| 1 | .TH fio 1 "August 2017" "User Manual" |
| 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 |
| 15 | .BI \-\-debug \fR=\fPtype |
| 16 | Enable verbose tracing \fItype\fR of various fio actions. May be `all' for all \fItype\fRs |
| 17 | or individual types separated by a comma (e.g. `\-\-debug=file,mem' will enable |
| 18 | file and memory debugging). `help' will list all available tracing options. |
| 19 | .TP |
| 20 | .BI \-\-parse\-only |
| 21 | Parse options only, don't start any I/O. |
| 22 | .TP |
| 23 | .BI \-\-output \fR=\fPfilename |
| 24 | Write output to \fIfilename\fR. |
| 25 | .TP |
| 26 | .BI \-\-output\-format \fR=\fPformat |
| 27 | Set the reporting \fIformat\fR to `normal', `terse', `json', or |
| 28 | `json+'. Multiple formats can be selected, separate by a comma. `terse' |
| 29 | is a CSV based format. `json+' is like `json', except it adds a full |
| 30 | dump of the latency buckets. |
| 31 | .TP |
| 32 | .BI \-\-bandwidth\-log |
| 33 | Generate aggregate bandwidth logs. |
| 34 | .TP |
| 35 | .BI \-\-minimal |
| 36 | Print statistics in a terse, semicolon\-delimited format. |
| 37 | .TP |
| 38 | .BI \-\-append\-terse |
| 39 | Print statistics in selected mode AND terse, semicolon\-delimited format. |
| 40 | \fBDeprecated\fR, use \fB\-\-output\-format\fR instead to select multiple formats. |
| 41 | .TP |
| 42 | .BI \-\-terse\-version \fR=\fPversion |
| 43 | Set terse \fIversion\fR output format (default `3', or `2', `4', `5'). |
| 44 | .TP |
| 45 | .BI \-\-version |
| 46 | Print version information and exit. |
| 47 | .TP |
| 48 | .BI \-\-help |
| 49 | Print a summary of the command line options and exit. |
| 50 | .TP |
| 51 | .BI \-\-cpuclock\-test |
| 52 | Perform test and validation of internal CPU clock. |
| 53 | .TP |
| 54 | .BI \-\-crctest \fR=\fP[test] |
| 55 | Test the speed of the built\-in checksumming functions. If no argument is given, |
| 56 | all of them are tested. Alternatively, a comma separated list can be passed, in which |
| 57 | case the given ones are tested. |
| 58 | .TP |
| 59 | .BI \-\-cmdhelp \fR=\fPcommand |
| 60 | Print help information for \fIcommand\fR. May be `all' for all commands. |
| 61 | .TP |
| 62 | .BI \-\-enghelp \fR=\fP[ioengine[,command]] |
| 63 | List all commands defined by \fIioengine\fR, or print help for \fIcommand\fR |
| 64 | defined by \fIioengine\fR. If no \fIioengine\fR is given, list all |
| 65 | available ioengines. |
| 66 | .TP |
| 67 | .BI \-\-showcmd \fR=\fPjobfile |
| 68 | Convert \fIjobfile\fR to a set of command\-line options. |
| 69 | .TP |
| 70 | .BI \-\-readonly |
| 71 | Turn on safety read\-only checks, preventing writes. The \fB\-\-readonly\fR |
| 72 | option is an extra safety guard to prevent users from accidentally starting |
| 73 | a write workload when that is not desired. Fio will only write if |
| 74 | `rw=write/randwrite/rw/randrw' is given. This extra safety net can be used |
| 75 | as an extra precaution as \fB\-\-readonly\fR will also enable a write check in |
| 76 | the I/O engine core to prevent writes due to unknown user space bug(s). |
| 77 | .TP |
| 78 | .BI \-\-eta \fR=\fPwhen |
| 79 | Specifies when real\-time ETA estimate should be printed. \fIwhen\fR may |
| 80 | be `always', `never' or `auto'. |
| 81 | .TP |
| 82 | .BI \-\-eta\-newline \fR=\fPtime |
| 83 | Force a new line for every \fItime\fR period passed. When the unit is omitted, |
| 84 | the value is interpreted in seconds. |
| 85 | .TP |
| 86 | .BI \-\-status\-interval \fR=\fPtime |
| 87 | Force a full status dump of cumulative (from job start) values at \fItime\fR |
| 88 | intervals. This option does *not* provide per-period measurements. So |
| 89 | values such as bandwidth are running averages. When the time unit is omitted, |
| 90 | \fItime\fR is interpreted in seconds. |
| 91 | .TP |
| 92 | .BI \-\-section \fR=\fPname |
| 93 | Only run specified section \fIname\fR in job file. Multiple sections can be specified. |
| 94 | The \fB\-\-section\fR option allows one to combine related jobs into one file. |
| 95 | E.g. one job file could define light, moderate, and heavy sections. Tell |
| 96 | fio to run only the "heavy" section by giving `\-\-section=heavy' |
| 97 | command line option. One can also specify the "write" operations in one |
| 98 | section and "verify" operation in another section. The \fB\-\-section\fR option |
| 99 | only applies to job sections. The reserved *global* section is always |
| 100 | parsed and used. |
| 101 | .TP |
| 102 | .BI \-\-alloc\-size \fR=\fPkb |
| 103 | Set the internal smalloc pool size to \fIkb\fR in KiB. The |
| 104 | \fB\-\-alloc\-size\fR switch allows one to use a larger pool size for smalloc. |
| 105 | If running large jobs with randommap enabled, fio can run out of memory. |
| 106 | Smalloc is an internal allocator for shared structures from a fixed size |
| 107 | memory pool and can grow to 16 pools. The pool size defaults to 16MiB. |
| 108 | NOTE: While running `.fio_smalloc.*' backing store files are visible |
| 109 | in `/tmp'. |
| 110 | .TP |
| 111 | .BI \-\-warnings\-fatal |
| 112 | All fio parser warnings are fatal, causing fio to exit with an error. |
| 113 | .TP |
| 114 | .BI \-\-max\-jobs \fR=\fPnr |
| 115 | Set the maximum number of threads/processes to support to \fInr\fR. |
| 116 | NOTE: On Linux, it may be necessary to increase the shared-memory limit |
| 117 | (`/proc/sys/kernel/shmmax') if fio runs into errors while creating jobs. |
| 118 | .TP |
| 119 | .BI \-\-server \fR=\fPargs |
| 120 | Start a backend server, with \fIargs\fR specifying what to listen to. |
| 121 | See \fBCLIENT/SERVER\fR section. |
| 122 | .TP |
| 123 | .BI \-\-daemonize \fR=\fPpidfile |
| 124 | Background a fio server, writing the pid to the given \fIpidfile\fR file. |
| 125 | .TP |
| 126 | .BI \-\-client \fR=\fPhostname |
| 127 | Instead of running the jobs locally, send and run them on the given \fIhostname\fR |
| 128 | or set of \fIhostname\fRs. See \fBCLIENT/SERVER\fR section. |
| 129 | .TP |
| 130 | .BI \-\-remote\-config \fR=\fPfile |
| 131 | Tell fio server to load this local \fIfile\fR. |
| 132 | .TP |
| 133 | .BI \-\-idle\-prof \fR=\fPoption |
| 134 | Report CPU idleness. \fIoption\fR is one of the following: |
| 135 | .RS |
| 136 | .RS |
| 137 | .TP |
| 138 | .B calibrate |
| 139 | Run unit work calibration only and exit. |
| 140 | .TP |
| 141 | .B system |
| 142 | Show aggregate system idleness and unit work. |
| 143 | .TP |
| 144 | .B percpu |
| 145 | As \fBsystem\fR but also show per CPU idleness. |
| 146 | .RE |
| 147 | .RE |
| 148 | .TP |
| 149 | .BI \-\-inflate\-log \fR=\fPlog |
| 150 | Inflate and output compressed \fIlog\fR. |
| 151 | .TP |
| 152 | .BI \-\-trigger\-file \fR=\fPfile |
| 153 | Execute trigger command when \fIfile\fR exists. |
| 154 | .TP |
| 155 | .BI \-\-trigger\-timeout \fR=\fPtime |
| 156 | Execute trigger at this \fItime\fR. |
| 157 | .TP |
| 158 | .BI \-\-trigger \fR=\fPcommand |
| 159 | Set this \fIcommand\fR as local trigger. |
| 160 | .TP |
| 161 | .BI \-\-trigger\-remote \fR=\fPcommand |
| 162 | Set this \fIcommand\fR as remote trigger. |
| 163 | .TP |
| 164 | .BI \-\-aux\-path \fR=\fPpath |
| 165 | Use this \fIpath\fR for fio state generated files. |
| 166 | .SH "JOB FILE FORMAT" |
| 167 | Any parameters following the options will be assumed to be job files, unless |
| 168 | they match a job file parameter. Multiple job files can be listed and each job |
| 169 | file will be regarded as a separate group. Fio will \fBstonewall\fR execution |
| 170 | between each group. |
| 171 | |
| 172 | Fio accepts one or more job files describing what it is |
| 173 | supposed to do. The job file format is the classic ini file, where the names |
| 174 | enclosed in [] brackets define the job name. You are free to use any ASCII name |
| 175 | you want, except *global* which has special meaning. Following the job name is |
| 176 | a sequence of zero or more parameters, one per line, that define the behavior of |
| 177 | the job. If the first character in a line is a ';' or a '#', the entire line is |
| 178 | discarded as a comment. |
| 179 | |
| 180 | A *global* section sets defaults for the jobs described in that file. A job may |
| 181 | override a *global* section parameter, and a job file may even have several |
| 182 | *global* sections if so desired. A job is only affected by a *global* section |
| 183 | residing above it. |
| 184 | |
| 185 | The \fB\-\-cmdhelp\fR option also lists all options. If used with an \fIcommand\fR |
| 186 | argument, \fB\-\-cmdhelp\fR will detail the given \fIcommand\fR. |
| 187 | |
| 188 | See the `examples/' directory for inspiration on how to write job files. Note |
| 189 | the copyright and license requirements currently apply to |
| 190 | `examples/' files. |
| 191 | .SH "JOB FILE PARAMETERS" |
| 192 | Some parameters take an option of a given type, such as an integer or a |
| 193 | string. Anywhere a numeric value is required, an arithmetic expression may be |
| 194 | used, provided it is surrounded by parentheses. Supported operators are: |
| 195 | .RS |
| 196 | .P |
| 197 | .B addition (+) |
| 198 | .P |
| 199 | .B subtraction (\-) |
| 200 | .P |
| 201 | .B multiplication (*) |
| 202 | .P |
| 203 | .B division (/) |
| 204 | .P |
| 205 | .B modulus (%) |
| 206 | .P |
| 207 | .B exponentiation (^) |
| 208 | .RE |
| 209 | .P |
| 210 | For time values in expressions, units are microseconds by default. This is |
| 211 | different than for time values not in expressions (not enclosed in |
| 212 | parentheses). |
| 213 | .SH "PARAMETER TYPES" |
| 214 | The following parameter types are used. |
| 215 | .TP |
| 216 | .I str |
| 217 | String. A sequence of alphanumeric characters. |
| 218 | .TP |
| 219 | .I time |
| 220 | Integer with possible time suffix. Without a unit value is interpreted as |
| 221 | seconds unless otherwise specified. Accepts a suffix of 'd' for days, 'h' for |
| 222 | hours, 'm' for minutes, 's' for seconds, 'ms' (or 'msec') for milliseconds and 'us' |
| 223 | (or 'usec') for microseconds. For example, use 10m for 10 minutes. |
| 224 | .TP |
| 225 | .I int |
| 226 | Integer. A whole number value, which may contain an integer prefix |
| 227 | and an integer suffix. |
| 228 | .RS |
| 229 | .RS |
| 230 | .P |
| 231 | [*integer prefix*] **number** [*integer suffix*] |
| 232 | .RE |
| 233 | .P |
| 234 | The optional *integer prefix* specifies the number's base. The default |
| 235 | is decimal. *0x* specifies hexadecimal. |
| 236 | .P |
| 237 | The optional *integer suffix* specifies the number's units, and includes an |
| 238 | optional unit prefix and an optional unit. For quantities of data, the |
| 239 | default unit is bytes. For quantities of time, the default unit is seconds |
| 240 | unless otherwise specified. |
| 241 | .P |
| 242 | With `kb_base=1000', fio follows international standards for unit |
| 243 | prefixes. To specify power\-of\-10 decimal values defined in the |
| 244 | International System of Units (SI): |
| 245 | .RS |
| 246 | .P |
| 247 | .PD 0 |
| 248 | K means kilo (K) or 1000 |
| 249 | .P |
| 250 | M means mega (M) or 1000**2 |
| 251 | .P |
| 252 | G means giga (G) or 1000**3 |
| 253 | .P |
| 254 | T means tera (T) or 1000**4 |
| 255 | .P |
| 256 | P means peta (P) or 1000**5 |
| 257 | .PD |
| 258 | .RE |
| 259 | .P |
| 260 | To specify power\-of\-2 binary values defined in IEC 80000\-13: |
| 261 | .RS |
| 262 | .P |
| 263 | .PD 0 |
| 264 | Ki means kibi (Ki) or 1024 |
| 265 | .P |
| 266 | Mi means mebi (Mi) or 1024**2 |
| 267 | .P |
| 268 | Gi means gibi (Gi) or 1024**3 |
| 269 | .P |
| 270 | Ti means tebi (Ti) or 1024**4 |
| 271 | .P |
| 272 | Pi means pebi (Pi) or 1024**5 |
| 273 | .PD |
| 274 | .RE |
| 275 | .P |
| 276 | With `kb_base=1024' (the default), the unit prefixes are opposite |
| 277 | from those specified in the SI and IEC 80000\-13 standards to provide |
| 278 | compatibility with old scripts. For example, 4k means 4096. |
| 279 | .P |
| 280 | For quantities of data, an optional unit of 'B' may be included |
| 281 | (e.g., 'kB' is the same as 'k'). |
| 282 | .P |
| 283 | The *integer suffix* is not case sensitive (e.g., m/mi mean mebi/mega, |
| 284 | not milli). 'b' and 'B' both mean byte, not bit. |
| 285 | .P |
| 286 | Examples with `kb_base=1000': |
| 287 | .RS |
| 288 | .P |
| 289 | .PD 0 |
| 290 | 4 KiB: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB |
| 291 | .P |
| 292 | 1 MiB: 1048576, 1m, 1024k |
| 293 | .P |
| 294 | 1 MB: 1000000, 1mi, 1000ki |
| 295 | .P |
| 296 | 1 TiB: 1073741824, 1t, 1024m, 1048576k |
| 297 | .P |
| 298 | 1 TB: 1000000000, 1ti, 1000mi, 1000000ki |
| 299 | .PD |
| 300 | .RE |
| 301 | .P |
| 302 | Examples with `kb_base=1024' (default): |
| 303 | .RS |
| 304 | .P |
| 305 | .PD 0 |
| 306 | 4 KiB: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB |
| 307 | .P |
| 308 | 1 MiB: 1048576, 1m, 1024k |
| 309 | .P |
| 310 | 1 MB: 1000000, 1mi, 1000ki |
| 311 | .P |
| 312 | 1 TiB: 1073741824, 1t, 1024m, 1048576k |
| 313 | .P |
| 314 | 1 TB: 1000000000, 1ti, 1000mi, 1000000ki |
| 315 | .PD |
| 316 | .RE |
| 317 | .P |
| 318 | To specify times (units are not case sensitive): |
| 319 | .RS |
| 320 | .P |
| 321 | .PD 0 |
| 322 | D means days |
| 323 | .P |
| 324 | H means hours |
| 325 | .P |
| 326 | M mean minutes |
| 327 | .P |
| 328 | s or sec means seconds (default) |
| 329 | .P |
| 330 | ms or msec means milliseconds |
| 331 | .P |
| 332 | us or usec means microseconds |
| 333 | .PD |
| 334 | .RE |
| 335 | .P |
| 336 | If the option accepts an upper and lower range, use a colon ':' or |
| 337 | minus '\-' to separate such values. See \fIirange\fR parameter type. |
| 338 | If the lower value specified happens to be larger than the upper value |
| 339 | the two values are swapped. |
| 340 | .RE |
| 341 | .TP |
| 342 | .I bool |
| 343 | Boolean. Usually parsed as an integer, however only defined for |
| 344 | true and false (1 and 0). |
| 345 | .TP |
| 346 | .I irange |
| 347 | Integer range with suffix. Allows value range to be given, such as |
| 348 | 1024\-4096. A colon may also be used as the separator, e.g. 1k:4k. If the |
| 349 | option allows two sets of ranges, they can be specified with a ',' or '/' |
| 350 | delimiter: 1k\-4k/8k\-32k. Also see \fIint\fR parameter type. |
| 351 | .TP |
| 352 | .I float_list |
| 353 | A list of floating point numbers, separated by a ':' character. |
| 354 | .SH "JOB PARAMETERS" |
| 355 | With the above in mind, here follows the complete list of fio job parameters. |
| 356 | .SS "Units" |
| 357 | .TP |
| 358 | .BI kb_base \fR=\fPint |
| 359 | Select the interpretation of unit prefixes in input parameters. |
| 360 | .RS |
| 361 | .RS |
| 362 | .TP |
| 363 | .B 1000 |
| 364 | Inputs comply with IEC 80000\-13 and the International |
| 365 | System of Units (SI). Use: |
| 366 | .RS |
| 367 | .P |
| 368 | .PD 0 |
| 369 | \- power\-of\-2 values with IEC prefixes (e.g., KiB) |
| 370 | .P |
| 371 | \- power\-of\-10 values with SI prefixes (e.g., kB) |
| 372 | .PD |
| 373 | .RE |
| 374 | .TP |
| 375 | .B 1024 |
| 376 | Compatibility mode (default). To avoid breaking old scripts: |
| 377 | .P |
| 378 | .RS |
| 379 | .PD 0 |
| 380 | \- power\-of\-2 values with SI prefixes |
| 381 | .P |
| 382 | \- power\-of\-10 values with IEC prefixes |
| 383 | .PD |
| 384 | .RE |
| 385 | .RE |
| 386 | .P |
| 387 | See \fBbs\fR for more details on input parameters. |
| 388 | .P |
| 389 | Outputs always use correct prefixes. Most outputs include both |
| 390 | side\-by\-side, like: |
| 391 | .P |
| 392 | .RS |
| 393 | bw=2383.3kB/s (2327.4KiB/s) |
| 394 | .RE |
| 395 | .P |
| 396 | If only one value is reported, then kb_base selects the one to use: |
| 397 | .P |
| 398 | .RS |
| 399 | .PD 0 |
| 400 | 1000 \-\- SI prefixes |
| 401 | .P |
| 402 | 1024 \-\- IEC prefixes |
| 403 | .PD |
| 404 | .RE |
| 405 | .RE |
| 406 | .TP |
| 407 | .BI unit_base \fR=\fPint |
| 408 | Base unit for reporting. Allowed values are: |
| 409 | .RS |
| 410 | .RS |
| 411 | .TP |
| 412 | .B 0 |
| 413 | Use auto\-detection (default). |
| 414 | .TP |
| 415 | .B 8 |
| 416 | Byte based. |
| 417 | .TP |
| 418 | .B 1 |
| 419 | Bit based. |
| 420 | .RE |
| 421 | .RE |
| 422 | .SS "Job description" |
| 423 | .TP |
| 424 | .BI name \fR=\fPstr |
