%, 20=11.33%, 50=5.37% | |
3473 | lat (msec) : 100=0.65% | |
40943b9a TK |
3474 | cpu : usr=0.27%, sys=0.18%, ctx=12072, majf=0, minf=21 |
3475 | IO depths : 1=85.0%, 2=13.1%, 4=1.8%, 8=0.1%, 16=0.0%, 32=0.0%, >=64=0.0% | |
3476 | submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% | |
3477 | complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% | |
3478 | issued rwt: total=0,4450,0, short=0,0,0, dropped=0,0,0 | |
3479 | latency : target=0, window=0, percentile=100.00%, depth=8 | |
3480 | .fi | |
3481 | .P | |
3482 | The job name (or first job's name when using \fBgroup_reporting\fR) is printed, | |
3483 | along with the group id, count of jobs being aggregated, last error id seen (which | |
3484 | is 0 when there are no errors), pid/tid of that thread and the time the job/group | |
3485 | completed. Below are the I/O statistics for each data direction performed (showing | |
3486 | writes in the example above). In the order listed, they denote: | |
d60e92d1 | 3487 | .RS |
d60e92d1 | 3488 | .TP |
40943b9a TK |
3489 | .B read/write/trim |
3490 | The string before the colon shows the I/O direction the statistics | |
3491 | are for. \fIIOPS\fR is the average I/Os performed per second. \fIBW\fR | |
3492 | is the average bandwidth rate shown as: value in power of 2 format | |
3493 | (value in power of 10 format). The last two values show: (total | |
3494 | I/O performed in power of 2 format / \fIruntime\fR of that thread). | |
d60e92d1 AC |
3495 | .TP |
3496 | .B slat | |
40943b9a TK |
3497 | Submission latency (\fImin\fR being the minimum, \fImax\fR being the |
3498 | maximum, \fIavg\fR being the average, \fIstdev\fR being the standard | |
3499 | deviation). This is the time it took to submit the I/O. For | |
3500 | sync I/O this row is not displayed as the slat is really the | |
3501 | completion latency (since queue/complete is one operation there). | |
3502 | This value can be in nanoseconds, microseconds or milliseconds \-\-\- | |
3503 | fio will choose the most appropriate base and print that (in the | |
3504 | example above nanoseconds was the best scale). Note: in \fB\-\-minimal\fR mode | |
3505 | latencies are always expressed in microseconds. | |
d60e92d1 AC |
3506 | .TP |
3507 | .B clat | |
40943b9a TK |
3508 | Completion latency. Same names as slat, this denotes the time from |
3509 | submission to completion of the I/O pieces. For sync I/O, clat will | |
3510 | usually be equal (or very close) to 0, as the time from submit to | |
3511 | complete is basically just CPU time (I/O has already been done, see slat | |
3512 | explanation). | |
d60e92d1 | 3513 | .TP |
d3b9694d VF |
3514 | .B lat |
3515 | Total latency. Same names as slat and clat, this denotes the time from | |
3516 | when fio created the I/O unit to completion of the I/O operation. | |
3517 | .TP | |
d60e92d1 | 3518 | .B bw |
40943b9a TK |
3519 | Bandwidth statistics based on samples. Same names as the xlat stats, |
3520 | but also includes the number of samples taken (\fIsamples\fR) and an | |
3521 | approximate percentage of total aggregate bandwidth this thread | |
3522 | received in its group (\fIper\fR). This last value is only really | |
3523 | useful if the threads in this group are on the same disk, since they | |
3524 | are then competing for disk access. | |
3525 | .TP | |
3526 | .B iops | |
3527 | IOPS statistics based on samples. Same names as \fBbw\fR. | |
d60e92d1 | 3528 | .TP |
d3b9694d VF |
3529 | .B lat (nsec/usec/msec) |
3530 | The distribution of I/O completion latencies. This is the time from when | |
3531 | I/O leaves fio and when it gets completed. Unlike the separate | |
3532 | read/write/trim sections above, the data here and in the remaining | |
3533 | sections apply to all I/Os for the reporting group. 250=0.04% means that | |
3534 | 0.04% of the I/Os completed in under 250us. 500=64.11% means that 64.11% | |
3535 | of the I/Os required 250 to 499us for completion. | |
3536 | .TP | |
d60e92d1 | 3537 | .B cpu |
40943b9a TK |
3538 | CPU usage. User and system time, along with the number of context |
3539 | switches this thread went through, usage of system and user time, and | |
3540 | finally the number of major and minor page faults. The CPU utilization | |
3541 | numbers are averages for the jobs in that reporting group, while the | |
3542 | context and fault counters are summed. | |
d60e92d1 AC |
3543 | .TP |
3544 | .B IO depths | |
40943b9a TK |
3545 | The distribution of I/O depths over the job lifetime. The numbers are |
3546 | divided into powers of 2 and each entry covers depths from that value | |
3547 | up to those that are lower than the next entry \-\- e.g., 16= covers | |
3548 | depths from 16 to 31. Note that the range covered by a depth | |
3549 | distribution entry can be different to the range covered by the | |