| 425 | ASCII name of the job. This may be used to override the name printed by fio |
| 426 | for this job. Otherwise the job name is used. On the command line this |
| 427 | parameter has the special purpose of also signaling the start of a new job. |
| 428 | .TP |
| 429 | .BI description \fR=\fPstr |
| 430 | Text description of the job. Doesn't do anything except dump this text |
| 431 | description when this job is run. It's not parsed. |
| 432 | .TP |
| 433 | .BI loops \fR=\fPint |
| 434 | Run the specified number of iterations of this job. Used to repeat the same |
| 435 | workload a given number of times. Defaults to 1. |
| 436 | .TP |
| 437 | .BI numjobs \fR=\fPint |
| 438 | Create the specified number of clones of this job. Each clone of job |
| 439 | is spawned as an independent thread or process. May be used to setup a |
| 440 | larger number of threads/processes doing the same thing. Each thread is |
| 441 | reported separately; to see statistics for all clones as a whole, use |
| 442 | \fBgroup_reporting\fR in conjunction with \fBnew_group\fR. |
| 443 | See \fB\-\-max\-jobs\fR. Default: 1. |
| 444 | .SS "Time related parameters" |
| 445 | .TP |
| 446 | .BI runtime \fR=\fPtime |
| 447 | Tell fio to terminate processing after the specified period of time. It |
| 448 | can be quite hard to determine for how long a specified job will run, so |
| 449 | this parameter is handy to cap the total runtime to a given time. When |
| 450 | the unit is omitted, the value is intepreted in seconds. |
| 451 | .TP |
| 452 | .BI time_based |
| 453 | If set, fio will run for the duration of the \fBruntime\fR specified |
| 454 | even if the file(s) are completely read or written. It will simply loop over |
| 455 | the same workload as many times as the \fBruntime\fR allows. |
| 456 | .TP |
| 457 | .BI startdelay \fR=\fPirange(int) |
| 458 | Delay the start of job for the specified amount of time. Can be a single |
| 459 | value or a range. When given as a range, each thread will choose a value |
| 460 | randomly from within the range. Value is in seconds if a unit is omitted. |
| 461 | .TP |
| 462 | .BI ramp_time \fR=\fPtime |
| 463 | If set, fio will run the specified workload for this amount of time before |
| 464 | logging any performance numbers. Useful for letting performance settle |
| 465 | before logging results, thus minimizing the runtime required for stable |
| 466 | results. Note that the \fBramp_time\fR is considered lead in time for a job, |
| 467 | thus it will increase the total runtime if a special timeout or |
| 468 | \fBruntime\fR is specified. When the unit is omitted, the value is |
| 469 | given in seconds. |
| 470 | .TP |
| 471 | .BI clocksource \fR=\fPstr |
| 472 | Use the given clocksource as the base of timing. The supported options are: |
| 473 | .RS |
| 474 | .RS |
| 475 | .TP |
| 476 | .B gettimeofday |
| 477 | \fBgettimeofday\fR\|(2) |
| 478 | .TP |
| 479 | .B clock_gettime |
| 480 | \fBclock_gettime\fR\|(2) |
| 481 | .TP |
| 482 | .B cpu |
| 483 | Internal CPU clock source |
| 484 | .RE |
| 485 | .P |
| 486 | \fBcpu\fR is the preferred clocksource if it is reliable, as it is very fast (and |
| 487 | fio is heavy on time calls). Fio will automatically use this clocksource if |
| 488 | it's supported and considered reliable on the system it is running on, |
| 489 | unless another clocksource is specifically set. For x86/x86\-64 CPUs, this |
| 490 | means supporting TSC Invariant. |
| 491 | .RE |
| 492 | .TP |
| 493 | .BI gtod_reduce \fR=\fPbool |
| 494 | Enable all of the \fBgettimeofday\fR\|(2) reducing options |
| 495 | (\fBdisable_clat\fR, \fBdisable_slat\fR, \fBdisable_bw_measurement\fR) plus |
| 496 | reduce precision of the timeout somewhat to really shrink the |
| 497 | \fBgettimeofday\fR\|(2) call count. With this option enabled, we only do |
| 498 | about 0.4% of the \fBgettimeofday\fR\|(2) calls we would have done if all |
| 499 | time keeping was enabled. |
| 500 | .TP |
| 501 | .BI gtod_cpu \fR=\fPint |
| 502 | Sometimes it's cheaper to dedicate a single thread of execution to just |
| 503 | getting the current time. Fio (and databases, for instance) are very |
| 504 | intensive on \fBgettimeofday\fR\|(2) calls. With this option, you can set |
| 505 | one CPU aside for doing nothing but logging current time to a shared memory |
| 506 | location. Then the other threads/processes that run I/O workloads need only |
| 507 | copy that segment, instead of entering the kernel with a |
| 508 | \fBgettimeofday\fR\|(2) call. The CPU set aside for doing these time |
| 509 | calls will be excluded from other uses. Fio will manually clear it from the |
| 510 | CPU mask of other jobs. |
| 511 | .SS "Target file/device" |
| 512 | .TP |
| 513 | .BI directory \fR=\fPstr |
| 514 | Prefix \fBfilename\fRs with this directory. Used to place files in a different |
| 515 | location than `./'. You can specify a number of directories by |
| 516 | separating the names with a ':' character. These directories will be |
| 517 | assigned equally distributed to job clones created by \fBnumjobs\fR as |
| 518 | long as they are using generated filenames. If specific \fBfilename\fR(s) are |
| 519 | set fio will use the first listed directory, and thereby matching the |
| 520 | \fBfilename\fR semantic which generates a file each clone if not specified, but |
| 521 | let all clones use the same if set. |
| 522 | .RS |
| 523 | .P |
| 524 | See the \fBfilename\fR option for information on how to escape ':' and '\' |
| 525 | characters within the directory path itself. |
| 526 | .RE |
| 527 | .TP |
| 528 | .BI filename \fR=\fPstr |
| 529 | Fio normally makes up a \fBfilename\fR based on the job name, thread number, and |
| 530 | file number (see \fBfilename_format\fR). If you want to share files |
| 531 | between threads in a job or several |
| 532 | jobs with fixed file paths, specify a \fBfilename\fR for each of them to override |
| 533 | the default. If the ioengine is file based, you can specify a number of files |
| 534 | by separating the names with a ':' colon. So if you wanted a job to open |
| 535 | `/dev/sda' and `/dev/sdb' as the two working files, you would use |
| 536 | `filename=/dev/sda:/dev/sdb'. This also means that whenever this option is |
| 537 | specified, \fBnrfiles\fR is ignored. The size of regular files specified |
| 538 | by this option will be \fBsize\fR divided by number of files unless an |
| 539 | explicit size is specified by \fBfilesize\fR. |
| 540 | .RS |
| 541 | .P |
| 542 | Each colon and backslash in the wanted path must be escaped with a '\' |
| 543 | character. For instance, if the path is `/dev/dsk/foo@3,0:c' then you |
| 544 | would use `filename=/dev/dsk/foo@3,0\\:c' and if the path is |
| 545 | `F:\\\\filename' then you would use `filename=F\\:\\\\filename'. |
| 546 | .P |
| 547 | On Windows, disk devices are accessed as `\\\\\\\\.\\\\PhysicalDrive0' for |
| 548 | the first device, `\\\\\\\\.\\\\PhysicalDrive1' for the second etc. |
| 549 | Note: Windows and FreeBSD prevent write access to areas |
| 550 | of the disk containing in\-use data (e.g. filesystems). |
| 551 | .P |
| 552 | The filename `\-' is a reserved name, meaning *stdin* or *stdout*. Which |
| 553 | of the two depends on the read/write direction set. |
| 554 | .RE |
| 555 | .TP |
| 556 | .BI filename_format \fR=\fPstr |
| 557 | If sharing multiple files between jobs, it is usually necessary to have fio |
| 558 | generate the exact names that you want. By default, fio will name a file |
| 559 | based on the default file format specification of |
| 560 | `jobname.jobnumber.filenumber'. With this option, that can be |
| 561 | customized. Fio will recognize and replace the following keywords in this |
| 562 | string: |
| 563 | .RS |
| 564 | .RS |
| 565 | .TP |
| 566 | .B $jobname |
| 567 | The name of the worker thread or process. |
| 568 | .TP |
| 569 | .B $jobnum |
| 570 | The incremental number of the worker thread or process. |
| 571 | .TP |
| 572 | .B $filenum |
| 573 | The incremental number of the file for that worker thread or process. |
| 574 | .RE |
| 575 | .P |
| 576 | To have dependent jobs share a set of files, this option can be set to have |
| 577 | fio generate filenames that are shared between the two. For instance, if |
| 578 | `testfiles.$filenum' is specified, file number 4 for any job will be |
| 579 | named `testfiles.4'. The default of `$jobname.$jobnum.$filenum' |
| 580 | will be used if no other format specifier is given. |
| 581 | .P |
| 582 | If you specify a path then the directories will be created up to the main |
| 583 | directory for the file. So for example if you specify `a/b/c/$jobnum` then the |
| 584 | directories a/b/c will be created before the file setup part of the job. If you |
| 585 | specify \fBdirectory\fR then the path will be relative that directory, otherwise |
| 586 | it is treated as the absolute path. |
| 587 | .RE |
| 588 | .TP |
| 589 | .BI unique_filename \fR=\fPbool |
| 590 | To avoid collisions between networked clients, fio defaults to prefixing any |
| 591 | generated filenames (with a directory specified) with the source of the |
| 592 | client connecting. To disable this behavior, set this option to 0. |
| 593 | .TP |
| 594 | .BI opendir \fR=\fPstr |
| 595 | Recursively open any files below directory \fIstr\fR. |
| 596 | .TP |
| 597 | .BI lockfile \fR=\fPstr |
| 598 | Fio defaults to not locking any files before it does I/O to them. If a file |
| 599 | or file descriptor is shared, fio can serialize I/O to that file to make the |
| 600 | end result consistent. This is usual for emulating real workloads that share |
| 601 | files. The lock modes are: |
| 602 | .RS |
| 603 | .RS |
| 604 | .TP |
| 605 | .B none |
| 606 | No locking. The default. |
| 607 | .TP |
| 608 | .B exclusive |
| 609 | Only one thread or process may do I/O at a time, excluding all others. |
| 610 | .TP |
| 611 | .B readwrite |
| 612 | Read\-write locking on the file. Many readers may |
| 613 | access the file at the same time, but writes get exclusive access. |
| 614 | .RE |
| 615 | .RE |
| 616 | .TP |
| 617 | .BI nrfiles \fR=\fPint |
| 618 | Number of files to use for this job. Defaults to 1. The size of files |
| 619 | will be \fBsize\fR divided by this unless explicit size is specified by |
| 620 | \fBfilesize\fR. Files are created for each thread separately, and each |
| 621 | file will have a file number within its name by default, as explained in |
| 622 | \fBfilename\fR section. |
| 623 | .TP |
| 624 | .BI openfiles \fR=\fPint |
| 625 | Number of files to keep open at the same time. Defaults to the same as |
| 626 | \fBnrfiles\fR, can be set smaller to limit the number simultaneous |
| 627 | opens. |
| 628 | .TP |
| 629 | .BI file_service_type \fR=\fPstr |
| 630 | Defines how fio decides which file from a job to service next. The following |
| 631 | types are defined: |
| 632 | .RS |
| 633 | .RS |
| 634 | .TP |
| 635 | .B random |
| 636 | Choose a file at random. |
| 637 | .TP |
| 638 | .B roundrobin |
| 639 | Round robin over opened files. This is the default. |
| 640 | .TP |
| 641 | .B sequential |
| 642 | Finish one file before moving on to the next. Multiple files can |
| 643 | still be open depending on \fBopenfiles\fR. |
| 644 | .TP |
| 645 | .B zipf |
| 646 | Use a Zipf distribution to decide what file to access. |
| 647 | .TP |
| 648 | .B pareto |
| 649 | Use a Pareto distribution to decide what file to access. |
| 650 | .TP |
| 651 | .B normal |
| 652 | Use a Gaussian (normal) distribution to decide what file to access. |
| 653 | .TP |
| 654 | .B gauss |
| 655 | Alias for normal. |
| 656 | .RE |
| 657 | .P |
| 658 | For \fBrandom\fR, \fBroundrobin\fR, and \fBsequential\fR, a postfix can be appended to |
| 659 | tell fio how many I/Os to issue before switching to a new file. For example, |
| 660 | specifying `file_service_type=random:8' would cause fio to issue |
| 661 | 8 I/Os before selecting a new file at random. For the non\-uniform |
| 662 | distributions, a floating point postfix can be given to influence how the |
| 663 | distribution is skewed. See \fBrandom_distribution\fR for a description |
| 664 | of how that would work. |
| 665 | .RE |
| 666 | .TP |
| 667 | .BI ioscheduler \fR=\fPstr |
| 668 | Attempt to switch the device hosting the file to the specified I/O scheduler |
| 669 | before running. |
| 670 | .TP |
| 671 | .BI create_serialize \fR=\fPbool |
| 672 | If true, serialize the file creation for the jobs. This may be handy to |
| 673 | avoid interleaving of data files, which may greatly depend on the filesystem |
| 674 | used and even the number of processors in the system. Default: true. |
| 675 | .TP |
| 676 | .BI create_fsync \fR=\fPbool |
| 677 | \fBfsync\fR\|(2) the data file after creation. This is the default. |
| 678 | .TP |
| 679 | .BI create_on_open \fR=\fPbool |
| 680 | If true, don't pre\-create files but allow the job's open() to create a file |
| 681 | when it's time to do I/O. Default: false \-\- pre\-create all necessary files |
| 682 | when the job starts. |
| 683 | .TP |
| 684 | .BI create_only \fR=\fPbool |
| 685 | If true, fio will only run the setup phase of the job. If files need to be |
| 686 | laid out or updated on disk, only that will be done \-\- the actual job contents |
| 687 | are not executed. Default: false. |
| 688 | .TP |
| 689 | .BI allow_file_create \fR=\fPbool |
| 690 | If true, fio is permitted to create files as part of its workload. If this |
| 691 | option is false, then fio will error out if |
| 692 | the files it needs to use don't already exist. Default: true. |
| 693 | .TP |
| 694 | .BI allow_mounted_write \fR=\fPbool |
| 695 | If this isn't set, fio will abort jobs that are destructive (e.g. that write) |
| 696 | to what appears to be a mounted device or partition. This should help catch |
| 697 | creating inadvertently destructive tests, not realizing that the test will |
| 698 | destroy data on the mounted file system. Note that some platforms don't allow |
| 699 | writing against a mounted device regardless of this option. Default: false. |
| 700 | .TP |
| 701 | .BI pre_read \fR=\fPbool |
| 702 | If this is given, files will be pre\-read into memory before starting the |
| 703 | given I/O operation. This will also clear the \fBinvalidate\fR flag, |
| 704 | since it is pointless to pre\-read and then drop the cache. This will only |
| 705 | work for I/O engines that are seek\-able, since they allow you to read the |
| 706 | same data multiple times. Thus it will not work on non\-seekable I/O engines |
| 707 | (e.g. network, splice). Default: false. |
| 708 | .TP |
| 709 | .BI unlink \fR=\fPbool |
| 710 | Unlink the job files when done. Not the default, as repeated runs of that |
| 711 | job would then waste time recreating the file set again and again. Default: |
| 712 | false. |
| 713 | .TP |
| 714 | .BI unlink_each_loop \fR=\fPbool |
| 715 | Unlink job files after each iteration or loop. Default: false. |
| 716 | .TP |
| 717 | .BI zonesize \fR=\fPint |
| 718 | Divide a file into zones of the specified size. See \fBzoneskip\fR. |
| 719 | .TP |
| 720 | .BI zonerange \fR=\fPint |
| 721 | Give size of an I/O zone. See \fBzoneskip\fR. |
| 722 | .TP |
| 723 | .BI zoneskip \fR=\fPint |
| 724 | Skip the specified number of bytes when \fBzonesize\fR data has been |
| 725 | read. The two zone options can be used to only do I/O on zones of a file. |
| 726 | .SS "I/O type" |
| 727 | .TP |
| 728 | .BI direct \fR=\fPbool |
| 729 | If value is true, use non\-buffered I/O. This is usually O_DIRECT. Note that |
| 730 | OpenBSD and ZFS on Solaris don't support direct I/O. On Windows the synchronous |
| 731 | ioengines don't support direct I/O. Default: false. |
| 732 | .TP |
| 733 | .BI atomic \fR=\fPbool |
| 734 | If value is true, attempt to use atomic direct I/O. Atomic writes are |
| 735 | guaranteed to be stable once acknowledged by the operating system. Only |
| 736 | Linux supports O_ATOMIC right now. |
| 737 | .TP |
| 738 | .BI buffered \fR=\fPbool |
| 739 | If value is true, use buffered I/O. This is the opposite of the |
| 740 | \fBdirect\fR option. Defaults to true. |
| 741 | .TP |
| 742 | .BI readwrite \fR=\fPstr "\fR,\fP rw" \fR=\fPstr |
| 743 | Type of I/O pattern. Accepted values are: |
| 744 | .RS |
| 745 | .RS |
| 746 | .TP |
| 747 | .B read |
| 748 | Sequential reads. |
| 749 | .TP |
| 750 | .B write |
| 751 | Sequential writes. |
| 752 | .TP |
| 753 | .B trim |
| 754 | Sequential trims (Linux block devices only). |
| 755 | .TP |
| 756 | .B randread |
| 757 | Random reads. |
| 758 | .TP |
| 759 | .B randwrite |
| 760 | Random writes. |
| 761 | .TP |
| 762 | .B randtrim |
| 763 | Random trims (Linux block devices only). |
| 764 | .TP |
| 765 | .B rw,readwrite |
| 766 | Sequential mixed reads and writes. |
| 767 | .TP |
| 768 | .B randrw |
| 769 | Random mixed reads and writes. |
| 770 | .TP |
| 771 | .B trimwrite |
| 772 | Sequential trim+write sequences. Blocks will be trimmed first, |
| 773 | then the same blocks will be written to. |
| 774 | .RE |
| 775 | .P |
| 776 | Fio defaults to read if the option is not specified. For the mixed I/O |
| 777 | types, the default is to split them 50/50. For certain types of I/O the |
| 778 | result may still be skewed a bit, since the speed may be different. |
| 779 | .P |
| 780 | It is possible to specify the number of I/Os to do before getting a new |
| 781 | offset by appending `:<nr>' to the end of the string given. For a |
| 782 | random read, it would look like `rw=randread:8' for passing in an offset |
| 783 | modifier with a value of 8. If the suffix is used with a sequential I/O |
| 784 | pattern, then the `<nr>' value specified will be added to the generated |
| 785 | offset for each I/O turning sequential I/O into sequential I/O with holes. |
| 786 | For instance, using `rw=write:4k' will skip 4k for every write. Also see |
| 787 | the \fBrw_sequencer\fR option. |
| 788 | .RE |
| 789 | .TP |
| 790 | .BI rw_sequencer \fR=\fPstr |
| 791 | If an offset modifier is given by appending a number to the `rw=\fIstr\fR' |
| 792 | line, then this option controls how that number modifies the I/O offset |
| 793 | being generated. Accepted values are: |
| 794 | .RS |
| 795 | .RS |
| 796 | .TP |
| 797 | .B sequential |
| 798 | Generate sequential offset. |
| 799 | .TP |
| 800 | .B identical |
| 801 | Generate the same offset. |
| 802 | .RE |
| 803 | .P |
| 804 | \fBsequential\fR is only useful for random I/O, where fio would normally |
| 805 | generate a new random offset for every I/O. If you append e.g. 8 to randread, |
| 806 | you would get a new random offset for every 8 I/Os. The result would be a |
| 807 | seek for only every 8 I/Os, instead of for every I/O. Use `rw=randread:8' |
| 808 | to specify that. As sequential I/O is already sequential, setting |
| 809 | \fBsequential\fR for that would not result in any differences. \fBidentical\fR |
| 810 | behaves in a similar fashion, except it sends the same offset 8 number of |
| 811 | times before generating a new offset. |
| 812 | .RE |
| 813 | .TP |
| 814 | .BI unified_rw_reporting \fR=\fPbool |
| 815 | Fio normally reports statistics on a per data direction basis, meaning that |
| 816 | reads, writes, and trims are accounted and reported separately. If this |
| 817 | option is set fio sums the results and report them as "mixed" instead. |
| 818 | .TP |
| 819 | .BI randrepeat \fR=\fPbool |
| 820 | Seed the random number generator used for random I/O patterns in a |
| 821 | predictable way so the pattern is repeatable across runs. Default: true. |
| 822 | .TP |
| 823 | .BI allrandrepeat \fR=\fPbool |
| 824 | Seed all random number generators in a predictable way so results are |
| 825 | repeatable across runs. Default: false. |
| 826 | .TP |
| 827 | .BI randseed \fR=\fPint |
| 828 | Seed the random number generators based on this seed value, to be able to |
| 829 | control what sequence of output is being generated. If not set, the random |
| 830 | sequence depends on the \fBrandrepeat\fR setting. |
| 831 | .TP |
| 832 | .BI fallocate \fR=\fPstr |
| 833 | Whether pre\-allocation is performed when laying down files. |
| 834 | Accepted values are: |
| 835 | .RS |
| 836 | .RS |
| 837 | .TP |
| 838 | .B none |
| 839 | Do not pre\-allocate space. |
| 840 | .TP |
| 841 | .B native |
| 842 | Use a platform's native pre\-allocation call but fall back to |
| 843 | \fBnone\fR behavior if it fails/is not implemented. |
| 844 | .TP |
| 845 | .B posix |
| 846 | Pre\-allocate via \fBposix_fallocate\fR\|(3). |
| 847 | .TP |
| 848 | .B keep |
| 849 | Pre\-allocate via \fBfallocate\fR\|(2) with |
| 850 | FALLOC_FL_KEEP_SIZE set. |
| 851 | .TP |
| 852 | .B 0 |
| 853 | Backward\-compatible alias for \fBnone\fR. |
| 854 | .TP |
| 855 | .B 1 |
| 856 | Backward\-compatible alias for \fBposix\fR. |
| 857 | .RE |
| 858 | .P |
| 859 | May not be available on all supported platforms. \fBkeep\fR is only available |
| 860 | on Linux. If using ZFS on Solaris this cannot be set to \fBposix\fR |
| 861 | because ZFS doesn't support pre\-allocation. Default: \fBnative\fR if any |
| 862 | pre\-allocation methods are available, \fBnone\fR if not. |
| 863 | .RE |
| 864 | .TP |
| 865 | .BI fadvise_hint \fR=\fPstr |
| 866 | Use \fBposix_fadvise\fR\|(2) to advise the kernel what I/O patterns |
| 867 | are likely to be issued. Accepted values are: |
| 868 | .RS |
| 869 | .RS |
| 870 | .TP |
| 871 | .B 0 |
| 872 | Backwards compatible hint for "no hint". |
| 873 | .TP |
| 874 | .B 1 |
| 875 | Backwards compatible hint for "advise with fio workload type". This |
| 876 | uses FADV_RANDOM for a random workload, and FADV_SEQUENTIAL |
| 877 | for a sequential workload. |
| 878 | .TP |
| 879 | .B sequential |
| 880 | Advise using FADV_SEQUENTIAL. |
| 881 | .TP |
| 882 | .B random |
| 883 | Advise using FADV_RANDOM. |
| 884 | .RE |
| 885 | .RE |
| 886 | .TP |
| 887 | .BI write_hint \fR=\fPstr |
| 888 | Use \fBfcntl\fR\|(2) to advise the kernel what life time to expect |
| 889 | from a write. Only supported on Linux, as of version 4.13. Accepted |
| 890 | values are: |
| 891 | .RS |
| 892 | .RS |
| 893 | .TP |
| 894 | .B none |
| 895 | No particular life time associated with this file. |
| 896 | .TP |
| 897 | .B short |
| 898 | Data written to this file has a short life time. |
| 899 | .TP |
| 900 | .B medium |
| 901 | Data written to this file has a medium life time. |
| 902 | .TP |
| 903 | .B long |
| 904 | Data written to this file has a long life time. |
| 905 | .TP |
| 906 | .B extreme |
| 907 | Data written to this file has a very long life time. |
| 908 | .RE |
| 909 | .P |
| 910 | The values are all relative to each other, and no absolute meaning |
| 911 | should be associated with them. |
| 912 | .RE |
| 913 | .TP |
| 914 | .BI offset \fR=\fPint |
| 915 | Start I/O at the provided offset in the file, given as either a fixed size in |
| 916 | bytes or a percentage. If a percentage is given, the generated offset will be |
| 917 | aligned to the minimum \fBblocksize\fR or to the value of \fBoffset_align\fR if |
| 918 | provided. Data before the given offset will not be touched. This |
| 919 | effectively caps the file size at `real_size \- offset'. Can be combined with |
| 920 | \fBsize\fR to constrain the start and end range of the I/O workload. |
| 921 | A percentage can be specified by a number between 1 and 100 followed by '%', |
| 922 | for example, `offset=20%' to specify 20%. |
| 923 | .TP |
| 924 | .BI offset_align \fR=\fPint |
| 925 | If set to non-zero value, the byte offset generated by a percentage \fBoffset\fR |
| 926 | is aligned upwards to this value. Defaults to 0 meaning that a percentage |
| 927 | offset is aligned to the minimum block size. |
| 928 | .TP |
| 929 | .BI offset_increment \fR=\fPint |
| 930 | If this is provided, then the real offset becomes `\fBoffset\fR + \fBoffset_increment\fR |
| 931 | * thread_number', where the thread number is a counter that starts at 0 and |
| 932 | is incremented for each sub\-job (i.e. when \fBnumjobs\fR option is |
| 933 | specified). This option is useful if there are several jobs which are |
| 934 | intended to operate on a file in parallel disjoint segments, with even |
| 935 | spacing between the starting points. |
| 936 | .TP |
| 937 | .BI number_ios \fR=\fPint |
| 938 | Fio will normally perform I/Os until it has exhausted the size of the region |
| 939 | set by \fBsize\fR, or if it exhaust the allocated time (or hits an error |
| 940 | condition). With this setting, the range/size can be set independently of |
| 941 | the number of I/Os to perform. When fio reaches this number, it will exit |
| 942 | normally and report status. Note that this does not extend the amount of I/O |
| 943 | that will be done, it will only stop fio if this condition is met before |
| 944 | other end\-of\-job criteria. |
| 945 | .TP |
| 946 | .BI fsync \fR=\fPint |
| 947 | If writing to a file, issue an \fBfsync\fR\|(2) (or its equivalent) of |
| 948 | the dirty data for every number of blocks given. For example, if you give 32 |
| 949 | as a parameter, fio will sync the file after every 32 writes issued. If fio is |
| 950 | using non\-buffered I/O, we may not sync the file. The exception is the sg |
| 951 | I/O engine, which synchronizes the disk cache anyway. Defaults to 0, which |
| 952 | means fio does not periodically issue and wait for a sync to complete. Also |
| 953 | see \fBend_fsync\fR and \fBfsync_on_close\fR. |
| 954 | .TP |
| 955 | .BI fdatasync \fR=\fPint |
| 956 | Like \fBfsync\fR but uses \fBfdatasync\fR\|(2) to only sync data and |
| 957 | not metadata blocks. In Windows, FreeBSD, and DragonFlyBSD there is no |
| 958 | \fBfdatasync\fR\|(2) so this falls back to using \fBfsync\fR\|(2). |
| 959 | Defaults to 0, which means fio does not periodically issue and wait for a |
| 960 | data\-only sync to complete. |
| 961 | .TP |
| 962 | .BI write_barrier \fR=\fPint |
| 963 | Make every N\-th write a barrier write. |
| 964 | .TP |
| 965 | .BI sync_file_range \fR=\fPstr:int |
| 966 | Use \fBsync_file_range\fR\|(2) for every \fIint\fR number of write |
| 967 | operations. Fio will track range of writes that have happened since the last |
| 968 | \fBsync_file_range\fR\|(2) call. \fIstr\fR can currently be one or more of: |
| 969 | .RS |
| 970 | .RS |
| 971 | .TP |
| 972 | .B wait_before |
| 973 | SYNC_FILE_RANGE_WAIT_BEFORE |
| 974 | .TP |
| 975 | .B write |
| 976 | SYNC_FILE_RANGE_WRITE |
| 977 | .TP |
| 978 | .B wait_after |
| 979 | SYNC_FILE_RANGE_WRITE_AFTER |
| 980 | .RE |
| 981 | .P |
| 982 | So if you do `sync_file_range=wait_before,write:8', fio would use |
| 983 | `SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE' for every 8 |
| 984 | writes. Also see the \fBsync_file_range\fR\|(2) man page. This option is |
| 985 | Linux specific. |
| 986 | .RE |
| 987 | .TP |
| 988 | .BI overwrite \fR=\fPbool |
| 989 | If true, writes to a file will always overwrite existing data. If the file |
| 990 | doesn't already exist, it will be created before the write phase begins. If |
| 991 | the file exists and is large enough for the specified write phase, nothing |
| 992 | will be done. Default: false. |
| 993 | .TP |
| 994 | .BI end_fsync \fR=\fPbool |
| 995 | If true, \fBfsync\fR\|(2) file contents when a write stage has completed. |
| 996 | Default: false. |
| 997 | .TP |
| 998 | .BI fsync_on_close \fR=\fPbool |
| 999 | If true, fio will \fBfsync\fR\|(2) a dirty file on close. This differs |
| 1000 | from \fBend_fsync\fR in that it will happen on every file close, not |
| 1001 | just at the end of the job. Default: false. |
| 1002 | .TP |
| 1003 | .BI rwmixread \fR=\fPint |
| 1004 | Percentage of a mixed workload that should be reads. Default: 50. |
| 1005 | .TP |
| 1006 | .BI rwmixwrite \fR=\fPint |
| 1007 | Percentage of a mixed workload that should be writes. If both |
| 1008 | \fBrwmixread\fR and \fBrwmixwrite\fR is given and the values do not |
| 1009 | add up to 100%, the latter of the two will be used to override the |
| 1010 | first. This may interfere with a given rate setting, if fio is asked to |
| 1011 | limit reads or writes to a certain rate. If that is the case, then the |
| 1012 | distribution may be skewed. Default: 50. |
| 1013 | .TP |
| 1014 | .BI random_distribution \fR=\fPstr:float[,str:float][,str:float] |
| 1015 | By default, fio will use a completely uniform random distribution when asked |
| 1016 | to perform random I/O. Sometimes it is useful to skew the distribution in |
| 1017 | specific ways, ensuring that some parts of the data is more hot than others. |
| 1018 | fio includes the following distribution models: |
| 1019 | .RS |
| 1020 | .RS |
| 1021 | .TP |
| 1022 | .B random |
| 1023 | Uniform random distribution |
| 1024 | .TP |
| 1025 | .B zipf |
| 1026 | Zipf distribution |
| 1027 | .TP |
| 1028 | .B pareto |
| 1029 | Pareto distribution |
| 1030 | .TP |
| 1031 | .B normal |
| 1032 | Normal (Gaussian) distribution |
| 1033 | .TP |
| 1034 | .B zoned |
| 1035 | Zoned random distribution |
| 1036 | .RE |
| 1037 | .P |
| 1038 | When using a \fBzipf\fR or \fBpareto\fR distribution, an input value is also |
| 1039 | needed to define the access pattern. For \fBzipf\fR, this is the `Zipf theta'. |
| 1040 | For \fBpareto\fR, it's the `Pareto power'. Fio includes a test |
| 1041 | program, \fBfio\-genzipf\fR, that can be used visualize what the given input |
| 1042 | values will yield in terms of hit rates. If you wanted to use \fBzipf\fR with |
| 1043 | a `theta' of 1.2, you would use `random_distribution=zipf:1.2' as the |
| 1044 | option. If a non\-uniform model is used, fio will disable use of the random |
| 1045 | map. For the \fBnormal\fR distribution, a normal (Gaussian) deviation is |
| 1046 | supplied as a value between 0 and 100. |
| 1047 | .P |
| 1048 | For a \fBzoned\fR distribution, fio supports specifying percentages of I/O |
| 1049 | access that should fall within what range of the file or device. For |
| 1050 | example, given a criteria of: |
| 1051 | .RS |
| 1052 | .P |
| 1053 | .PD 0 |
| 1054 | 60% of accesses should be to the first 10% |
| 1055 | .P |
| 1056 | 30% of accesses should be to the next 20% |
| 1057 | .P |
| 1058 | 8% of accesses should be to the next 30% |
| 1059 | .P |
| 1060 | 2% of accesses should be to the next 40% |
| 1061 | .PD |
| 1062 | .RE |
| 1063 | .P |
| 1064 | we can define that through zoning of the random accesses. For the above |
| 1065 | example, the user would do: |
| 1066 | .RS |
| 1067 | .P |
| 1068 | random_distribution=zoned:60/10:30/20:8/30:2/40 |
| 1069 | .RE |
| 1070 | .P |
| 1071 | similarly to how \fBbssplit\fR works for setting ranges and percentages |
| 1072 | of block sizes. Like \fBbssplit\fR, it's possible to specify separate |
| 1073 | zones for reads, writes, and trims. If just one set is given, it'll apply to |
| 1074 | all of them. |
| 1075 | .RE |
| 1076 | .TP |
| 1077 | .BI percentage_random \fR=\fPint[,int][,int] |
| 1078 | For a random workload, set how big a percentage should be random. This |
| 1079 | defaults to 100%, in which case the workload is fully random. It can be set |
| 1080 | from anywhere from 0 to 100. Setting it to 0 would make the workload fully |
| 1081 | sequential. Any setting in between will result in a random mix of sequential |
| 1082 | and random I/O, at the given percentages. Comma\-separated values may be |
| 1083 | specified for reads, writes, and trims as described in \fBblocksize\fR. |
| 1084 | .TP |
| 1085 | .BI norandommap |
| 1086 | Normally fio will cover every block of the file when doing random I/O. If |
| 1087 | this option is given, fio will just get a new random offset without looking |
| 1088 | at past I/O history. This means that some blocks may not be read or written, |
| 1089 | and that some blocks may be read/written more than once. If this option is |
| 1090 | used with \fBverify\fR and multiple blocksizes (via \fBbsrange\fR), |
| 1091 | only intact blocks are verified, i.e., partially\-overwritten blocks are |
| 1092 | ignored. |
| 1093 | .TP |
| 1094 | .BI softrandommap \fR=\fPbool |
| 1095 | See \fBnorandommap\fR. If fio runs with the random block map enabled and |
| 1096 | it fails to allocate the map, if this option is set it will continue without |
| 1097 | a random block map. As coverage will not be as complete as with random maps, |
| 1098 | this option is disabled by default. |
| 1099 | .TP |
| 1100 | .BI random_generator \fR=\fPstr |