3550 | equivalent \fBsubmit\fR/\fBcomplete\fR distribution entry. | |
3551 | .TP | |
3552 | .B IO submit | |
3553 | How many pieces of I/O were submitting in a single submit call. Each | |
3554 | entry denotes that amount and below, until the previous entry \-\- e.g., | |
3555 | 16=100% means that we submitted anywhere between 9 to 16 I/Os per submit | |
3556 | call. Note that the range covered by a \fBsubmit\fR distribution entry can | |
3557 | be different to the range covered by the equivalent depth distribution | |
3558 | entry. | |
3559 | .TP | |
3560 | .B IO complete | |
3561 | Like the above \fBsubmit\fR number, but for completions instead. | |
3562 | .TP | |
3563 | .B IO issued rwt | |
3564 | The number of \fBread/write/trim\fR requests issued, and how many of them were | |
3565 | short or dropped. | |
d60e92d1 | 3566 | .TP |
d3b9694d | 3567 | .B IO latency |
ee21ebee | 3568 | These values are for \fBlatency_target\fR and related options. When |
d3b9694d VF |
3569 | these options are engaged, this section describes the I/O depth required |
3570 | to meet the specified latency target. | |
d60e92d1 | 3571 | .RE |
d60e92d1 | 3572 | .P |
40943b9a TK |
3573 | After each client has been listed, the group statistics are printed. They |
3574 | will look like this: | |
3575 | .P | |
3576 | .nf | |
3577 | Run status group 0 (all jobs): | |
3578 | READ: bw=20.9MiB/s (21.9MB/s), 10.4MiB/s\-10.8MiB/s (10.9MB/s\-11.3MB/s), io=64.0MiB (67.1MB), run=2973\-3069msec | |
3579 | WRITE: bw=1231KiB/s (1261kB/s), 616KiB/s\-621KiB/s (630kB/s\-636kB/s), io=64.0MiB (67.1MB), run=52747\-53223msec | |
3580 | .fi | |
3581 | .P | |
3582 | For each data direction it prints: | |
d60e92d1 AC |
3583 | .RS |
3584 | .TP | |
40943b9a TK |
3585 | .B bw |
3586 | Aggregate bandwidth of threads in this group followed by the | |
3587 | minimum and maximum bandwidth of all the threads in this group. | |
338f2db5 SW |
3588 | Values outside of brackets are power-of-2 format and those |
3589 | within are the equivalent value in a power-of-10 format. | |
d60e92d1 | 3590 | .TP |
40943b9a TK |
3591 | .B io |
3592 | Aggregate I/O performed of all threads in this group. The | |
3593 | format is the same as \fBbw\fR. | |
d60e92d1 | 3594 | .TP |
40943b9a TK |
3595 | .B run |
3596 | The smallest and longest runtimes of the threads in this group. | |
d60e92d1 | 3597 | .RE |
d60e92d1 | 3598 | .P |
40943b9a TK |
3599 | And finally, the disk statistics are printed. This is Linux specific. |
3600 | They will look like this: | |
3601 | .P | |
3602 | .nf | |
3603 | Disk stats (read/write): | |
3604 | sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00% | |
3605 | .fi | |
3606 | .P | |
3607 | Each value is printed for both reads and writes, with reads first. The | |
3608 | numbers denote: | |
d60e92d1 AC |
3609 | .RS |
3610 | .TP | |
3611 | .B ios | |
3612 | Number of I/Os performed by all groups. | |
3613 | .TP | |
3614 | .B merge | |
007c7be9 | 3615 | Number of merges performed by the I/O scheduler. |
d60e92d1 AC |
3616 | .TP |
3617 | .B ticks | |
3618 | Number of ticks we kept the disk busy. | |
3619 | .TP | |
40943b9a | 3620 | .B in_queue |
d60e92d1 AC |
3621 | Total time spent in the disk queue. |
3622 | .TP | |
3623 | .B util | |
40943b9a TK |
3624 | The disk utilization. A value of 100% means we kept the disk |
3625 | busy constantly, 50% would be a disk idling half of the time. | |
d60e92d1 | 3626 | .RE |
8423bd11 | 3627 | .P |
40943b9a TK |
3628 | It is also possible to get fio to dump the current output while it is running, |
3629 | without terminating the job. To do that, send fio the USR1 signal. You can | |
3630 | also get regularly timed dumps by using the \fB\-\-status\-interval\fR | |
3631 | parameter, or by creating a file in `/tmp' named | |
3632 | `fio\-dump\-status'. If fio sees this file, it will unlink it and dump the | |
3633 | current output status. | |
d60e92d1 | 3634 | .SH TERSE OUTPUT |
40943b9a TK |
3635 | For scripted usage where you typically want to generate tables or graphs of the |
3636 | results, fio can output the results in a semicolon separated format. The format | |
3637 | is one long line of values, such as: | |
d60e92d1 | 3638 | .P |
40943b9a TK |
3639 | .nf |
3640 | 2;card0;0;0;7139336;121836;60004;1;10109;27.932460;116.933948;220;126861;3495.446807;1085.368601;226;126864;3523.635629;1089.012448;24063;99944;50.275485%;59818.274627;5540.657370;7155060;122104;60004;1;8338;29.086342;117.839068;388;128077;5032.488518;1234.785715;391;128085;5061.839412;1236.909129;23436;100928;50.287926%;59964.832030;5644.844189;14.595833%;19.394167%;123706;0;7313;0.1%;0.1%;0.1%;0.1%;0.1%;0.1%;100.0%;0.00%;0.00%;0.00%;0.00%;0.00%;0.00%;0.01%;0.02%;0.05%;0.16%;6.04%;40.40%;52.68%;0.64%;0.01%;0.00%;0.01%;0.00%;0.00%;0.00%;0.00%;0.00% | |