| 1101 | Fio supports the following engines for generating I/O offsets for random I/O: |
| 1102 | .RS |
| 1103 | .RS |
| 1104 | .TP |
| 1105 | .B tausworthe |
| 1106 | Strong 2^88 cycle random number generator. |
| 1107 | .TP |
| 1108 | .B lfsr |
| 1109 | Linear feedback shift register generator. |
| 1110 | .TP |
| 1111 | .B tausworthe64 |
| 1112 | Strong 64\-bit 2^258 cycle random number generator. |
| 1113 | .RE |
| 1114 | .P |
| 1115 | \fBtausworthe\fR is a strong random number generator, but it requires tracking |
| 1116 | on the side if we want to ensure that blocks are only read or written |
| 1117 | once. \fBlfsr\fR guarantees that we never generate the same offset twice, and |
| 1118 | it's also less computationally expensive. It's not a true random generator, |
| 1119 | however, though for I/O purposes it's typically good enough. \fBlfsr\fR only |
| 1120 | works with single block sizes, not with workloads that use multiple block |
| 1121 | sizes. If used with such a workload, fio may read or write some blocks |
| 1122 | multiple times. The default value is \fBtausworthe\fR, unless the required |
| 1123 | space exceeds 2^32 blocks. If it does, then \fBtausworthe64\fR is |
| 1124 | selected automatically. |
| 1125 | .RE |
| 1126 | .SS "Block size" |
| 1127 | .TP |
| 1128 | .BI blocksize \fR=\fPint[,int][,int] "\fR,\fB bs" \fR=\fPint[,int][,int] |
| 1129 | The block size in bytes used for I/O units. Default: 4096. A single value |
| 1130 | applies to reads, writes, and trims. Comma\-separated values may be |
| 1131 | specified for reads, writes, and trims. A value not terminated in a comma |
| 1132 | applies to subsequent types. Examples: |
| 1133 | .RS |
| 1134 | .RS |
| 1135 | .P |
| 1136 | .PD 0 |
| 1137 | bs=256k means 256k for reads, writes and trims. |
| 1138 | .P |
| 1139 | bs=8k,32k means 8k for reads, 32k for writes and trims. |
| 1140 | .P |
| 1141 | bs=8k,32k, means 8k for reads, 32k for writes, and default for trims. |
| 1142 | .P |
| 1143 | bs=,8k means default for reads, 8k for writes and trims. |
| 1144 | .P |
| 1145 | bs=,8k, means default for reads, 8k for writes, and default for trims. |
| 1146 | .PD |
| 1147 | .RE |
| 1148 | .RE |
| 1149 | .TP |
| 1150 | .BI blocksize_range \fR=\fPirange[,irange][,irange] "\fR,\fB bsrange" \fR=\fPirange[,irange][,irange] |
| 1151 | A range of block sizes in bytes for I/O units. The issued I/O unit will |
| 1152 | always be a multiple of the minimum size, unless |
| 1153 | \fBblocksize_unaligned\fR is set. |
| 1154 | Comma\-separated ranges may be specified for reads, writes, and trims as |
| 1155 | described in \fBblocksize\fR. Example: |
| 1156 | .RS |
| 1157 | .RS |
| 1158 | .P |
| 1159 | bsrange=1k\-4k,2k\-8k |
| 1160 | .RE |
| 1161 | .RE |
| 1162 | .TP |
| 1163 | .BI bssplit \fR=\fPstr[,str][,str] |
| 1164 | Sometimes you want even finer grained control of the block sizes issued, not |
| 1165 | just an even split between them. This option allows you to weight various |
| 1166 | block sizes, so that you are able to define a specific amount of block sizes |
| 1167 | issued. The format for this option is: |
| 1168 | .RS |
| 1169 | .RS |
| 1170 | .P |
| 1171 | bssplit=blocksize/percentage:blocksize/percentage |
| 1172 | .RE |
| 1173 | .P |
| 1174 | for as many block sizes as needed. So if you want to define a workload that |
| 1175 | has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would write: |
| 1176 | .RS |
| 1177 | .P |
| 1178 | bssplit=4k/10:64k/50:32k/40 |
| 1179 | .RE |
| 1180 | .P |
| 1181 | Ordering does not matter. If the percentage is left blank, fio will fill in |
| 1182 | the remaining values evenly. So a bssplit option like this one: |
| 1183 | .RS |
| 1184 | .P |
| 1185 | bssplit=4k/50:1k/:32k/ |
| 1186 | .RE |
| 1187 | .P |
| 1188 | would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always add up |
| 1189 | to 100, if bssplit is given a range that adds up to more, it will error out. |
| 1190 | .P |
| 1191 | Comma\-separated values may be specified for reads, writes, and trims as |
| 1192 | described in \fBblocksize\fR. |
| 1193 | .P |
| 1194 | If you want a workload that has 50% 2k reads and 50% 4k reads, while having |
| 1195 | 90% 4k writes and 10% 8k writes, you would specify: |
| 1196 | .RS |
| 1197 | .P |
| 1198 | bssplit=2k/50:4k/50,4k/90,8k/10 |
| 1199 | .RE |
| 1200 | .RE |
| 1201 | .TP |
| 1202 | .BI blocksize_unaligned "\fR,\fB bs_unaligned" |
| 1203 | If set, fio will issue I/O units with any size within |
| 1204 | \fBblocksize_range\fR, not just multiples of the minimum size. This |
| 1205 | typically won't work with direct I/O, as that normally requires sector |
| 1206 | alignment. |
| 1207 | .TP |
| 1208 | .BI bs_is_seq_rand \fR=\fPbool |
| 1209 | If this option is set, fio will use the normal read,write blocksize settings |
| 1210 | as sequential,random blocksize settings instead. Any random read or write |
| 1211 | will use the WRITE blocksize settings, and any sequential read or write will |
| 1212 | use the READ blocksize settings. |
| 1213 | .TP |
| 1214 | .BI blockalign \fR=\fPint[,int][,int] "\fR,\fB ba" \fR=\fPint[,int][,int] |
| 1215 | Boundary to which fio will align random I/O units. Default: |
| 1216 | \fBblocksize\fR. Minimum alignment is typically 512b for using direct |
| 1217 | I/O, though it usually depends on the hardware block size. This option is |
| 1218 | mutually exclusive with using a random map for files, so it will turn off |
| 1219 | that option. Comma\-separated values may be specified for reads, writes, and |
| 1220 | trims as described in \fBblocksize\fR. |
| 1221 | .SS "Buffers and memory" |
| 1222 | .TP |
| 1223 | .BI zero_buffers |
| 1224 | Initialize buffers with all zeros. Default: fill buffers with random data. |
| 1225 | .TP |
| 1226 | .BI refill_buffers |
| 1227 | If this option is given, fio will refill the I/O buffers on every |
| 1228 | submit. The default is to only fill it at init time and reuse that |
| 1229 | data. Only makes sense if zero_buffers isn't specified, naturally. If data |
| 1230 | verification is enabled, \fBrefill_buffers\fR is also automatically enabled. |
| 1231 | .TP |
| 1232 | .BI scramble_buffers \fR=\fPbool |
| 1233 | If \fBrefill_buffers\fR is too costly and the target is using data |
| 1234 | deduplication, then setting this option will slightly modify the I/O buffer |
| 1235 | contents to defeat normal de\-dupe attempts. This is not enough to defeat |
| 1236 | more clever block compression attempts, but it will stop naive dedupe of |
| 1237 | blocks. Default: true. |
| 1238 | .TP |
| 1239 | .BI buffer_compress_percentage \fR=\fPint |
| 1240 | If this is set, then fio will attempt to provide I/O buffer content (on |
| 1241 | WRITEs) that compresses to the specified level. Fio does this by providing a |
| 1242 | mix of random data and a fixed pattern. The fixed pattern is either zeros, |
| 1243 | or the pattern specified by \fBbuffer_pattern\fR. If the pattern option |
| 1244 | is used, it might skew the compression ratio slightly. Note that this is per |
| 1245 | block size unit, for file/disk wide compression level that matches this |
| 1246 | setting, you'll also want to set \fBrefill_buffers\fR. |
| 1247 | .TP |
| 1248 | .BI buffer_compress_chunk \fR=\fPint |
| 1249 | See \fBbuffer_compress_percentage\fR. This setting allows fio to manage |
| 1250 | how big the ranges of random data and zeroed data is. Without this set, fio |
| 1251 | will provide \fBbuffer_compress_percentage\fR of blocksize random data, |
| 1252 | followed by the remaining zeroed. With this set to some chunk size smaller |
| 1253 | than the block size, fio can alternate random and zeroed data throughout the |
| 1254 | I/O buffer. |
| 1255 | .TP |
| 1256 | .BI buffer_pattern \fR=\fPstr |
| 1257 | If set, fio will fill the I/O buffers with this pattern or with the contents |
| 1258 | of a file. If not set, the contents of I/O buffers are defined by the other |
| 1259 | options related to buffer contents. The setting can be any pattern of bytes, |
| 1260 | and can be prefixed with 0x for hex values. It may also be a string, where |
| 1261 | the string must then be wrapped with "". Or it may also be a filename, |
| 1262 | where the filename must be wrapped with '' in which case the file is |
| 1263 | opened and read. Note that not all the file contents will be read if that |
| 1264 | would cause the buffers to overflow. So, for example: |
| 1265 | .RS |
| 1266 | .RS |
| 1267 | .P |
| 1268 | .PD 0 |
| 1269 | buffer_pattern='filename' |
| 1270 | .P |
| 1271 | or: |
| 1272 | .P |
| 1273 | buffer_pattern="abcd" |
| 1274 | .P |
| 1275 | or: |
| 1276 | .P |
| 1277 | buffer_pattern=\-12 |
| 1278 | .P |
| 1279 | or: |
| 1280 | .P |
| 1281 | buffer_pattern=0xdeadface |
| 1282 | .PD |
| 1283 | .RE |
| 1284 | .P |
| 1285 | Also you can combine everything together in any order: |
| 1286 | .RS |
| 1287 | .P |
| 1288 | buffer_pattern=0xdeadface"abcd"\-12'filename' |
| 1289 | .RE |
| 1290 | .RE |
| 1291 | .TP |
| 1292 | .BI dedupe_percentage \fR=\fPint |
| 1293 | If set, fio will generate this percentage of identical buffers when |
| 1294 | writing. These buffers will be naturally dedupable. The contents of the |
| 1295 | buffers depend on what other buffer compression settings have been set. It's |
| 1296 | possible to have the individual buffers either fully compressible, or not at |
| 1297 | all. This option only controls the distribution of unique buffers. |
| 1298 | .TP |
| 1299 | .BI invalidate \fR=\fPbool |
| 1300 | Invalidate the buffer/page cache parts of the files to be used prior to |
| 1301 | starting I/O if the platform and file type support it. Defaults to true. |
| 1302 | This will be ignored if \fBpre_read\fR is also specified for the |
| 1303 | same job. |
| 1304 | .TP |
| 1305 | .BI sync \fR=\fPbool |
| 1306 | Use synchronous I/O for buffered writes. For the majority of I/O engines, |
| 1307 | this means using O_SYNC. Default: false. |
| 1308 | .TP |
| 1309 | .BI iomem \fR=\fPstr "\fR,\fP mem" \fR=\fPstr |
| 1310 | Fio can use various types of memory as the I/O unit buffer. The allowed |
| 1311 | values are: |
| 1312 | .RS |
| 1313 | .RS |
| 1314 | .TP |
| 1315 | .B malloc |
| 1316 | Use memory from \fBmalloc\fR\|(3) as the buffers. Default memory type. |
| 1317 | .TP |
| 1318 | .B shm |
| 1319 | Use shared memory as the buffers. Allocated through \fBshmget\fR\|(2). |
| 1320 | .TP |
| 1321 | .B shmhuge |
| 1322 | Same as \fBshm\fR, but use huge pages as backing. |
| 1323 | .TP |
| 1324 | .B mmap |
| 1325 | Use \fBmmap\fR\|(2) to allocate buffers. May either be anonymous memory, or can |
| 1326 | be file backed if a filename is given after the option. The format |
| 1327 | is `mem=mmap:/path/to/file'. |
| 1328 | .TP |
| 1329 | .B mmaphuge |
| 1330 | Use a memory mapped huge file as the buffer backing. Append filename |
| 1331 | after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file'. |
| 1332 | .TP |
| 1333 | .B mmapshared |
| 1334 | Same as \fBmmap\fR, but use a MMAP_SHARED mapping. |
| 1335 | .TP |
| 1336 | .B cudamalloc |
| 1337 | Use GPU memory as the buffers for GPUDirect RDMA benchmark. |
| 1338 | The \fBioengine\fR must be \fBrdma\fR. |
| 1339 | .RE |
| 1340 | .P |
| 1341 | The area allocated is a function of the maximum allowed bs size for the job, |
| 1342 | multiplied by the I/O depth given. Note that for \fBshmhuge\fR and |
| 1343 | \fBmmaphuge\fR to work, the system must have free huge pages allocated. This |
| 1344 | can normally be checked and set by reading/writing |
| 1345 | `/proc/sys/vm/nr_hugepages' on a Linux system. Fio assumes a huge page |
| 1346 | is 4MiB in size. So to calculate the number of huge pages you need for a |
| 1347 | given job file, add up the I/O depth of all jobs (normally one unless |
| 1348 | \fBiodepth\fR is used) and multiply by the maximum bs set. Then divide |
| 1349 | that number by the huge page size. You can see the size of the huge pages in |
| 1350 | `/proc/meminfo'. If no huge pages are allocated by having a non\-zero |
| 1351 | number in `nr_hugepages', using \fBmmaphuge\fR or \fBshmhuge\fR will fail. Also |
| 1352 | see \fBhugepage\-size\fR. |
| 1353 | .P |
| 1354 | \fBmmaphuge\fR also needs to have hugetlbfs mounted and the file location |
| 1355 | should point there. So if it's mounted in `/huge', you would use |
| 1356 | `mem=mmaphuge:/huge/somefile'. |
| 1357 | .RE |
| 1358 | .TP |
| 1359 | .BI iomem_align \fR=\fPint "\fR,\fP mem_align" \fR=\fPint |
| 1360 | This indicates the memory alignment of the I/O memory buffers. Note that |
| 1361 | the given alignment is applied to the first I/O unit buffer, if using |
| 1362 | \fBiodepth\fR the alignment of the following buffers are given by the |
| 1363 | \fBbs\fR used. In other words, if using a \fBbs\fR that is a |
| 1364 | multiple of the page sized in the system, all buffers will be aligned to |
| 1365 | this value. If using a \fBbs\fR that is not page aligned, the alignment |
| 1366 | of subsequent I/O memory buffers is the sum of the \fBiomem_align\fR and |
| 1367 | \fBbs\fR used. |
| 1368 | .TP |
| 1369 | .BI hugepage\-size \fR=\fPint |
| 1370 | Defines the size of a huge page. Must at least be equal to the system |
| 1371 | setting, see `/proc/meminfo'. Defaults to 4MiB. Should probably |
| 1372 | always be a multiple of megabytes, so using `hugepage\-size=Xm' is the |
| 1373 | preferred way to set this to avoid setting a non\-pow\-2 bad value. |
| 1374 | .TP |
| 1375 | .BI lockmem \fR=\fPint |
| 1376 | Pin the specified amount of memory with \fBmlock\fR\|(2). Can be used to |
| 1377 | simulate a smaller amount of memory. The amount specified is per worker. |
| 1378 | .SS "I/O size" |
| 1379 | .TP |
| 1380 | .BI size \fR=\fPint |
| 1381 | The total size of file I/O for each thread of this job. Fio will run until |
| 1382 | this many bytes has been transferred, unless runtime is limited by other options |
| 1383 | (such as \fBruntime\fR, for instance, or increased/decreased by \fBio_size\fR). |
| 1384 | Fio will divide this size between the available files determined by options |
| 1385 | such as \fBnrfiles\fR, \fBfilename\fR, unless \fBfilesize\fR is |
| 1386 | specified by the job. If the result of division happens to be 0, the size is |
| 1387 | set to the physical size of the given files or devices if they exist. |
| 1388 | If this option is not specified, fio will use the full size of the given |
| 1389 | files or devices. If the files do not exist, size must be given. It is also |
| 1390 | possible to give size as a percentage between 1 and 100. If `size=20%' is |
| 1391 | given, fio will use 20% of the full size of the given files or devices. |
| 1392 | Can be combined with \fBoffset\fR to constrain the start and end range |
| 1393 | that I/O will be done within. |
| 1394 | .TP |
| 1395 | .BI io_size \fR=\fPint "\fR,\fB io_limit" \fR=\fPint |
| 1396 | Normally fio operates within the region set by \fBsize\fR, which means |
| 1397 | that the \fBsize\fR option sets both the region and size of I/O to be |
| 1398 | performed. Sometimes that is not what you want. With this option, it is |
| 1399 | possible to define just the amount of I/O that fio should do. For instance, |
| 1400 | if \fBsize\fR is set to 20GiB and \fBio_size\fR is set to 5GiB, fio |
| 1401 | will perform I/O within the first 20GiB but exit when 5GiB have been |
| 1402 | done. The opposite is also possible \-\- if \fBsize\fR is set to 20GiB, |
| 1403 | and \fBio_size\fR is set to 40GiB, then fio will do 40GiB of I/O within |
| 1404 | the 0..20GiB region. |
| 1405 | .TP |
| 1406 | .BI filesize \fR=\fPirange(int) |
| 1407 | Individual file sizes. May be a range, in which case fio will select sizes |
| 1408 | for files at random within the given range and limited to \fBsize\fR in |
| 1409 | total (if that is given). If not given, each created file is the same size. |
| 1410 | This option overrides \fBsize\fR in terms of file size, which means |
| 1411 | this value is used as a fixed size or possible range of each file. |
| 1412 | .TP |
| 1413 | .BI file_append \fR=\fPbool |
| 1414 | Perform I/O after the end of the file. Normally fio will operate within the |
| 1415 | size of a file. If this option is set, then fio will append to the file |
| 1416 | instead. This has identical behavior to setting \fBoffset\fR to the size |
| 1417 | of a file. This option is ignored on non\-regular files. |
| 1418 | .TP |
| 1419 | .BI fill_device \fR=\fPbool "\fR,\fB fill_fs" \fR=\fPbool |
| 1420 | Sets size to something really large and waits for ENOSPC (no space left on |
| 1421 | device) as the terminating condition. Only makes sense with sequential |
| 1422 | write. For a read workload, the mount point will be filled first then I/O |
| 1423 | started on the result. This option doesn't make sense if operating on a raw |
| 1424 | device node, since the size of that is already known by the file system. |
| 1425 | Additionally, writing beyond end\-of\-device will not return ENOSPC there. |
| 1426 | .SS "I/O engine" |
| 1427 | .TP |
| 1428 | .BI ioengine \fR=\fPstr |
| 1429 | Defines how the job issues I/O to the file. The following types are defined: |
| 1430 | .RS |
| 1431 | .RS |
| 1432 | .TP |
| 1433 | .B sync |
| 1434 | Basic \fBread\fR\|(2) or \fBwrite\fR\|(2) |
| 1435 | I/O. \fBlseek\fR\|(2) is used to position the I/O location. |
| 1436 | See \fBfsync\fR and \fBfdatasync\fR for syncing write I/Os. |
| 1437 | .TP |
| 1438 | .B psync |
| 1439 | Basic \fBpread\fR\|(2) or \fBpwrite\fR\|(2) I/O. Default on |
| 1440 | all supported operating systems except for Windows. |
| 1441 | .TP |
| 1442 | .B vsync |
| 1443 | Basic \fBreadv\fR\|(2) or \fBwritev\fR\|(2) I/O. Will emulate |
| 1444 | queuing by coalescing adjacent I/Os into a single submission. |
| 1445 | .TP |
| 1446 | .B pvsync |
| 1447 | Basic \fBpreadv\fR\|(2) or \fBpwritev\fR\|(2) I/O. |
| 1448 | .TP |
| 1449 | .B pvsync2 |
| 1450 | Basic \fBpreadv2\fR\|(2) or \fBpwritev2\fR\|(2) I/O. |
| 1451 | .TP |
| 1452 | .B libaio |
| 1453 | Linux native asynchronous I/O. Note that Linux may only support |
| 1454 | queued behavior with non\-buffered I/O (set `direct=1' or |
| 1455 | `buffered=0'). |
| 1456 | This engine defines engine specific options. |
| 1457 | .TP |
| 1458 | .B posixaio |
| 1459 | POSIX asynchronous I/O using \fBaio_read\fR\|(3) and |
| 1460 | \fBaio_write\fR\|(3). |
| 1461 | .TP |
| 1462 | .B solarisaio |
| 1463 | Solaris native asynchronous I/O. |
| 1464 | .TP |
| 1465 | .B windowsaio |
| 1466 | Windows native asynchronous I/O. Default on Windows. |
| 1467 | .TP |
| 1468 | .B mmap |
| 1469 | File is memory mapped with \fBmmap\fR\|(2) and data copied |
| 1470 | to/from using \fBmemcpy\fR\|(3). |
| 1471 | .TP |
| 1472 | .B splice |
| 1473 | \fBsplice\fR\|(2) is used to transfer the data and |
| 1474 | \fBvmsplice\fR\|(2) to transfer data from user space to the |
| 1475 | kernel. |
| 1476 | .TP |
| 1477 | .B sg |
| 1478 | SCSI generic sg v3 I/O. May either be synchronous using the SG_IO |
| 1479 | ioctl, or if the target is an sg character device we use |
| 1480 | \fBread\fR\|(2) and \fBwrite\fR\|(2) for asynchronous |
| 1481 | I/O. Requires \fBfilename\fR option to specify either block or |
| 1482 | character devices. |
| 1483 | .TP |
| 1484 | .B null |
| 1485 | Doesn't transfer any data, just pretends to. This is mainly used to |
| 1486 | exercise fio itself and for debugging/testing purposes. |
| 1487 | .TP |
| 1488 | .B net |
| 1489 | Transfer over the network to given `host:port'. Depending on the |
| 1490 | \fBprotocol\fR used, the \fBhostname\fR, \fBport\fR, |
| 1491 | \fBlisten\fR and \fBfilename\fR options are used to specify |
| 1492 | what sort of connection to make, while the \fBprotocol\fR option |
| 1493 | determines which protocol will be used. This engine defines engine |
| 1494 | specific options. |
| 1495 | .TP |
| 1496 | .B netsplice |
| 1497 | Like \fBnet\fR, but uses \fBsplice\fR\|(2) and |
| 1498 | \fBvmsplice\fR\|(2) to map data and send/receive. |
| 1499 | This engine defines engine specific options. |
| 1500 | .TP |
| 1501 | .B cpuio |
| 1502 | Doesn't transfer any data, but burns CPU cycles according to the |
| 1503 | \fBcpuload\fR and \fBcpuchunks\fR options. Setting |
| 1504 | \fBcpuload\fR\=85 will cause that job to do nothing but burn 85% |
| 1505 | of the CPU. In case of SMP machines, use `numjobs=<nr_of_cpu>' |
| 1506 | to get desired CPU usage, as the cpuload only loads a |
| 1507 | single CPU at the desired rate. A job never finishes unless there is |
| 1508 | at least one non\-cpuio job. |
| 1509 | .TP |
| 1510 | .B guasi |
| 1511 | The GUASI I/O engine is the Generic Userspace Asyncronous Syscall |
| 1512 | Interface approach to async I/O. See \fIhttp://www.xmailserver.org/guasi\-lib.html\fR |
| 1513 | for more info on GUASI. |
| 1514 | .TP |
| 1515 | .B rdma |
| 1516 | The RDMA I/O engine supports both RDMA memory semantics |
| 1517 | (RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the |
| 1518 | InfiniBand, RoCE and iWARP protocols. |
| 1519 | .TP |
| 1520 | .B falloc |
| 1521 | I/O engine that does regular fallocate to simulate data transfer as |
| 1522 | fio ioengine. |
| 1523 | .RS |
| 1524 | .P |
| 1525 | .PD 0 |
| 1526 | DDIR_READ does fallocate(,mode = FALLOC_FL_KEEP_SIZE,). |
| 1527 | .P |
| 1528 | DIR_WRITE does fallocate(,mode = 0). |
| 1529 | .P |
| 1530 | DDIR_TRIM does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE). |
| 1531 | .PD |
| 1532 | .RE |
| 1533 | .TP |
| 1534 | .B ftruncate |
| 1535 | I/O engine that sends \fBftruncate\fR\|(2) operations in response |
| 1536 | to write (DDIR_WRITE) events. Each ftruncate issued sets the file's |
| 1537 | size to the current block offset. \fBblocksize\fR is ignored. |
| 1538 | .TP |
| 1539 | .B e4defrag |
| 1540 | I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate |
| 1541 | defragment activity in request to DDIR_WRITE event. |
| 1542 | .TP |
| 1543 | .B rbd |
| 1544 | I/O engine supporting direct access to Ceph Rados Block Devices |
| 1545 | (RBD) via librbd without the need to use the kernel rbd driver. This |
| 1546 | ioengine defines engine specific options. |
| 1547 | .TP |
| 1548 | .B gfapi |
| 1549 | Using GlusterFS libgfapi sync interface to direct access to |
| 1550 | GlusterFS volumes without having to go through FUSE. This ioengine |
| 1551 | defines engine specific options. |
| 1552 | .TP |
| 1553 | .B gfapi_async |
| 1554 | Using GlusterFS libgfapi async interface to direct access to |
| 1555 | GlusterFS volumes without having to go through FUSE. This ioengine |
| 1556 | defines engine specific options. |
| 1557 | .TP |
| 1558 | .B libhdfs |
| 1559 | Read and write through Hadoop (HDFS). The \fBfilename\fR option |
| 1560 | is used to specify host,port of the hdfs name\-node to connect. This |
| 1561 | engine interprets offsets a little differently. In HDFS, files once |
| 1562 | created cannot be modified so random writes are not possible. To |
| 1563 | imitate this the libhdfs engine expects a bunch of small files to be |
| 1564 | created over HDFS and will randomly pick a file from them |
| 1565 | based on the offset generated by fio backend (see the example |
| 1566 | job file to create such files, use `rw=write' option). Please |
| 1567 | note, it may be necessary to set environment variables to work |
| 1568 | with HDFS/libhdfs properly. Each job uses its own connection to |
| 1569 | HDFS. |
| 1570 | .TP |
| 1571 | .B mtd |
| 1572 | Read, write and erase an MTD character device (e.g., |
| 1573 | `/dev/mtd0'). Discards are treated as erases. Depending on the |
| 1574 | underlying device type, the I/O may have to go in a certain pattern, |
| 1575 | e.g., on NAND, writing sequentially to erase blocks and discarding |
| 1576 | before overwriting. The \fBtrimwrite\fR mode works well for this |
| 1577 | constraint. |
| 1578 | .TP |
| 1579 | .B pmemblk |
| 1580 | Read and write using filesystem DAX to a file on a filesystem |
| 1581 | mounted with DAX on a persistent memory device through the NVML |
| 1582 | libpmemblk library. |
| 1583 | .TP |
| 1584 | .B dev\-dax |
| 1585 | Read and write using device DAX to a persistent memory device (e.g., |
| 1586 | /dev/dax0.0) through the NVML libpmem library. |
| 1587 | .TP |
| 1588 | .B external |
| 1589 | Prefix to specify loading an external I/O engine object file. Append |
| 1590 | the engine filename, e.g. `ioengine=external:/tmp/foo.o' to load |
| 1591 | ioengine `foo.o' in `/tmp'. The path can be either |
| 1592 | absolute or relative. See `engines/skeleton_external.c' in the fio source for |
| 1593 | details of writing an external I/O engine. |
| 1594 | .TP |
| 1595 | .B filecreate |
| 1596 | Create empty files only. \fBfilesize\fR still needs to be specified so that fio |
| 1597 | will run and grab latency results, but no IO will actually be done on the files. |
| 1598 | .SS "I/O engine specific parameters" |
| 1599 | In addition, there are some parameters which are only valid when a specific |
| 1600 | \fBioengine\fR is in use. These are used identically to normal parameters, |
| 1601 | with the caveat that when used on the command line, they must come after the |
| 1602 | \fBioengine\fR that defines them is selected. |
| 1603 | .TP |
| 1604 | .BI (libaio)userspace_reap |
| 1605 | Normally, with the libaio engine in use, fio will use the |
| 1606 | \fBio_getevents\fR\|(3) system call to reap newly returned events. With |
| 1607 | this flag turned on, the AIO ring will be read directly from user\-space to |
| 1608 | reap events. The reaping mode is only enabled when polling for a minimum of |
| 1609 | 0 events (e.g. when `iodepth_batch_complete=0'). |
| 1610 | .TP |
| 1611 | .BI (pvsync2)hipri |
| 1612 | Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority |
| 1613 | than normal. |
| 1614 | .TP |
| 1615 | .BI (pvsync2)hipri_percentage |
| 1616 | When hipri is set this determines the probability of a pvsync2 I/O being high |
| 1617 | priority. The default is 100%. |
| 1618 | .TP |
| 1619 | .BI (cpuio)cpuload \fR=\fPint |
| 1620 | Attempt to use the specified percentage of CPU cycles. This is a mandatory |
| 1621 | option when using cpuio I/O engine. |
| 1622 | .TP |
| 1623 | .BI (cpuio)cpuchunks \fR=\fPint |
| 1624 | Split the load into cycles of the given time. In microseconds. |
| 1625 | .TP |
| 1626 | .BI (cpuio)exit_on_io_done \fR=\fPbool |
| 1627 | Detect when I/O threads are done, then exit. |
| 1628 | .TP |
| 1629 | .BI (libhdfs)namenode \fR=\fPstr |
| 1630 | The hostname or IP address of a HDFS cluster namenode to contact. |
| 1631 | .TP |
| 1632 | .BI (libhdfs)port |
| 1633 | The listening port of the HFDS cluster namenode. |
| 1634 | .TP |
| 1635 | .BI (netsplice,net)port |
| 1636 | The TCP or UDP port to bind to or connect to. If this is used with |
| 1637 | \fBnumjobs\fR to spawn multiple instances of the same job type, then |
| 1638 | this will be the starting port number since fio will use a range of |
| 1639 | ports. |
| 1640 | .TP |
| 1641 | .BI (netsplice,net)hostname \fR=\fPstr |
| 1642 | The hostname or IP address to use for TCP or UDP based I/O. If the job is |
| 1643 | a TCP listener or UDP reader, the hostname is not used and must be omitted |
| 1644 | unless it is a valid UDP multicast address. |
| 1645 | .TP |
| 1646 | .BI (netsplice,net)interface \fR=\fPstr |
| 1647 | The IP address of the network interface used to send or receive UDP |
| 1648 | multicast. |
| 1649 | .TP |
| 1650 | .BI (netsplice,net)ttl \fR=\fPint |
| 1651 | Time\-to\-live value for outgoing UDP multicast packets. Default: 1. |
| 1652 | .TP |
| 1653 | .BI (netsplice,net)nodelay \fR=\fPbool |
| 1654 | Set TCP_NODELAY on TCP connections. |
| 1655 | .TP |
| 1656 | .BI (netsplice,net)protocol \fR=\fPstr "\fR,\fP proto" \fR=\fPstr |
| 1657 | The network protocol to use. Accepted values are: |
| 1658 | .RS |
| 1659 | .RS |
| 1660 | .TP |
| 1661 | .B tcp |
| 1662 | Transmission control protocol. |
| 1663 | .TP |
| 1664 | .B tcpv6 |
| 1665 | Transmission control protocol V6. |
| 1666 | .TP |
| 1667 | .B udp |
| 1668 | User datagram protocol. |
| 1669 | .TP |
| 1670 | .B udpv6 |
| 1671 | User datagram protocol V6. |
| 1672 | .TP |
| 1673 | .B unix |
| 1674 | UNIX domain socket. |
| 1675 | .RE |
| 1676 | .P |
| 1677 | When the protocol is TCP or UDP, the port must also be given, as well as the |
| 1678 | hostname if the job is a TCP listener or UDP reader. For unix sockets, the |
| 1679 | normal \fBfilename\fR option should be used and the port is invalid. |
| 1680 | .RE |
| 1681 | .TP |
| 1682 | .BI (netsplice,net)listen |
| 1683 | For TCP network connections, tell fio to listen for incoming connections |
| 1684 | rather than initiating an outgoing connection. The \fBhostname\fR must |
| 1685 | be omitted if this option is used. |
| 1686 | .TP |
| 1687 | .BI (netsplice,net)pingpong |
| 1688 | Normally a network writer will just continue writing data, and a network |
| 1689 | reader will just consume packages. If `pingpong=1' is set, a writer will |
| 1690 | send its normal payload to the reader, then wait for the reader to send the |
| 1691 | same payload back. This allows fio to measure network latencies. The |
| 1692 | submission and completion latencies then measure local time spent sending or |
| 1693 | receiving, and the completion latency measures how long it took for the |
| 1694 | other end to receive and send back. For UDP multicast traffic |
| 1695 | `pingpong=1' should only be set for a single reader when multiple readers |
| 1696 | are listening to the same address. |
| 1697 | .TP |
| 1698 | .BI (netsplice,net)window_size \fR=\fPint |
| 1699 | Set the desired socket buffer size for the connection. |
| 1700 | .TP |
| 1701 | .BI (netsplice,net)mss \fR=\fPint |
| 1702 | Set the TCP maximum segment size (TCP_MAXSEG). |
| 1703 | .TP |
| 1704 | .BI (e4defrag)donorname \fR=\fPstr |
| 1705 | File will be used as a block donor (swap extents between files). |
| 1706 | .TP |
| 1707 | .BI (e4defrag)inplace \fR=\fPint |
| 1708 | Configure donor file blocks allocation strategy: |
| 1709 | .RS |
| 1710 | .RS |
| 1711 | .TP |
| 1712 | .B 0 |
| 1713 | Default. Preallocate donor's file on init. |
| 1714 | .TP |
| 1715 | .B 1 |
| 1716 | Allocate space immediately inside defragment event, and free right |
| 1717 | after event. |
| 1718 | .RE |
| 1719 | .RE |
| 1720 | .TP |
| 1721 | .BI (rbd)clustername \fR=\fPstr |
| 1722 | Specifies the name of the Ceph cluster. |
| 1723 | .TP |
| 1724 | .BI (rbd)rbdname \fR=\fPstr |
| 1725 | Specifies the name of the RBD. |
| 1726 | .TP |
| 1727 | .BI (rbd)pool \fR=\fPstr |
| 1728 | Specifies the name of the Ceph pool containing RBD. |
| 1729 | .TP |
| 1730 | .BI (rbd)clientname \fR=\fPstr |
| 1731 | Specifies the username (without the 'client.' prefix) used to access the |
| 1732 | Ceph cluster. If the \fBclustername\fR is specified, the \fBclientname\fR shall be |
| 1733 | the full *type.id* string. If no type. prefix is given, fio will add 'client.' |
| 1734 | by default. |
| 1735 | .TP |
| 1736 | .BI (mtd)skip_bad \fR=\fPbool |
| 1737 | Skip operations against known bad blocks. |
| 1738 | .TP |
| 1739 | .BI (libhdfs)hdfsdirectory |
| 1740 | libhdfs will create chunk in this HDFS directory. |
| 1741 | .TP |
| 1742 | .BI (libhdfs)chunk_size |
| 1743 | The size of the chunk to use for each file. |
| 1744 | .SS "I/O depth" |
| 1745 | .TP |
| 1746 | .BI iodepth \fR=\fPint |
| 1747 | Number of I/O units to keep in flight against the file. Note that |
| 1748 | increasing \fBiodepth\fR beyond 1 will not affect synchronous ioengines (except |
| 1749 | for small degrees when \fBverify_async\fR is in use). Even async |
| 1750 | engines may impose OS restrictions causing the desired depth not to be |
| 1751 | achieved. This may happen on Linux when using libaio and not setting |
| 1752 | `direct=1', since buffered I/O is not async on that OS. Keep an |
| 1753 | eye on the I/O depth distribution in the fio output to verify that the |
| 1754 | achieved depth is as expected. Default: 1. |
| 1755 | .TP |
| 1756 | .BI iodepth_batch_submit \fR=\fPint "\fR,\fP iodepth_batch" \fR=\fPint |
| 1757 | This defines how many pieces of I/O to submit at once. It defaults to 1 |
| 1758 | which means that we submit each I/O as soon as it is available, but can be |
| 1759 | raised to submit bigger batches of I/O at the time. If it is set to 0 the |
| 1760 | \fBiodepth\fR value will be used. |
| 1761 | .TP |
| 1762 | .BI iodepth_batch_complete_min \fR=\fPint "\fR,\fP iodepth_batch_complete" \fR=\fPint |