3641 | A description of this job goes here. | |
3642 | .fi | |
d60e92d1 | 3643 | .P |
4e757af1 VF |
3644 | The job description (if provided) follows on a second line for terse v2. |
3645 | It appears on the same line for other terse versions. | |
d60e92d1 | 3646 | .P |
40943b9a TK |
3647 | To enable terse output, use the \fB\-\-minimal\fR or |
3648 | `\-\-output\-format=terse' command line options. The | |
3649 | first value is the version of the terse output format. If the output has to be | |
3650 | changed for some reason, this number will be incremented by 1 to signify that | |
3651 | change. | |
d60e92d1 | 3652 | .P |
40943b9a TK |
3653 | Split up, the format is as follows (comments in brackets denote when a |
3654 | field was introduced or whether it's specific to some terse version): | |
d60e92d1 | 3655 | .P |
40943b9a TK |
3656 | .nf |
3657 | terse version, fio version [v3], jobname, groupid, error | |
3658 | .fi | |
525c2bfa | 3659 | .RS |
40943b9a TK |
3660 | .P |
3661 | .B | |
3662 | READ status: | |
525c2bfa | 3663 | .RE |
40943b9a TK |
3664 | .P |
3665 | .nf | |
3666 | Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec) | |
3667 | Submission latency: min, max, mean, stdev (usec) | |
3668 | Completion latency: min, max, mean, stdev (usec) | |
3669 | Completion latency percentiles: 20 fields (see below) | |
3670 | Total latency: min, max, mean, stdev (usec) | |
3671 | Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5] | |
3672 | IOPS [v5]: min, max, mean, stdev, number of samples | |
3673 | .fi | |
d60e92d1 | 3674 | .RS |
40943b9a TK |
3675 | .P |
3676 | .B | |
3677 | WRITE status: | |
a2c95580 | 3678 | .RE |
40943b9a TK |
3679 | .P |
3680 | .nf | |
3681 | Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec) | |
3682 | Submission latency: min, max, mean, stdev (usec) | |
3683 | Completion latency: min, max, mean, stdev (usec) | |
3684 | Completion latency percentiles: 20 fields (see below) | |
3685 | Total latency: min, max, mean, stdev (usec) | |
3686 | Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5] | |
3687 | IOPS [v5]: min, max, mean, stdev, number of samples | |
3688 | .fi | |
a2c95580 | 3689 | .RS |
40943b9a TK |
3690 | .P |
3691 | .B | |
3692 | TRIM status [all but version 3]: | |
d60e92d1 AC |
3693 | .RE |
3694 | .P | |
40943b9a TK |
3695 | .nf |
3696 | Fields are similar to \fBREAD/WRITE\fR status. | |
3697 | .fi | |
a2c95580 | 3698 | .RS |
a2c95580 | 3699 | .P |
40943b9a | 3700 | .B |
d1429b5c | 3701 | CPU usage: |
d60e92d1 AC |
3702 | .RE |
3703 | .P | |
40943b9a TK |
3704 | .nf |
3705 | user, system, context switches, major faults, minor faults | |
3706 | .fi | |
d60e92d1 | 3707 | .RS |
40943b9a TK |
3708 | .P |
3709 | .B | |
3710 | I/O depths: | |
d60e92d1 AC |
3711 | .RE |
3712 | .P | |
40943b9a TK |
3713 | .nf |
3714 | <=1, 2, 4, 8, 16, 32, >=64 | |
3715 | .fi | |
562c2d2f | 3716 | .RS |
40943b9a TK |
3717 | .P |
3718 | .B | |
3719 | I/O latencies microseconds: | |
562c2d2f | 3720 | .RE |
40943b9a TK |
3721 | .P |
3722 | .nf | |
3723 | <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000 | |
3724 | .fi | |
562c2d2f | 3725 | .RS |
40943b9a TK |
3726 | .P |
3727 | .B | |
3728 | I/O latencies milliseconds: | |
562c2d2f DN |
3729 | .RE |
3730 | .P | |
40943b9a TK |
3731 | .nf |
3732 | <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000 | |
3733 | .fi | |
f2f788dd | 3734 | .RS |
40943b9a TK |
3735 | .P |
3736 | .B | |
3737 | Disk utilization [v3]: | |
f2f788dd JA |
3738 | .RE |
3739 | .P | |
40943b9a TK |
3740 | .nf |
3741 | disk name, read ios, write ios, read merges, write merges, read ticks, write ticks, time spent in queue, disk utilization percentage | |
3742 | .fi | |
562c2d2f | 3743 | .RS |
d60e92d1 | 3744 | .P |
40943b9a TK |
3745 | .B |
3746 | Additional Info (dependent on continue_on_error, default off): | |
d60e92d1 | 3747 | .RE |
2fc26c3d | 3748 | .P |
40943b9a TK |
3749 | .nf |
3750 | total # errors, first error code | |
3751 | .fi | |
2fc26c3d IC |
3752 | .RS |
3753 | .P | |
40943b9a TK |
3754 | .B |
3755 | Additional Info (dependent on description being set): | |
3756 | .RE | |
3757 | .P | |
2fc26c3d | 3758 | .nf |
40943b9a TK |
3759 | Text description |
3760 | .fi | |
3761 | .P | |
3762 | Completion latency percentiles can be a grouping of up to 20 sets, so for the | |
3763 | terse output fio writes all of them. Each field will look like this: | |
3764 | .P | |
3765 | .nf | |
3766 | 1.00%=6112 | |
3767 | .fi | |
3768 | .P | |
3769 | which is the Xth percentile, and the `usec' latency associated with it. | |
3770 | .P | |
3771 | For \fBDisk utilization\fR, all disks used by fio are shown. So for each disk there | |
3772 | will be a disk utilization section. | |
3773 | .P | |
3774 | Below is a single line containing short names for each of the fields in the | |