| 1763 | This defines how many pieces of I/O to retrieve at once. It defaults to 1 |
| 1764 | which means that we'll ask for a minimum of 1 I/O in the retrieval process |
| 1765 | from the kernel. The I/O retrieval will go on until we hit the limit set by |
| 1766 | \fBiodepth_low\fR. If this variable is set to 0, then fio will always |
| 1767 | check for completed events before queuing more I/O. This helps reduce I/O |
| 1768 | latency, at the cost of more retrieval system calls. |
| 1769 | .TP |
| 1770 | .BI iodepth_batch_complete_max \fR=\fPint |
| 1771 | This defines maximum pieces of I/O to retrieve at once. This variable should |
| 1772 | be used along with \fBiodepth_batch_complete_min\fR=\fIint\fR variable, |
| 1773 | specifying the range of min and max amount of I/O which should be |
| 1774 | retrieved. By default it is equal to \fBiodepth_batch_complete_min\fR |
| 1775 | value. Example #1: |
| 1776 | .RS |
| 1777 | .RS |
| 1778 | .P |
| 1779 | .PD 0 |
| 1780 | iodepth_batch_complete_min=1 |
| 1781 | .P |
| 1782 | iodepth_batch_complete_max=<iodepth> |
| 1783 | .PD |
| 1784 | .RE |
| 1785 | .P |
| 1786 | which means that we will retrieve at least 1 I/O and up to the whole |
| 1787 | submitted queue depth. If none of I/O has been completed yet, we will wait. |
| 1788 | Example #2: |
| 1789 | .RS |
| 1790 | .P |
| 1791 | .PD 0 |
| 1792 | iodepth_batch_complete_min=0 |
| 1793 | .P |
| 1794 | iodepth_batch_complete_max=<iodepth> |
| 1795 | .PD |
| 1796 | .RE |
| 1797 | .P |
| 1798 | which means that we can retrieve up to the whole submitted queue depth, but |
| 1799 | if none of I/O has been completed yet, we will NOT wait and immediately exit |
| 1800 | the system call. In this example we simply do polling. |
| 1801 | .RE |
| 1802 | .TP |
| 1803 | .BI iodepth_low \fR=\fPint |
| 1804 | The low water mark indicating when to start filling the queue |
| 1805 | again. Defaults to the same as \fBiodepth\fR, meaning that fio will |
| 1806 | attempt to keep the queue full at all times. If \fBiodepth\fR is set to |
| 1807 | e.g. 16 and \fBiodepth_low\fR is set to 4, then after fio has filled the queue of |
| 1808 | 16 requests, it will let the depth drain down to 4 before starting to fill |
| 1809 | it again. |
| 1810 | .TP |
| 1811 | .BI serialize_overlap \fR=\fPbool |
| 1812 | Serialize in-flight I/Os that might otherwise cause or suffer from data races. |
| 1813 | When two or more I/Os are submitted simultaneously, there is no guarantee that |
| 1814 | the I/Os will be processed or completed in the submitted order. Further, if |
| 1815 | two or more of those I/Os are writes, any overlapping region between them can |
| 1816 | become indeterminate/undefined on certain storage. These issues can cause |
| 1817 | verification to fail erratically when at least one of the racing I/Os is |
| 1818 | changing data and the overlapping region has a non-zero size. Setting |
| 1819 | \fBserialize_overlap\fR tells fio to avoid provoking this behavior by explicitly |
| 1820 | serializing in-flight I/Os that have a non-zero overlap. Note that setting |
| 1821 | this option can reduce both performance and the \fBiodepth\fR achieved. |
| 1822 | Additionally this option does not work when \fBio_submit_mode\fR is set to |
| 1823 | offload. Default: false. |
| 1824 | .TP |
| 1825 | .BI io_submit_mode \fR=\fPstr |
| 1826 | This option controls how fio submits the I/O to the I/O engine. The default |
| 1827 | is `inline', which means that the fio job threads submit and reap I/O |
| 1828 | directly. If set to `offload', the job threads will offload I/O submission |
| 1829 | to a dedicated pool of I/O threads. This requires some coordination and thus |
| 1830 | has a bit of extra overhead, especially for lower queue depth I/O where it |
| 1831 | can increase latencies. The benefit is that fio can manage submission rates |
| 1832 | independently of the device completion rates. This avoids skewed latency |
| 1833 | reporting if I/O gets backed up on the device side (the coordinated omission |
| 1834 | problem). |
| 1835 | .SS "I/O rate" |
| 1836 | .TP |
| 1837 | .BI thinktime \fR=\fPtime |
| 1838 | Stall the job for the specified period of time after an I/O has completed before issuing the |
| 1839 | next. May be used to simulate processing being done by an application. |
| 1840 | When the unit is omitted, the value is interpreted in microseconds. See |
| 1841 | \fBthinktime_blocks\fR and \fBthinktime_spin\fR. |
| 1842 | .TP |
| 1843 | .BI thinktime_spin \fR=\fPtime |
| 1844 | Only valid if \fBthinktime\fR is set \- pretend to spend CPU time doing |
| 1845 | something with the data received, before falling back to sleeping for the |
| 1846 | rest of the period specified by \fBthinktime\fR. When the unit is |
| 1847 | omitted, the value is interpreted in microseconds. |
| 1848 | .TP |
| 1849 | .BI thinktime_blocks \fR=\fPint |
| 1850 | Only valid if \fBthinktime\fR is set \- control how many blocks to issue, |
| 1851 | before waiting \fBthinktime\fR usecs. If not set, defaults to 1 which will make |
| 1852 | fio wait \fBthinktime\fR usecs after every block. This effectively makes any |
| 1853 | queue depth setting redundant, since no more than 1 I/O will be queued |
| 1854 | before we have to complete it and do our \fBthinktime\fR. In other words, this |
| 1855 | setting effectively caps the queue depth if the latter is larger. |
| 1856 | .TP |
| 1857 | .BI rate \fR=\fPint[,int][,int] |
| 1858 | Cap the bandwidth used by this job. The number is in bytes/sec, the normal |
| 1859 | suffix rules apply. Comma\-separated values may be specified for reads, |
| 1860 | writes, and trims as described in \fBblocksize\fR. |
| 1861 | .RS |
| 1862 | .P |
| 1863 | For example, using `rate=1m,500k' would limit reads to 1MiB/sec and writes to |
| 1864 | 500KiB/sec. Capping only reads or writes can be done with `rate=,500k' or |
| 1865 | `rate=500k,' where the former will only limit writes (to 500KiB/sec) and the |
| 1866 | latter will only limit reads. |
| 1867 | .RE |
| 1868 | .TP |
| 1869 | .BI rate_min \fR=\fPint[,int][,int] |
| 1870 | Tell fio to do whatever it can to maintain at least this bandwidth. Failing |
| 1871 | to meet this requirement will cause the job to exit. Comma\-separated values |
| 1872 | may be specified for reads, writes, and trims as described in |
| 1873 | \fBblocksize\fR. |
| 1874 | .TP |
| 1875 | .BI rate_iops \fR=\fPint[,int][,int] |
| 1876 | Cap the bandwidth to this number of IOPS. Basically the same as |
| 1877 | \fBrate\fR, just specified independently of bandwidth. If the job is |
| 1878 | given a block size range instead of a fixed value, the smallest block size |
| 1879 | is used as the metric. Comma\-separated values may be specified for reads, |
| 1880 | writes, and trims as described in \fBblocksize\fR. |
| 1881 | .TP |
| 1882 | .BI rate_iops_min \fR=\fPint[,int][,int] |
| 1883 | If fio doesn't meet this rate of I/O, it will cause the job to exit. |
| 1884 | Comma\-separated values may be specified for reads, writes, and trims as |
| 1885 | described in \fBblocksize\fR. |
| 1886 | .TP |
| 1887 | .BI rate_process \fR=\fPstr |
| 1888 | This option controls how fio manages rated I/O submissions. The default is |
| 1889 | `linear', which submits I/O in a linear fashion with fixed delays between |
| 1890 | I/Os that gets adjusted based on I/O completion rates. If this is set to |
| 1891 | `poisson', fio will submit I/O based on a more real world random request |
| 1892 | flow, known as the Poisson process |
| 1893 | (\fIhttps://en.wikipedia.org/wiki/Poisson_point_process\fR). The lambda will be |
| 1894 | 10^6 / IOPS for the given workload. |
| 1895 | .SS "I/O latency" |
| 1896 | .TP |
| 1897 | .BI latency_target \fR=\fPtime |
| 1898 | If set, fio will attempt to find the max performance point that the given |
| 1899 | workload will run at while maintaining a latency below this target. When |
| 1900 | the unit is omitted, the value is interpreted in microseconds. See |
| 1901 | \fBlatency_window\fR and \fBlatency_percentile\fR. |
| 1902 | .TP |
| 1903 | .BI latency_window \fR=\fPtime |
| 1904 | Used with \fBlatency_target\fR to specify the sample window that the job |
| 1905 | is run at varying queue depths to test the performance. When the unit is |
| 1906 | omitted, the value is interpreted in microseconds. |
| 1907 | .TP |
| 1908 | .BI latency_percentile \fR=\fPfloat |
| 1909 | The percentage of I/Os that must fall within the criteria specified by |
| 1910 | \fBlatency_target\fR and \fBlatency_window\fR. If not set, this |
| 1911 | defaults to 100.0, meaning that all I/Os must be equal or below to the value |
| 1912 | set by \fBlatency_target\fR. |
| 1913 | .TP |
| 1914 | .BI max_latency \fR=\fPtime |
| 1915 | If set, fio will exit the job with an ETIMEDOUT error if it exceeds this |
| 1916 | maximum latency. When the unit is omitted, the value is interpreted in |
| 1917 | microseconds. |
| 1918 | .TP |
| 1919 | .BI rate_cycle \fR=\fPint |
| 1920 | Average bandwidth for \fBrate\fR and \fBrate_min\fR over this number |
| 1921 | of milliseconds. Defaults to 1000. |
| 1922 | .SS "I/O replay" |
| 1923 | .TP |
| 1924 | .BI write_iolog \fR=\fPstr |
| 1925 | Write the issued I/O patterns to the specified file. See |
| 1926 | \fBread_iolog\fR. Specify a separate file for each job, otherwise the |
| 1927 | iologs will be interspersed and the file may be corrupt. |
| 1928 | .TP |
| 1929 | .BI read_iolog \fR=\fPstr |
| 1930 | Open an iolog with the specified filename and replay the I/O patterns it |
| 1931 | contains. This can be used to store a workload and replay it sometime |
| 1932 | later. The iolog given may also be a blktrace binary file, which allows fio |
| 1933 | to replay a workload captured by blktrace. See |
| 1934 | \fBblktrace\fR\|(8) for how to capture such logging data. For blktrace |
| 1935 | replay, the file needs to be turned into a blkparse binary data file first |
| 1936 | (`blkparse <device> \-o /dev/null \-d file_for_fio.bin'). |
| 1937 | .TP |
| 1938 | .BI replay_no_stall \fR=\fPbool |
| 1939 | When replaying I/O with \fBread_iolog\fR the default behavior is to |
| 1940 | attempt to respect the timestamps within the log and replay them with the |
| 1941 | appropriate delay between IOPS. By setting this variable fio will not |
| 1942 | respect the timestamps and attempt to replay them as fast as possible while |
| 1943 | still respecting ordering. The result is the same I/O pattern to a given |
| 1944 | device, but different timings. |
| 1945 | .TP |
| 1946 | .BI replay_redirect \fR=\fPstr |
| 1947 | While replaying I/O patterns using \fBread_iolog\fR the default behavior |
| 1948 | is to replay the IOPS onto the major/minor device that each IOP was recorded |
| 1949 | from. This is sometimes undesirable because on a different machine those |
| 1950 | major/minor numbers can map to a different device. Changing hardware on the |
| 1951 | same system can also result in a different major/minor mapping. |
| 1952 | \fBreplay_redirect\fR causes all I/Os to be replayed onto the single specified |
| 1953 | device regardless of the device it was recorded |
| 1954 | from. i.e. `replay_redirect=/dev/sdc' would cause all I/O |
| 1955 | in the blktrace or iolog to be replayed onto `/dev/sdc'. This means |
| 1956 | multiple devices will be replayed onto a single device, if the trace |
| 1957 | contains multiple devices. If you want multiple devices to be replayed |
| 1958 | concurrently to multiple redirected devices you must blkparse your trace |
| 1959 | into separate traces and replay them with independent fio invocations. |
| 1960 | Unfortunately this also breaks the strict time ordering between multiple |
| 1961 | device accesses. |
| 1962 | .TP |
| 1963 | .BI replay_align \fR=\fPint |
| 1964 | Force alignment of I/O offsets and lengths in a trace to this power of 2 |
| 1965 | value. |
| 1966 | .TP |
| 1967 | .BI replay_scale \fR=\fPint |
| 1968 | Scale sector offsets down by this factor when replaying traces. |
| 1969 | .SS "Threads, processes and job synchronization" |
| 1970 | .TP |
| 1971 | .BI thread |
| 1972 | Fio defaults to creating jobs by using fork, however if this option is |
| 1973 | given, fio will create jobs by using POSIX Threads' function |
| 1974 | \fBpthread_create\fR\|(3) to create threads instead. |
| 1975 | .TP |
| 1976 | .BI wait_for \fR=\fPstr |
| 1977 | If set, the current job won't be started until all workers of the specified |
| 1978 | waitee job are done. |
| 1979 | .\" ignore blank line here from HOWTO as it looks normal without it |
| 1980 | \fBwait_for\fR operates on the job name basis, so there are a few |
| 1981 | limitations. First, the waitee must be defined prior to the waiter job |
| 1982 | (meaning no forward references). Second, if a job is being referenced as a |
| 1983 | waitee, it must have a unique name (no duplicate waitees). |
| 1984 | .TP |
| 1985 | .BI nice \fR=\fPint |
| 1986 | Run the job with the given nice value. See man \fBnice\fR\|(2). |
| 1987 | .\" ignore blank line here from HOWTO as it looks normal without it |
| 1988 | On Windows, values less than \-15 set the process class to "High"; \-1 through |
| 1989 | \-15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle" |
| 1990 | priority class. |
| 1991 | .TP |
| 1992 | .BI prio \fR=\fPint |
| 1993 | Set the I/O priority value of this job. Linux limits us to a positive value |
| 1994 | between 0 and 7, with 0 being the highest. See man |
| 1995 | \fBionice\fR\|(1). Refer to an appropriate manpage for other operating |
| 1996 | systems since meaning of priority may differ. |
| 1997 | .TP |
| 1998 | .BI prioclass \fR=\fPint |
| 1999 | Set the I/O priority class. See man \fBionice\fR\|(1). |
| 2000 | .TP |
| 2001 | .BI cpumask \fR=\fPint |
| 2002 | Set the CPU affinity of this job. The parameter given is a bit mask of |
| 2003 | allowed CPUs the job may run on. So if you want the allowed CPUs to be 1 |
| 2004 | and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man |
| 2005 | \fBsched_setaffinity\fR\|(2). This may not work on all supported |
| 2006 | operating systems or kernel versions. This option doesn't work well for a |
| 2007 | higher CPU count than what you can store in an integer mask, so it can only |
| 2008 | control cpus 1\-32. For boxes with larger CPU counts, use |
| 2009 | \fBcpus_allowed\fR. |
| 2010 | .TP |
| 2011 | .BI cpus_allowed \fR=\fPstr |
| 2012 | Controls the same options as \fBcpumask\fR, but accepts a textual |
| 2013 | specification of the permitted CPUs instead. So to use CPUs 1 and 5 you |
| 2014 | would specify `cpus_allowed=1,5'. This option also allows a range of CPUs |
| 2015 | to be specified \-\- say you wanted a binding to CPUs 1, 5, and 8 to 15, you |
| 2016 | would set `cpus_allowed=1,5,8\-15'. |
| 2017 | .TP |
| 2018 | .BI cpus_allowed_policy \fR=\fPstr |
| 2019 | Set the policy of how fio distributes the CPUs specified by |
| 2020 | \fBcpus_allowed\fR or \fBcpumask\fR. Two policies are supported: |
| 2021 | .RS |
| 2022 | .RS |
| 2023 | .TP |
| 2024 | .B shared |
| 2025 | All jobs will share the CPU set specified. |
| 2026 | .TP |
| 2027 | .B split |
| 2028 | Each job will get a unique CPU from the CPU set. |
| 2029 | .RE |
| 2030 | .P |
| 2031 | \fBshared\fR is the default behavior, if the option isn't specified. If |
| 2032 | \fBsplit\fR is specified, then fio will will assign one cpu per job. If not |
| 2033 | enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs |
| 2034 | in the set. |
| 2035 | .RE |
| 2036 | .TP |
| 2037 | .BI numa_cpu_nodes \fR=\fPstr |
| 2038 | Set this job running on specified NUMA nodes' CPUs. The arguments allow |
| 2039 | comma delimited list of cpu numbers, A\-B ranges, or `all'. Note, to enable |
| 2040 | NUMA options support, fio must be built on a system with libnuma\-dev(el) |
| 2041 | installed. |
| 2042 | .TP |
| 2043 | .BI numa_mem_policy \fR=\fPstr |
| 2044 | Set this job's memory policy and corresponding NUMA nodes. Format of the |
| 2045 | arguments: |
| 2046 | .RS |
| 2047 | .RS |
| 2048 | .P |
| 2049 | <mode>[:<nodelist>] |
| 2050 | .RE |
| 2051 | .P |
| 2052 | `mode' is one of the following memory poicies: `default', `prefer', |