3775 | minimal output v3, separated by semicolons: | |
3776 | .P | |
3777 | .nf | |
f95689d3 | 3778 | terse_version_3;fio_version;jobname;groupid;error;read_kb;read_bandwidth_kb;read_iops;read_runtime_ms;read_slat_min_us;read_slat_max_us;read_slat_mean_us;read_slat_dev_us;read_clat_min_us;read_clat_max_us;read_clat_mean_us;read_clat_dev_us;read_clat_pct01;read_clat_pct02;read_clat_pct03;read_clat_pct04;read_clat_pct05;read_clat_pct06;read_clat_pct07;read_clat_pct08;read_clat_pct09;read_clat_pct10;read_clat_pct11;read_clat_pct12;read_clat_pct13;read_clat_pct14;read_clat_pct15;read_clat_pct16;read_clat_pct17;read_clat_pct18;read_clat_pct19;read_clat_pct20;read_tlat_min_us;read_lat_max_us;read_lat_mean_us;read_lat_dev_us;read_bw_min_kb;read_bw_max_kb;read_bw_agg_pct;read_bw_mean_kb;read_bw_dev_kb;write_kb;write_bandwidth_kb;write_iops;write_runtime_ms;write_slat_min_us;write_slat_max_us;write_slat_mean_us;write_slat_dev_us;write_clat_min_us;write_clat_max_us;write_clat_mean_us;write_clat_dev_us;write_clat_pct01;write_clat_pct02;write_clat_pct03;write_clat_pct04;write_clat_pct05;write_clat_pct06;write_clat_pct07;write_clat_pct08;write_clat_pct09;write_clat_pct10;write_clat_pct11;write_clat_pct12;write_clat_pct13;write_clat_pct14;write_clat_pct15;write_clat_pct16;write_clat_pct17;write_clat_pct18;write_clat_pct19;write_clat_pct20;write_tlat_min_us;write_lat_max_us;write_lat_mean_us;write_lat_dev_us;write_bw_min_kb;write_bw_max_kb;write_bw_agg_pct;write_bw_mean_kb;write_bw_dev_kb;cpu_user;cpu_sys;cpu_csw;cpu_mjf;cpu_minf;iodepth_1;iodepth_2;iodepth_4;iodepth_8;iodepth_16;iodepth_32;iodepth_64;lat_2us;lat_4us;lat_10us;lat_20us;lat_50us;lat_100us;lat_250us;lat_500us;lat_750us;lat_1000us;lat_2ms;lat_4ms;lat_10ms;lat_20ms;lat_50ms;lat_100ms;lat_250ms;lat_500ms;lat_750ms;lat_1000ms;lat_2000ms;lat_over_2000ms;disk_name;disk_read_iops;disk_write_iops;disk_read_merges;disk_write_merges;disk_read_ticks;write_ticks;disk_queue_time;disk_util |
2fc26c3d | 3779 | .fi |
4e757af1 VF |
3780 | .P |
3781 | In client/server mode terse output differs from what appears when jobs are run | |
3782 | locally. Disk utilization data is omitted from the standard terse output and | |
3783 | for v3 and later appears on its own separate line at the end of each terse | |
3784 | reporting cycle. | |
44c82dba VF |
3785 | .SH JSON OUTPUT |
3786 | The \fBjson\fR output format is intended to be both human readable and convenient | |
3787 | for automated parsing. For the most part its sections mirror those of the | |
3788 | \fBnormal\fR output. The \fBruntime\fR value is reported in msec and the \fBbw\fR value is | |
3789 | reported in 1024 bytes per second units. | |
3790 | .fi | |
d9e557ab VF |
3791 | .SH JSON+ OUTPUT |
3792 | The \fBjson+\fR output format is identical to the \fBjson\fR output format except that it | |
3793 | adds a full dump of the completion latency bins. Each \fBbins\fR object contains a | |
3794 | set of (key, value) pairs where keys are latency durations and values count how | |
3795 | many I/Os had completion latencies of the corresponding duration. For example, | |
3796 | consider: | |
d9e557ab | 3797 | .RS |
40943b9a | 3798 | .P |
d9e557ab VF |
3799 | "bins" : { "87552" : 1, "89600" : 1, "94720" : 1, "96768" : 1, "97792" : 1, "99840" : 1, "100864" : 2, "103936" : 6, "104960" : 534, "105984" : 5995, "107008" : 7529, ... } |
3800 | .RE | |
40943b9a | 3801 | .P |
d9e557ab VF |
3802 | This data indicates that one I/O required 87,552ns to complete, two I/Os required |
3803 | 100,864ns to complete, and 7529 I/Os required 107,008ns to complete. | |
40943b9a | 3804 | .P |
d9e557ab | 3805 | Also included with fio is a Python script \fBfio_jsonplus_clat2csv\fR that takes |
338f2db5 | 3806 | json+ output and generates CSV-formatted latency data suitable for plotting. |
40943b9a | 3807 | .P |
d9e557ab | 3808 | The latency durations actually represent the midpoints of latency intervals. |
40943b9a | 3809 | For details refer to `stat.h' in the fio source. |
29dbd1e5 | 3810 | .SH TRACE FILE FORMAT |
40943b9a TK |
3811 | There are two trace file format that you can encounter. The older (v1) format is |
3812 | unsupported since version 1.20\-rc3 (March 2008). It will still be described | |
29dbd1e5 | 3813 | below in case that you get an old trace and want to understand it. |
29dbd1e5 | 3814 | .P |
40943b9a TK |
3815 | In any case the trace is a simple text file with a single action per line. |
3816 | .TP | |
29dbd1e5 | 3817 | .B Trace file format v1 |
40943b9a | 3818 | Each line represents a single I/O action in the following format: |
29dbd1e5 | 3819 | .RS |
40943b9a TK |
3820 | .RS |
3821 | .P | |
29dbd1e5 | 3822 | rw, offset, length |
29dbd1e5 JA |
3823 | .RE |
3824 | .P | |
40943b9a TK |
3825 | where `rw=0/1' for read/write, and the `offset' and `length' entries being in bytes. |