| 2053 | `bind', `interleave' or `local'. For `default' and `local' memory |
| 2054 | policies, no node needs to be specified. For `prefer', only one node is |
| 2055 | allowed. For `bind' and `interleave' the `nodelist' may be as |
| 2056 | follows: a comma delimited list of numbers, A\-B ranges, or `all'. |
| 2057 | .RE |
| 2058 | .TP |
| 2059 | .BI cgroup \fR=\fPstr |
| 2060 | Add job to this control group. If it doesn't exist, it will be created. The |
| 2061 | system must have a mounted cgroup blkio mount point for this to work. If |
| 2062 | your system doesn't have it mounted, you can do so with: |
| 2063 | .RS |
| 2064 | .RS |
| 2065 | .P |
| 2066 | # mount \-t cgroup \-o blkio none /cgroup |
| 2067 | .RE |
| 2068 | .RE |
| 2069 | .TP |
| 2070 | .BI cgroup_weight \fR=\fPint |
| 2071 | Set the weight of the cgroup to this value. See the documentation that comes |
| 2072 | with the kernel, allowed values are in the range of 100..1000. |
| 2073 | .TP |
| 2074 | .BI cgroup_nodelete \fR=\fPbool |
| 2075 | Normally fio will delete the cgroups it has created after the job |
| 2076 | completion. To override this behavior and to leave cgroups around after the |
| 2077 | job completion, set `cgroup_nodelete=1'. This can be useful if one wants |
| 2078 | to inspect various cgroup files after job completion. Default: false. |
| 2079 | .TP |
| 2080 | .BI flow_id \fR=\fPint |
| 2081 | The ID of the flow. If not specified, it defaults to being a global |
| 2082 | flow. See \fBflow\fR. |
| 2083 | .TP |
| 2084 | .BI flow \fR=\fPint |
| 2085 | Weight in token\-based flow control. If this value is used, then there is |
| 2086 | a 'flow counter' which is used to regulate the proportion of activity between |
| 2087 | two or more jobs. Fio attempts to keep this flow counter near zero. The |
| 2088 | \fBflow\fR parameter stands for how much should be added or subtracted to the |
| 2089 | flow counter on each iteration of the main I/O loop. That is, if one job has |
| 2090 | `flow=8' and another job has `flow=\-1', then there will be a roughly 1:8 |
| 2091 | ratio in how much one runs vs the other. |
| 2092 | .TP |
| 2093 | .BI flow_watermark \fR=\fPint |
| 2094 | The maximum value that the absolute value of the flow counter is allowed to |
| 2095 | reach before the job must wait for a lower value of the counter. |
| 2096 | .TP |
| 2097 | .BI flow_sleep \fR=\fPint |
| 2098 | The period of time, in microseconds, to wait after the flow watermark has |
| 2099 | been exceeded before retrying operations. |
| 2100 | .TP |
| 2101 | .BI stonewall "\fR,\fB wait_for_previous" |
| 2102 | Wait for preceding jobs in the job file to exit, before starting this |
| 2103 | one. Can be used to insert serialization points in the job file. A stone |
| 2104 | wall also implies starting a new reporting group, see |
| 2105 | \fBgroup_reporting\fR. |
| 2106 | .TP |
| 2107 | .BI exitall |
| 2108 | By default, fio will continue running all other jobs when one job finishes |
| 2109 | but sometimes this is not the desired action. Setting \fBexitall\fR will |
| 2110 | instead make fio terminate all other jobs when one job finishes. |
| 2111 | .TP |
| 2112 | .BI exec_prerun \fR=\fPstr |
| 2113 | Before running this job, issue the command specified through |
| 2114 | \fBsystem\fR\|(3). Output is redirected in a file called `jobname.prerun.txt'. |
| 2115 | .TP |
| 2116 | .BI exec_postrun \fR=\fPstr |
| 2117 | After the job completes, issue the command specified though |
| 2118 | \fBsystem\fR\|(3). Output is redirected in a file called `jobname.postrun.txt'. |
| 2119 | .TP |
| 2120 | .BI uid \fR=\fPint |
| 2121 | Instead of running as the invoking user, set the user ID to this value |
| 2122 | before the thread/process does any work. |
| 2123 | .TP |
| 2124 | .BI gid \fR=\fPint |
| 2125 | Set group ID, see \fBuid\fR. |
| 2126 | .SS "Verification" |
| 2127 | .TP |
| 2128 | .BI verify_only |
| 2129 | Do not perform specified workload, only verify data still matches previous |
| 2130 | invocation of this workload. This option allows one to check data multiple |
| 2131 | times at a later date without overwriting it. This option makes sense only |
| 2132 | for workloads that write data, and does not support workloads with the |
| 2133 | \fBtime_based\fR option set. |
| 2134 | .TP |
| 2135 | .BI do_verify \fR=\fPbool |
| 2136 | Run the verify phase after a write phase. Only valid if \fBverify\fR is |
| 2137 | set. Default: true. |
| 2138 | .TP |
| 2139 | .BI verify \fR=\fPstr |
| 2140 | If writing to a file, fio can verify the file contents after each iteration |
| 2141 | of the job. Each verification method also implies verification of special |
| 2142 | header, which is written to the beginning of each block. This header also |
| 2143 | includes meta information, like offset of the block, block number, timestamp |
| 2144 | when block was written, etc. \fBverify\fR can be combined with |
| 2145 | \fBverify_pattern\fR option. The allowed values are: |
| 2146 | .RS |
| 2147 | .RS |
| 2148 | .TP |
| 2149 | .B md5 |
| 2150 | Use an md5 sum of the data area and store it in the header of |
| 2151 | each block. |
| 2152 | .TP |
| 2153 | .B crc64 |
| 2154 | Use an experimental crc64 sum of the data area and store it in the |
| 2155 | header of each block. |
| 2156 | .TP |
| 2157 | .B crc32c |
| 2158 | Use a crc32c sum of the data area and store it in the header of |
| 2159 | each block. This will automatically use hardware acceleration |
| 2160 | (e.g. SSE4.2 on an x86 or CRC crypto extensions on ARM64) but will |
| 2161 | fall back to software crc32c if none is found. Generally the |
| 2162 | fatest checksum fio supports when hardware accelerated. |
| 2163 | .TP |
| 2164 | .B crc32c\-intel |
| 2165 | Synonym for crc32c. |
| 2166 | .TP |
| 2167 | .B crc32 |
| 2168 | Use a crc32 sum of the data area and store it in the header of each |
| 2169 | block. |
| 2170 | .TP |
| 2171 | .B crc16 |
| 2172 | Use a crc16 sum of the data area and store it in the header of each |
| 2173 | block. |
| 2174 | .TP |
| 2175 | .B crc7 |
| 2176 | Use a crc7 sum of the data area and store it in the header of each |
| 2177 | block. |
| 2178 | .TP |
| 2179 | .B xxhash |
| 2180 | Use xxhash as the checksum function. Generally the fastest software |
| 2181 | checksum that fio supports. |
| 2182 | .TP |
| 2183 | .B sha512 |
| 2184 | Use sha512 as the checksum function. |
| 2185 | .TP |
| 2186 | .B sha256 |
| 2187 | Use sha256 as the checksum function. |
| 2188 | .TP |
| 2189 | .B sha1 |
| 2190 | Use optimized sha1 as the checksum function. |
| 2191 | .TP |
| 2192 | .B sha3\-224 |
| 2193 | Use optimized sha3\-224 as the checksum function. |
| 2194 | .TP |
| 2195 | .B sha3\-256 |
| 2196 | Use optimized sha3\-256 as the checksum function. |
| 2197 | .TP |
| 2198 | .B sha3\-384 |
| 2199 | Use optimized sha3\-384 as the checksum function. |
| 2200 | .TP |
| 2201 | .B sha3\-512 |
| 2202 | Use optimized sha3\-512 as the checksum function. |
| 2203 | .TP |
| 2204 | .B meta |
| 2205 | This option is deprecated, since now meta information is included in |
| 2206 | generic verification header and meta verification happens by |
| 2207 | default. For detailed information see the description of the |
| 2208 | \fBverify\fR setting. This option is kept because of |
| 2209 | compatibility's sake with old configurations. Do not use it. |
| 2210 | .TP |
| 2211 | .B pattern |
| 2212 | Verify a strict pattern. Normally fio includes a header with some |
| 2213 | basic information and checksumming, but if this option is set, only |
| 2214 | the specific pattern set with \fBverify_pattern\fR is verified. |
| 2215 | .TP |
| 2216 | .B null |
| 2217 | Only pretend to verify. Useful for testing internals with |
| 2218 | `ioengine=null', not for much else. |
| 2219 | .RE |
| 2220 | .P |
| 2221 | This option can be used for repeated burn\-in tests of a system to make sure |
| 2222 | that the written data is also correctly read back. If the data direction |
| 2223 | given is a read or random read, fio will assume that it should verify a |
| 2224 | previously written file. If the data direction includes any form of write, |
| 2225 | the verify will be of the newly written data. |
| 2226 | .RE |
| 2227 | .TP |
| 2228 | .BI verifysort \fR=\fPbool |
| 2229 | If true, fio will sort written verify blocks when it deems it faster to read |
| 2230 | them back in a sorted manner. This is often the case when overwriting an |
| 2231 | existing file, since the blocks are already laid out in the file system. You |
| 2232 | can ignore this option unless doing huge amounts of really fast I/O where |
| 2233 | the red\-black tree sorting CPU time becomes significant. Default: true. |
| 2234 | .TP |
| 2235 | .BI verifysort_nr \fR=\fPint |
| 2236 | Pre\-load and sort verify blocks for a read workload. |
| 2237 | .TP |
| 2238 | .BI verify_offset \fR=\fPint |
| 2239 | Swap the verification header with data somewhere else in the block before |
| 2240 | writing. It is swapped back before verifying. |
| 2241 | .TP |
| 2242 | .BI verify_interval \fR=\fPint |
| 2243 | Write the verification header at a finer granularity than the |
| 2244 | \fBblocksize\fR. It will be written for chunks the size of |
| 2245 | \fBverify_interval\fR. \fBblocksize\fR should divide this evenly. |
| 2246 | .TP |
| 2247 | .BI verify_pattern \fR=\fPstr |
| 2248 | If set, fio will fill the I/O buffers with this pattern. Fio defaults to |
| 2249 | filling with totally random bytes, but sometimes it's interesting to fill |
| 2250 | with a known pattern for I/O verification purposes. Depending on the width |
| 2251 | of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time (it can |
| 2252 | be either a decimal or a hex number). The \fBverify_pattern\fR if larger than |
| 2253 | a 32\-bit quantity has to be a hex number that starts with either "0x" or |
| 2254 | "0X". Use with \fBverify\fR. Also, \fBverify_pattern\fR supports %o |
| 2255 | format, which means that for each block offset will be written and then |
| 2256 | verified back, e.g.: |
| 2257 | .RS |
| 2258 | .RS |
| 2259 | .P |
| 2260 | verify_pattern=%o |
| 2261 | .RE |
| 2262 | .P |
| 2263 | Or use combination of everything: |
| 2264 | .RS |
| 2265 | .P |
| 2266 | verify_pattern=0xff%o"abcd"\-12 |
| 2267 | .RE |
| 2268 | .RE |
| 2269 | .TP |
| 2270 | .BI verify_fatal \fR=\fPbool |
| 2271 | Normally fio will keep checking the entire contents before quitting on a |
| 2272 | block verification failure. If this option is set, fio will exit the job on |
| 2273 | the first observed failure. Default: false. |
| 2274 | .TP |
| 2275 | .BI verify_dump \fR=\fPbool |
| 2276 | If set, dump the contents of both the original data block and the data block |
| 2277 | we read off disk to files. This allows later analysis to inspect just what |
| 2278 | kind of data corruption occurred. Off by default. |
| 2279 | .TP |
| 2280 | .BI verify_async \fR=\fPint |
| 2281 | Fio will normally verify I/O inline from the submitting thread. This option |
| 2282 | takes an integer describing how many async offload threads to create for I/O |
| 2283 | verification instead, causing fio to offload the duty of verifying I/O |
| 2284 | contents to one or more separate threads. If using this offload option, even |
| 2285 | sync I/O engines can benefit from using an \fBiodepth\fR setting higher |
| 2286 | than 1, as it allows them to have I/O in flight while verifies are running. |
| 2287 | Defaults to 0 async threads, i.e. verification is not asynchronous. |
| 2288 | .TP |
| 2289 | .BI verify_async_cpus \fR=\fPstr |
| 2290 | Tell fio to set the given CPU affinity on the async I/O verification |
| 2291 | threads. See \fBcpus_allowed\fR for the format used. |
| 2292 | .TP |
| 2293 | .BI verify_backlog \fR=\fPint |
| 2294 | Fio will normally verify the written contents of a job that utilizes verify |
| 2295 | once that job has completed. In other words, everything is written then |
| 2296 | everything is read back and verified. You may want to verify continually |
| 2297 | instead for a variety of reasons. Fio stores the meta data associated with |
| 2298 | an I/O block in memory, so for large verify workloads, quite a bit of memory |
| 2299 | would be used up holding this meta data. If this option is enabled, fio will |
| 2300 | write only N blocks before verifying these blocks. |
| 2301 | .TP |
| 2302 | .BI verify_backlog_batch \fR=\fPint |
| 2303 | Control how many blocks fio will verify if \fBverify_backlog\fR is |
| 2304 | set. If not set, will default to the value of \fBverify_backlog\fR |
| 2305 | (meaning the entire queue is read back and verified). If |
| 2306 | \fBverify_backlog_batch\fR is less than \fBverify_backlog\fR then not all |
| 2307 | blocks will be verified, if \fBverify_backlog_batch\fR is larger than |
| 2308 | \fBverify_backlog\fR, some blocks will be verified more than once. |
| 2309 | .TP |
| 2310 | .BI verify_state_save \fR=\fPbool |
| 2311 | When a job exits during the write phase of a verify workload, save its |
| 2312 | current state. This allows fio to replay up until that point, if the verify |
| 2313 | state is loaded for the verify read phase. The format of the filename is, |
| 2314 | roughly: |
| 2315 | .RS |
| 2316 | .RS |
| 2317 | .P |
| 2318 | <type>\-<jobname>\-<jobindex>\-verify.state. |
| 2319 | .RE |
| 2320 | .P |
| 2321 | <type> is "local" for a local run, "sock" for a client/server socket |
| 2322 | connection, and "ip" (192.168.0.1, for instance) for a networked |
| 2323 | client/server connection. Defaults to true. |
| 2324 | .RE |
| 2325 | .TP |
| 2326 | .BI verify_state_load \fR=\fPbool |
| 2327 | If a verify termination trigger was used, fio stores the current write state |
| 2328 | of each thread. This can be used at verification time so that fio knows how |
| 2329 | far it should verify. Without this information, fio will run a full |
| 2330 | verification pass, according to the settings in the job file used. Default |
| 2331 | false. |
| 2332 | .TP |
| 2333 | .BI trim_percentage \fR=\fPint |
| 2334 | Number of verify blocks to discard/trim. |
| 2335 | .TP |
| 2336 | .BI trim_verify_zero \fR=\fPbool |
| 2337 | Verify that trim/discarded blocks are returned as zeros. |
| 2338 | .TP |
| 2339 | .BI trim_backlog \fR=\fPint |
| 2340 | Verify that trim/discarded blocks are returned as zeros. |
| 2341 | .TP |
| 2342 | .BI trim_backlog_batch \fR=\fPint |
| 2343 | Trim this number of I/O blocks. |
| 2344 | .TP |
| 2345 | .BI experimental_verify \fR=\fPbool |
| 2346 | Enable experimental verification. |
| 2347 | .SS "Steady state" |
| 2348 | .TP |
| 2349 | .BI steadystate \fR=\fPstr:float "\fR,\fP ss" \fR=\fPstr:float |
| 2350 | Define the criterion and limit for assessing steady state performance. The |
| 2351 | first parameter designates the criterion whereas the second parameter sets |
| 2352 | the threshold. When the criterion falls below the threshold for the |
| 2353 | specified duration, the job will stop. For example, `iops_slope:0.1%' will |
| 2354 | direct fio to terminate the job when the least squares regression slope |
| 2355 | falls below 0.1% of the mean IOPS. If \fBgroup_reporting\fR is enabled |
| 2356 | this will apply to all jobs in the group. Below is the list of available |
| 2357 | steady state assessment criteria. All assessments are carried out using only |
| 2358 | data from the rolling collection window. Threshold limits can be expressed |
| 2359 | as a fixed value or as a percentage of the mean in the collection window. |
| 2360 | .RS |
| 2361 | .RS |
| 2362 | .TP |
| 2363 | .B iops |
| 2364 | Collect IOPS data. Stop the job if all individual IOPS measurements |
| 2365 | are within the specified limit of the mean IOPS (e.g., `iops:2' |
| 2366 | means that all individual IOPS values must be within 2 of the mean, |
| 2367 | whereas `iops:0.2%' means that all individual IOPS values must be |
| 2368 | within 0.2% of the mean IOPS to terminate the job). |
| 2369 | .TP |
| 2370 | .B iops_slope |
| 2371 | Collect IOPS data and calculate the least squares regression |
| 2372 | slope. Stop the job if the slope falls below the specified limit. |
| 2373 | .TP |
| 2374 | .B bw |
| 2375 | Collect bandwidth data. Stop the job if all individual bandwidth |