3826 | .P | |
3827 | This format is not supported in fio versions >= 1.20\-rc3. | |
3828 | .RE | |
3829 | .TP | |
29dbd1e5 | 3830 | .B Trace file format v2 |
40943b9a TK |
3831 | The second version of the trace file format was added in fio version 1.17. It |
3832 | allows to access more then one file per trace and has a bigger set of possible | |
3833 | file actions. | |
29dbd1e5 | 3834 | .RS |
40943b9a | 3835 | .P |
29dbd1e5 | 3836 | The first line of the trace file has to be: |
40943b9a TK |
3837 | .RS |
3838 | .P | |
3839 | "fio version 2 iolog" | |
3840 | .RE | |
3841 | .P | |
29dbd1e5 | 3842 | Following this can be lines in two different formats, which are described below. |
40943b9a TK |
3843 | .P |
3844 | .B | |
29dbd1e5 | 3845 | The file management format: |
40943b9a TK |
3846 | .RS |
3847 | filename action | |
29dbd1e5 | 3848 | .P |
40943b9a | 3849 | The `filename' is given as an absolute path. The `action' can be one of these: |
29dbd1e5 JA |
3850 | .RS |
3851 | .TP | |
3852 | .B add | |
40943b9a | 3853 | Add the given `filename' to the trace. |
29dbd1e5 JA |
3854 | .TP |
3855 | .B open | |
40943b9a TK |
3856 | Open the file with the given `filename'. The `filename' has to have |
3857 | been added with the \fBadd\fR action before. | |
29dbd1e5 JA |
3858 | .TP |
3859 | .B close | |
40943b9a TK |
3860 | Close the file with the given `filename'. The file has to have been |
3861 | \fBopen\fRed before. | |
3862 | .RE | |
29dbd1e5 | 3863 | .RE |
29dbd1e5 | 3864 | .P |
40943b9a TK |
3865 | .B |
3866 | The file I/O action format: | |
3867 | .RS | |
3868 | filename action offset length | |
29dbd1e5 | 3869 | .P |
40943b9a TK |
3870 | The `filename' is given as an absolute path, and has to have been \fBadd\fRed and |
3871 | \fBopen\fRed before it can be used with this format. The `offset' and `length' are | |
3872 | given in bytes. The `action' can be one of these: | |
29dbd1e5 JA |
3873 | .RS |
3874 | .TP | |
3875 | .B wait | |
40943b9a TK |
3876 | Wait for `offset' microseconds. Everything below 100 is discarded. |
3877 | The time is relative to the previous `wait' statement. | |
29dbd1e5 JA |
3878 | .TP |
3879 | .B read | |
40943b9a | 3880 | Read `length' bytes beginning from `offset'. |
29dbd1e5 JA |
3881 | .TP |
3882 | .B write | |
40943b9a | 3883 | Write `length' bytes beginning from `offset'. |
29dbd1e5 JA |
3884 | .TP |
3885 | .B sync | |
40943b9a | 3886 | \fBfsync\fR\|(2) the file. |
29dbd1e5 JA |
3887 | .TP |
3888 | .B datasync | |
40943b9a | 3889 | \fBfdatasync\fR\|(2) the file. |
29dbd1e5 JA |
3890 | .TP |
3891 | .B trim | |
40943b9a TK |
3892 | Trim the given file from the given `offset' for `length' bytes. |
3893 | .RE | |
29dbd1e5 | 3894 | .RE |
b9921d1a DZ |
3895 | .SH I/O REPLAY \- MERGING TRACES |
3896 | Colocation is a common practice used to get the most out of a machine. | |
3897 | Knowing which workloads play nicely with each other and which ones don't is | |
3898 | a much harder task. While fio can replay workloads concurrently via multiple | |
3899 | jobs, it leaves some variability up to the scheduler making results harder to | |
3900 | reproduce. Merging is a way to make the order of events consistent. | |
3901 | .P | |
3902 | Merging is integrated into I/O replay and done when a \fBmerge_blktrace_file\fR | |
3903 | is specified. The list of files passed to \fBread_iolog\fR go through the merge | |
3904 | process and output a single file stored to the specified file. The output file is | |
3905 | passed on as if it were the only file passed to \fBread_iolog\fR. An example would | |
3906 | look like: | |
3907 | .RS | |
3908 | .P | |
3909 | $ fio \-\-read_iolog="<file1>:<file2>" \-\-merge_blktrace_file="<output_file>" | |
3910 | .RE | |
3911 | .P | |
3912 | Creating only the merged file can be done by passing the command line argument | |
3913 | \fBmerge-blktrace-only\fR. | |
87a48ada DZ |
3914 | .P |
3915 | Scaling traces can be done to see the relative impact of any particular trace | |
3916 | being slowed down or sped up. \fBmerge_blktrace_scalars\fR takes in a colon | |
3917 | separated list of percentage scalars. It is index paired with the files passed | |
3918 | to \fBread_iolog\fR. | |
55bfd8c8 DZ |
3919 | .P |
3920 | With scaling, it may be desirable to match the running time of all traces. | |
3921 | This can be done with \fBmerge_blktrace_iters\fR. It is index paired with | |
3922 | \fBread_iolog\fR just like \fBmerge_blktrace_scalars\fR. | |
3923 | .P | |
3924 | In an example, given two traces, A and B, each 60s long. If we want to see | |
3925 | the impact of trace A issuing IOs twice as fast and repeat trace A over the | |
3926 | runtime of trace B, the following can be done: | |
3927 | .RS | |
3928 | .P | |
3929 | $ fio \-\-read_iolog="<trace_a>:"<trace_b>" \-\-merge_blktrace_file"<output_file>" \-\-merge_blktrace_scalars="50:100" \-\-merge_blktrace_iters="2:1" | |
3930 | .RE | |
3931 | .P | |
3932 | This runs trace A at 2x the speed twice for approximately the same runtime as | |
3933 | a single run of trace B. | |
29dbd1e5 | 3934 | .SH CPU IDLENESS PROFILING |
40943b9a TK |
3935 | In some cases, we want to understand CPU overhead in a test. For example, we |
3936 | test patches for the specific goodness of whether they reduce CPU usage. | |
3937 | Fio implements a balloon approach to create a thread per CPU that runs at idle | |
3938 | priority, meaning that it only runs when nobody else needs the cpu. | |
3939 | By measuring the amount of work completed by the thread, idleness of each CPU | |
3940 | can be derived accordingly. | |
3941 | .P | |
3942 | An unit work is defined as touching a full page of unsigned characters. Mean and | |
3943 | standard deviation of time to complete an unit work is reported in "unit work" | |
3944 | section. Options can be chosen to report detailed percpu idleness or overall | |
3945 | system idleness by aggregating percpu stats. | |
29dbd1e5 | 3946 | .SH VERIFICATION AND TRIGGERS |
40943b9a TK |
3947 | Fio is usually run in one of two ways, when data verification is done. The first |
3948 | is a normal write job of some sort with verify enabled. When the write phase has | |
3949 | completed, fio switches to reads and verifies everything it wrote. The second | |
3950 | model is running just the write phase, and then later on running the same job | |
3951 | (but with reads instead of writes) to repeat the same I/O patterns and verify | |
3952 | the contents. Both of these methods depend on the write phase being completed, | |
3953 | as fio otherwise has no idea how much data was written. | |
3954 | .P | |
3955 | With verification triggers, fio supports dumping the current write state to | |
3956 | local files. Then a subsequent read verify workload can load this state and know | |
3957 | exactly where to stop. This is useful for testing cases where power is cut to a | |
3958 | server in a managed fashion, for instance. | |
3959 | .P | |
29dbd1e5 | 3960 | A verification trigger consists of two things: |
29dbd1e5 | 3961 | .RS |
40943b9a TK |
3962 | .P |
3963 | 1) Storing the write state of each job. | |
3964 | .P | |
3965 | 2) Executing a trigger command. | |
29dbd1e5 | 3966 | .RE |
40943b9a TK |
3967 | .P |
3968 | The write state is relatively small, on the order of hundreds of bytes to single | |
3969 | kilobytes. It contains information on the number of completions done, the last X | |
3970 | completions, etc. | |
3971 | .P | |
3972 | A trigger is invoked either through creation ('touch') of a specified file in | |
3973 | the system, or through a timeout setting. If fio is run with | |
3974 | `\-\-trigger\-file=/tmp/trigger\-file', then it will continually | |
3975 | check for the existence of `/tmp/trigger\-file'. When it sees this file, it | |
3976 | will fire off the trigger (thus saving state, and executing the trigger | |
29dbd1e5 | 3977 | command). |
40943b9a TK |
3978 | .P |
3979 | For client/server runs, there's both a local and remote trigger. If fio is | |
3980 | running as a server backend, it will send the job states back to the client for | |
3981 | safe storage, then execute the remote trigger, if specified. If a local trigger | |
3982 | is specified, the server will still send back the write state, but the client | |
3983 | will then execute the trigger. | |
29dbd1e5 JA |
3984 | .RE |
3985 | .P | |
3986 | .B Verification trigger example | |
3987 | .RS | |
40943b9a TK |
3988 | Let's say we want to run a powercut test on the remote Linux machine 'server'. |
3989 | Our write workload is in `write\-test.fio'. We want to cut power to 'server' at | |
3990 | some point during the run, and we'll run this test from the safety or our local | |
3991 | machine, 'localbox'. On the server, we'll start the fio backend normally: | |
3992 | .RS | |
3993 | .P | |
3994 | server# fio \-\-server | |
3995 | .RE | |
3996 | .P | |
29dbd1e5 | 3997 | and on the client, we'll fire off the workload: |
40943b9a TK |
3998 | .RS |
3999 | .P | |
4000 | localbox$ fio \-\-client=server \-\-trigger\-file=/tmp/my\-trigger \-\-trigger\-remote="bash \-c "echo b > /proc/sysrq\-triger"" | |
4001 | .RE | |
4002 | .P | |
4003 | We set `/tmp/my\-trigger' as the trigger file, and we tell fio to execute: | |
4004 | .RS | |
4005 | .P | |
4006 | echo b > /proc/sysrq\-trigger | |
4007 | .RE | |
4008 | .P | |
4009 | on the server once it has received the trigger and sent us the write state. This | |
4010 | will work, but it's not really cutting power to the server, it's merely | |
4011 | abruptly rebooting it. If we have a remote way of cutting power to the server | |
4012 | through IPMI or similar, we could do that through a local trigger command | |
4013 | instead. Let's assume we have a script that does IPMI reboot of a given hostname, | |
4014 | ipmi\-reboot. On localbox, we could then have run fio with a local trigger | |
4015 | instead: | |
4016 | .RS | |
4017 | .P | |
4018 | localbox$ fio \-\-client=server \-\-trigger\-file=/tmp/my\-trigger \-\-trigger="ipmi\-reboot server" | |
4019 | .RE | |
4020 | .P | |
4021 | For this case, fio would wait for the server to send us the write state, then | |
4022 | execute `ipmi\-reboot server' when that happened. | |
29dbd1e5 JA |
4023 | .RE |
4024 | .P | |
4025 | .B Loading verify state | |
4026 | .RS | |
40943b9a TK |
4027 | To load stored write state, a read verification job file must contain the |
4028 | \fBverify_state_load\fR option. If that is set, fio will load the previously | |
29dbd1e5 | 4029 | stored state. For a local fio run this is done by loading the files directly, |
40943b9a TK |
4030 | and on a client/server run, the server backend will ask the client to send the |
4031 | files over and load them from there. | |
29dbd1e5 | 4032 | .RE |
a3ae5b05 | 4033 | .SH LOG FILE FORMATS |
a3ae5b05 JA |
4034 | Fio supports a variety of log file formats, for logging latencies, bandwidth, |
4035 | and IOPS. The logs share a common format, which looks like this: | |
40943b9a | 4036 | .RS |
a3ae5b05 | 4037 | .P |
1a953d97 PC |
4038 | time (msec), value, data direction, block size (bytes), offset (bytes), |
4039 | command priority | |
40943b9a TK |
4040 | .RE |
4041 | .P | |
4042 | `Time' for the log entry is always in milliseconds. The `value' logged depends | |
4043 | on the type of log, it will be one of the following: | |
4044 | .RS | |
a3ae5b05 JA |
4045 | .TP |
4046 | .B Latency log | |
168bb587 | 4047 | Value is latency in nsecs |
a3ae5b05 JA |
4048 | .TP |
4049 | .B Bandwidth log | |
6d500c2e | 4050 | Value is in KiB/sec |
a3ae5b05 JA |
4051 | .TP |
4052 | .B IOPS log | |
40943b9a TK |
4053 | Value is IOPS |
4054 | .RE | |
a3ae5b05 | 4055 | .P |
40943b9a TK |
4056 | `Data direction' is one of the following: |
4057 | .RS | |
a3ae5b05 JA |
4058 | .TP |
4059 | .B 0 | |
40943b9a | 4060 | I/O is a READ |
a3ae5b05 JA |
4061 | .TP |
4062 | .B 1 | |
40943b9a | 4063 | I/O is a WRITE |
a3ae5b05 JA |
4064 | .TP |
4065 | .B 2 | |
40943b9a | 4066 | I/O is a TRIM |
a3ae5b05 | 4067 | .RE |
40943b9a | 4068 | .P |
15417073 SW |
4069 | The entry's `block size' is always in bytes. The `offset' is the position in bytes |
4070 | from the start of the file for that particular I/O. The logging of the offset can be | |
40943b9a TK |
4071 | toggled with \fBlog_offset\fR. |
4072 | .P | |
1a953d97 PC |
4073 | `Command priority` is 0 for normal priority and 1 for high priority. This is controlled |
4074 | by the ioengine specific \fBcmdprio_percentage\fR. | |
4075 | .P | |
15417073 SW |
4076 | Fio defaults to logging every individual I/O but when windowed logging is set |
4077 | through \fBlog_avg_msec\fR, either the average (by default) or the maximum | |
4078 | (\fBlog_max_value\fR is set) `value' seen over the specified period of time | |
4079 | is recorded. Each `data direction' seen within the window period will aggregate | |
4080 | its values in a separate row. Further, when using windowed logging the `block | |
4081 | size' and `offset' entries will always contain 0. | |
49da1240 | 4082 | .SH CLIENT / SERVER |
338f2db5 | 4083 | Normally fio is invoked as a stand-alone application on the machine where the |
40943b9a TK |
4084 | I/O workload should be generated. However, the backend and frontend of fio can |
4085 | be run separately i.e., the fio server can generate an I/O workload on the "Device | |
4086 | Under Test" while being controlled by a client on another machine. | |
4087 | .P | |
4088 | Start the server on the machine which has access to the storage DUT: | |
4089 | .RS | |
4090 | .P | |
4091 | $ fio \-\-server=args | |
4092 | .RE | |
4093 | .P | |
4094 | where `args' defines what fio listens to. The arguments are of the form | |
4095 | `type,hostname' or `IP,port'. `type' is either `ip' (or ip4) for TCP/IP | |
4096 | v4, `ip6' for TCP/IP v6, or `sock' for a local unix domain socket. | |
4097 | `hostname' is either a hostname or IP address, and `port' is the port to listen | |
4098 | to (only valid for TCP/IP, not a local socket). Some examples: | |
4099 | .RS | |
4100 | .TP | |
e0ee7a8b | 4101 | 1) \fBfio \-\-server\fR |
40943b9a TK |
4102 | Start a fio server, listening on all interfaces on the default port (8765). |
4103 | .TP | |
e0ee7a8b | 4104 | 2) \fBfio \-\-server=ip:hostname,4444\fR |
40943b9a TK |
4105 | Start a fio server, listening on IP belonging to hostname and on port 4444. |
4106 | .TP | |
e0ee7a8b | 4107 | 3) \fBfio \-\-server=ip6:::1,4444\fR |
40943b9a TK |
4108 | Start a fio server, listening on IPv6 localhost ::1 and on port 4444. |
4109 | .TP | |
e0ee7a8b | 4110 | 4) \fBfio \-\-server=,4444\fR |
40943b9a TK |
4111 | Start a fio server, listening on all interfaces on port 4444. |
4112 | .TP | |
e0ee7a8b | 4113 | 5) \fBfio \-\-server=1.2.3.4\fR |
40943b9a TK |
4114 | Start a fio server, listening on IP 1.2.3.4 on the default port. |
4115 | .TP | |
e0ee7a8b | 4116 | 6) \fBfio \-\-server=sock:/tmp/fio.sock\fR |
40943b9a TK |
4117 | Start a fio server, listening on the local socket `/tmp/fio.sock'. |
4118 | .RE | |
4119 | .P | |
4120 | Once a server is running, a "client" can connect to the fio server with: | |
4121 | .RS | |
4122 | .P | |
4123 | $ fio <local\-args> \-\-client=<server> <remote\-args> <job file(s)> | |
4124 | .RE | |
4125 | .P | |
4126 | where `local\-args' are arguments for the client where it is running, `server' | |
4127 | is the connect string, and `remote\-args' and `job file(s)' are sent to the | |
4128 | server. The `server' string follows the same format as it does on the server | |
4129 | side, to allow IP/hostname/socket and port strings. | |
4130 | .P | |
4131 | Fio can connect to multiple servers this way: | |
4132 | .RS | |
4133 | .P | |
4134 | $ fio \-\-client=<server1> <job file(s)> \-\-client=<server2> <job file(s)> | |
4135 | .RE | |
4136 | .P | |
4137 | If the job file is located on the fio server, then you can tell the server to | |
4138 | load a local file as well. This is done by using \fB\-\-remote\-config\fR: | |
4139 | .RS | |
4140 | .P | |
4141 | $ fio \-\-client=server \-\-remote\-config /path/to/file.fio | |
4142 | .RE | |
4143 | .P | |
4144 | Then fio will open this local (to the server) job file instead of being passed | |
4145 | one from the client. | |
4146 | .P | |
ff6bb260 | 4147 | If you have many servers (example: 100 VMs/containers), you can input a pathname |
40943b9a TK |
4148 | of a file containing host IPs/names as the parameter value for the |
4149 | \fB\-\-client\fR option. For example, here is an example `host.list' | |
4150 | file containing 2 hostnames: | |
4151 | .RS | |
4152 | .P | |
4153 | .PD 0 | |
39b5f61e | 4154 | host1.your.dns.domain |
40943b9a | 4155 | .P |
39b5f61e | 4156 | host2.your.dns.domain |
40943b9a TK |
4157 | .PD |
4158 | .RE | |
4159 | .P | |
39b5f61e | 4160 | The fio command would then be: |
40943b9a TK |
4161 | .RS |
4162 | .P | |
4163 | $ fio \-\-client=host.list <job file(s)> | |
4164 | .RE | |
4165 | .P | |
338f2db5 | 4166 | In this mode, you cannot input server-specific parameters or job files \-\- all |
39b5f61e | 4167 | servers receive the same job file. |
40943b9a TK |
4168 | .P |
4169 | In order to let `fio \-\-client' runs use a shared filesystem from multiple | |
4170 | hosts, `fio \-\-client' now prepends the IP address of the server to the | |
4171 | filename. For example, if fio is using the directory `/mnt/nfs/fio' and is | |
4172 | writing filename `fileio.tmp', with a \fB\-\-client\fR `hostfile' | |
4173 | containing two hostnames `h1' and `h2' with IP addresses 192.168.10.120 and | |
4174 | 192.168.10.121, then fio will create two files: | |
4175 | .RS | |
4176 | .P | |
4177 | .PD 0 | |
39b5f61e | 4178 | /mnt/nfs/fio/192.168.10.120.fileio.tmp |
40943b9a | 4179 | .P |
39b5f61e | 4180 | /mnt/nfs/fio/192.168.10.121.fileio.tmp |
40943b9a TK |
4181 | .PD |
4182 | .RE | |
4e757af1 VF |
4183 | .P |
4184 | Terse output in client/server mode will differ slightly from what is produced | |
4185 | when fio is run in stand-alone mode. See the terse output section for details. | |
d60e92d1 AC |
4186 | .SH AUTHORS |
4187 | .B fio | |
d292596c | 4188 | was written by Jens Axboe <axboe@kernel.dk>. |
d1429b5c AC |
4189 | .br |
4190 | This man page was written by Aaron Carroll <aaronc@cse.unsw.edu.au> based | |
d60e92d1 | 4191 | on documentation by Jens Axboe. |
40943b9a TK |
4192 | .br |
4193 | This man page was rewritten by Tomohiro Kusumi <tkusumi@tuxera.com> based | |
4194 | on documentation by Jens Axboe. | |
d60e92d1 | 4195 | .SH "REPORTING BUGS" |
482900c9 | 4196 | Report bugs to the \fBfio\fR mailing list <fio@vger.kernel.org>. |
6468020d | 4197 | .br |
40943b9a TK |
4198 | See \fBREPORTING\-BUGS\fR. |
4199 | .P | |
4200 | \fBREPORTING\-BUGS\fR: \fIhttp://git.kernel.dk/cgit/fio/plain/REPORTING\-BUGS\fR | |
d60e92d1 | 4201 | .SH "SEE ALSO" |
d1429b5c AC |
4202 | For further documentation see \fBHOWTO\fR and \fBREADME\fR. |
4203 | .br | |
40943b9a | 4204 | Sample jobfiles are available in the `examples/' directory. |
9040e236 | 4205 | .br |
40943b9a TK |
4206 | These are typically located under `/usr/share/doc/fio'. |
4207 | .P | |
4208 | \fBHOWTO\fR: \fIhttp://git.kernel.dk/cgit/fio/plain/HOWTO\fR | |
9040e236 | 4209 | .br |
40943b9a | 4210 | \fBREADME\fR: \fIhttp://git.kernel.dk/cgit/fio/plain/README\fR |