| 2376 | measurements are within the specified limit of the mean bandwidth. |
| 2377 | .TP |
| 2378 | .B bw_slope |
| 2379 | Collect bandwidth data and calculate the least squares regression |
| 2380 | slope. Stop the job if the slope falls below the specified limit. |
| 2381 | .RE |
| 2382 | .RE |
| 2383 | .TP |
| 2384 | .BI steadystate_duration \fR=\fPtime "\fR,\fP ss_dur" \fR=\fPtime |
| 2385 | A rolling window of this duration will be used to judge whether steady state |
| 2386 | has been reached. Data will be collected once per second. The default is 0 |
| 2387 | which disables steady state detection. When the unit is omitted, the |
| 2388 | value is interpreted in seconds. |
| 2389 | .TP |
| 2390 | .BI steadystate_ramp_time \fR=\fPtime "\fR,\fP ss_ramp" \fR=\fPtime |
| 2391 | Allow the job to run for the specified duration before beginning data |
| 2392 | collection for checking the steady state job termination criterion. The |
| 2393 | default is 0. When the unit is omitted, the value is interpreted in seconds. |
| 2394 | .SS "Measurements and reporting" |
| 2395 | .TP |
| 2396 | .BI per_job_logs \fR=\fPbool |
| 2397 | If set, this generates bw/clat/iops log with per file private filenames. If |
| 2398 | not set, jobs with identical names will share the log filename. Default: |
| 2399 | true. |
| 2400 | .TP |
| 2401 | .BI group_reporting |
| 2402 | It may sometimes be interesting to display statistics for groups of jobs as |
| 2403 | a whole instead of for each individual job. This is especially true if |
| 2404 | \fBnumjobs\fR is used; looking at individual thread/process output |
| 2405 | quickly becomes unwieldy. To see the final report per\-group instead of |
| 2406 | per\-job, use \fBgroup_reporting\fR. Jobs in a file will be part of the |
| 2407 | same reporting group, unless if separated by a \fBstonewall\fR, or by |
| 2408 | using \fBnew_group\fR. |
| 2409 | .TP |
| 2410 | .BI new_group |
| 2411 | Start a new reporting group. See: \fBgroup_reporting\fR. If not given, |
| 2412 | all jobs in a file will be part of the same reporting group, unless |
| 2413 | separated by a \fBstonewall\fR. |
| 2414 | .TP |
| 2415 | .BI stats \fR=\fPbool |
| 2416 | By default, fio collects and shows final output results for all jobs |
| 2417 | that run. If this option is set to 0, then fio will ignore it in |
| 2418 | the final stat output. |
| 2419 | .TP |
| 2420 | .BI write_bw_log \fR=\fPstr |
| 2421 | If given, write a bandwidth log for this job. Can be used to store data of |
| 2422 | the bandwidth of the jobs in their lifetime. The included |
| 2423 | \fBfio_generate_plots\fR script uses gnuplot to turn these |
| 2424 | text files into nice graphs. See \fBwrite_lat_log\fR for behavior of |
| 2425 | given filename. For this option, the postfix is `_bw.x.log', where `x' |
| 2426 | is the index of the job (1..N, where N is the number of jobs). If |
| 2427 | \fBper_job_logs\fR is false, then the filename will not include the job |
| 2428 | index. See \fBLOG FILE FORMATS\fR section. |
| 2429 | .TP |
| 2430 | .BI write_lat_log \fR=\fPstr |
| 2431 | Same as \fBwrite_bw_log\fR, except that this option stores I/O |
| 2432 | submission, completion, and total latencies instead. If no filename is given |
| 2433 | with this option, the default filename of `jobname_type.log' is |
| 2434 | used. Even if the filename is given, fio will still append the type of |
| 2435 | log. So if one specifies: |
| 2436 | .RS |
| 2437 | .RS |
| 2438 | .P |
| 2439 | write_lat_log=foo |
| 2440 | .RE |
| 2441 | .P |
| 2442 | The actual log names will be `foo_slat.x.log', `foo_clat.x.log', |
| 2443 | and `foo_lat.x.log', where `x' is the index of the job (1..N, where N |
| 2444 | is the number of jobs). This helps \fBfio_generate_plots\fR find the |
| 2445 | logs automatically. If \fBper_job_logs\fR is false, then the filename |
| 2446 | will not include the job index. See \fBLOG FILE FORMATS\fR section. |
| 2447 | .RE |
| 2448 | .TP |
| 2449 | .BI write_hist_log \fR=\fPstr |
| 2450 | Same as \fBwrite_lat_log\fR, but writes I/O completion latency |
| 2451 | histograms. If no filename is given with this option, the default filename |
| 2452 | of `jobname_clat_hist.x.log' is used, where `x' is the index of the |
| 2453 | job (1..N, where N is the number of jobs). Even if the filename is given, |
| 2454 | fio will still append the type of log. If \fBper_job_logs\fR is false, |
| 2455 | then the filename will not include the job index. See \fBLOG FILE FORMATS\fR section. |
| 2456 | .TP |
| 2457 | .BI write_iops_log \fR=\fPstr |
| 2458 | Same as \fBwrite_bw_log\fR, but writes IOPS. If no filename is given |
| 2459 | with this option, the default filename of `jobname_type.x.log' is |
| 2460 | used, where `x' is the index of the job (1..N, where N is the number of |
| 2461 | jobs). Even if the filename is given, fio will still append the type of |
| 2462 | log. If \fBper_job_logs\fR is false, then the filename will not include |
| 2463 | the job index. See \fBLOG FILE FORMATS\fR section. |
| 2464 | .TP |
| 2465 | .BI log_avg_msec \fR=\fPint |
| 2466 | By default, fio will log an entry in the iops, latency, or bw log for every |
| 2467 | I/O that completes. When writing to the disk log, that can quickly grow to a |
| 2468 | very large size. Setting this option makes fio average the each log entry |
| 2469 | over the specified period of time, reducing the resolution of the log. See |
| 2470 | \fBlog_max_value\fR as well. Defaults to 0, logging all entries. |
| 2471 | Also see \fBLOG FILE FORMATS\fR section. |
| 2472 | .TP |
| 2473 | .BI log_hist_msec \fR=\fPint |
| 2474 | Same as \fBlog_avg_msec\fR, but logs entries for completion latency |
| 2475 | histograms. Computing latency percentiles from averages of intervals using |
| 2476 | \fBlog_avg_msec\fR is inaccurate. Setting this option makes fio log |
| 2477 | histogram entries over the specified period of time, reducing log sizes for |
| 2478 | high IOPS devices while retaining percentile accuracy. See |
| 2479 | \fBlog_hist_coarseness\fR as well. Defaults to 0, meaning histogram |
| 2480 | logging is disabled. |
| 2481 | .TP |
| 2482 | .BI log_hist_coarseness \fR=\fPint |
| 2483 | Integer ranging from 0 to 6, defining the coarseness of the resolution of |
| 2484 | the histogram logs enabled with \fBlog_hist_msec\fR. For each increment |
| 2485 | in coarseness, fio outputs half as many bins. Defaults to 0, for which |
| 2486 | histogram logs contain 1216 latency bins. See \fBLOG FILE FORMATS\fR section. |
| 2487 | .TP |
| 2488 | .BI log_max_value \fR=\fPbool |
| 2489 | If \fBlog_avg_msec\fR is set, fio logs the average over that window. If |
| 2490 | you instead want to log the maximum value, set this option to 1. Defaults to |
| 2491 | 0, meaning that averaged values are logged. |
| 2492 | .TP |
| 2493 | .BI log_offset \fR=\fPbool |
| 2494 | If this is set, the iolog options will include the byte offset for the I/O |
| 2495 | entry as well as the other data values. Defaults to 0 meaning that |
| 2496 | offsets are not present in logs. Also see \fBLOG FILE FORMATS\fR section. |
| 2497 | .TP |
| 2498 | .BI log_compression \fR=\fPint |
| 2499 | If this is set, fio will compress the I/O logs as it goes, to keep the |
| 2500 | memory footprint lower. When a log reaches the specified size, that chunk is |
| 2501 | removed and compressed in the background. Given that I/O logs are fairly |
| 2502 | highly compressible, this yields a nice memory savings for longer runs. The |
| 2503 | downside is that the compression will consume some background CPU cycles, so |
| 2504 | it may impact the run. This, however, is also true if the logging ends up |
| 2505 | consuming most of the system memory. So pick your poison. The I/O logs are |
| 2506 | saved normally at the end of a run, by decompressing the chunks and storing |
| 2507 | them in the specified log file. This feature depends on the availability of |
| 2508 | zlib. |
| 2509 | .TP |
| 2510 | .BI log_compression_cpus \fR=\fPstr |
| 2511 | Define the set of CPUs that are allowed to handle online log compression for |
| 2512 | the I/O jobs. This can provide better isolation between performance |
| 2513 | sensitive jobs, and background compression work. |
| 2514 | .TP |
| 2515 | .BI log_store_compressed \fR=\fPbool |
| 2516 | If set, fio will store the log files in a compressed format. They can be |
| 2517 | decompressed with fio, using the \fB\-\-inflate\-log\fR command line |
| 2518 | parameter. The files will be stored with a `.fz' suffix. |
| 2519 | .TP |
| 2520 | .BI log_unix_epoch \fR=\fPbool |
| 2521 | If set, fio will log Unix timestamps to the log files produced by enabling |
| 2522 | write_type_log for each log type, instead of the default zero\-based |
| 2523 | timestamps. |
| 2524 | .TP |
| 2525 | .BI block_error_percentiles \fR=\fPbool |
| 2526 | If set, record errors in trim block\-sized units from writes and trims and |
| 2527 | output a histogram of how many trims it took to get to errors, and what kind |
| 2528 | of error was encountered. |
| 2529 | .TP |
| 2530 | .BI bwavgtime \fR=\fPint |
| 2531 | Average the calculated bandwidth over the given time. Value is specified in |
| 2532 | milliseconds. If the job also does bandwidth logging through |
| 2533 | \fBwrite_bw_log\fR, then the minimum of this option and |
| 2534 | \fBlog_avg_msec\fR will be used. Default: 500ms. |
| 2535 | .TP |
| 2536 | .BI iopsavgtime \fR=\fPint |
| 2537 | Average the calculated IOPS over the given time. Value is specified in |
| 2538 | milliseconds. If the job also does IOPS logging through |
| 2539 | \fBwrite_iops_log\fR, then the minimum of this option and |
| 2540 | \fBlog_avg_msec\fR will be used. Default: 500ms. |
| 2541 | .TP |
| 2542 | .BI disk_util \fR=\fPbool |
| 2543 | Generate disk utilization statistics, if the platform supports it. |
| 2544 | Default: true. |
| 2545 | .TP |
| 2546 | .BI disable_lat \fR=\fPbool |
| 2547 | Disable measurements of total latency numbers. Useful only for cutting back |
| 2548 | the number of calls to \fBgettimeofday\fR\|(2), as that does impact |
| 2549 | performance at really high IOPS rates. Note that to really get rid of a |
| 2550 | large amount of these calls, this option must be used with |
| 2551 | \fBdisable_slat\fR and \fBdisable_bw_measurement\fR as well. |
| 2552 | .TP |
| 2553 | .BI disable_clat \fR=\fPbool |
| 2554 | Disable measurements of completion latency numbers. See |
| 2555 | \fBdisable_lat\fR. |
| 2556 | .TP |
| 2557 | .BI disable_slat \fR=\fPbool |
| 2558 | Disable measurements of submission latency numbers. See |
| 2559 | \fBdisable_lat\fR. |
| 2560 | .TP |
| 2561 | .BI disable_bw_measurement \fR=\fPbool "\fR,\fP disable_bw" \fR=\fPbool |
| 2562 | Disable measurements of throughput/bandwidth numbers. See |
| 2563 | \fBdisable_lat\fR. |
| 2564 | .TP |
| 2565 | .BI clat_percentiles \fR=\fPbool |
| 2566 | Enable the reporting of percentiles of completion latencies. This option is |
| 2567 | mutually exclusive with \fBlat_percentiles\fR. |
| 2568 | .TP |
| 2569 | .BI lat_percentiles \fR=\fPbool |
| 2570 | Enable the reporting of percentiles of IO latencies. This is similar to |
| 2571 | \fBclat_percentiles\fR, except that this includes the submission latency. |
| 2572 | This option is mutually exclusive with \fBclat_percentiles\fR. |
| 2573 | .TP |
| 2574 | .BI percentile_list \fR=\fPfloat_list |
| 2575 | Overwrite the default list of percentiles for completion latencies and the |
| 2576 | block error histogram. Each number is a floating number in the range |
| 2577 | (0,100], and the maximum length of the list is 20. Use ':' to separate the |
| 2578 | numbers, and list the numbers in ascending order. For example, |
| 2579 | `\-\-percentile_list=99.5:99.9' will cause fio to report the values of |
| 2580 | completion latency below which 99.5% and 99.9% of the observed latencies |
| 2581 | fell, respectively. |
| 2582 | .SS "Error handling" |
| 2583 | .TP |
| 2584 | .BI exitall_on_error |
| 2585 | When one job finishes in error, terminate the rest. The default is to wait |
| 2586 | for each job to finish. |
| 2587 | .TP |
| 2588 | .BI continue_on_error \fR=\fPstr |
| 2589 | Normally fio will exit the job on the first observed failure. If this option |
| 2590 | is set, fio will continue the job when there is a 'non\-fatal error' (EIO or |
| 2591 | EILSEQ) until the runtime is exceeded or the I/O size specified is |
| 2592 | completed. If this option is used, there are two more stats that are |
| 2593 | appended, the total error count and the first error. The error field given |
| 2594 | in the stats is the first error that was hit during the run. |
| 2595 | The allowed values are: |
| 2596 | .RS |
| 2597 | .RS |
| 2598 | .TP |
| 2599 | .B none |
| 2600 | Exit on any I/O or verify errors. |
| 2601 | .TP |
| 2602 | .B read |
| 2603 | Continue on read errors, exit on all others. |
| 2604 | .TP |
| 2605 | .B write |
| 2606 | Continue on write errors, exit on all others. |
| 2607 | .TP |
| 2608 | .B io |
| 2609 | Continue on any I/O error, exit on all others. |
| 2610 | .TP |
| 2611 | .B verify |
| 2612 | Continue on verify errors, exit on all others. |
| 2613 | .TP |
| 2614 | .B all |
| 2615 | Continue on all errors. |
| 2616 | .TP |
| 2617 | .B 0 |
| 2618 | Backward\-compatible alias for 'none'. |
| 2619 | .TP |
| 2620 | .B 1 |
| 2621 | Backward\-compatible alias for 'all'. |
| 2622 | .RE |
| 2623 | .RE |
| 2624 | .TP |
| 2625 | .BI ignore_error \fR=\fPstr |
| 2626 | Sometimes you want to ignore some errors during test in that case you can |
| 2627 | specify error list for each error type, instead of only being able to |
| 2628 | ignore the default 'non\-fatal error' using \fBcontinue_on_error\fR. |
| 2629 | `ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST' errors for |
| 2630 | given error type is separated with ':'. Error may be symbol ('ENOSPC', 'ENOMEM') |
| 2631 | or integer. Example: |
| 2632 | .RS |
| 2633 | .RS |
| 2634 | .P |
| 2635 | ignore_error=EAGAIN,ENOSPC:122 |
| 2636 | .RE |
| 2637 | .P |
| 2638 | This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from |
| 2639 | WRITE. This option works by overriding \fBcontinue_on_error\fR with |
| 2640 | the list of errors for each error type if any. |
| 2641 | .RE |
| 2642 | .TP |
| 2643 | .BI error_dump \fR=\fPbool |
| 2644 | If set dump every error even if it is non fatal, true by default. If |
| 2645 | disabled only fatal error will be dumped. |
| 2646 | .SS "Running predefined workloads" |
| 2647 | Fio includes predefined profiles that mimic the I/O workloads generated by |
| 2648 | other tools. |
| 2649 | .TP |
| 2650 | .BI profile \fR=\fPstr |
| 2651 | The predefined workload to run. Current profiles are: |
| 2652 | .RS |
| 2653 | .RS |
| 2654 | .TP |
| 2655 | .B tiobench |
| 2656 | Threaded I/O bench (tiotest/tiobench) like workload. |
| 2657 | .TP |
| 2658 | .B act |
| 2659 | Aerospike Certification Tool (ACT) like workload. |
| 2660 | .RE |
| 2661 | .RE |
| 2662 | .P |
| 2663 | To view a profile's additional options use \fB\-\-cmdhelp\fR after specifying |
| 2664 | the profile. For example: |
| 2665 | .RS |
| 2666 | .TP |
| 2667 | $ fio \-\-profile=act \-\-cmdhelp |
| 2668 | .RE |
| 2669 | .SS "Act profile options" |
| 2670 | .TP |
| 2671 | .BI device\-names \fR=\fPstr |
| 2672 | Devices to use. |
| 2673 | .TP |
| 2674 | .BI load \fR=\fPint |
| 2675 | ACT load multiplier. Default: 1. |
| 2676 | .TP |
| 2677 | .BI test\-duration\fR=\fPtime |
| 2678 | How long the entire test takes to run. When the unit is omitted, the value |
| 2679 | is given in seconds. Default: 24h. |
| 2680 | .TP |
| 2681 | .BI threads\-per\-queue\fR=\fPint |
| 2682 | Number of read I/O threads per device. Default: 8. |
| 2683 | .TP |
| 2684 | .BI read\-req\-num\-512\-blocks\fR=\fPint |
| 2685 | Number of 512B blocks to read at the time. Default: 3. |
| 2686 | .TP |
| 2687 | .BI large\-block\-op\-kbytes\fR=\fPint |
| 2688 | Size of large block ops in KiB (writes). Default: 131072. |
| 2689 | .TP |
| 2690 | .BI prep |
| 2691 | Set to run ACT prep phase. |
| 2692 | .SS "Tiobench profile options" |
| 2693 | .TP |
| 2694 | .BI size\fR=\fPstr |
| 2695 | Size in MiB. |
| 2696 | .TP |
| 2697 | .BI block\fR=\fPint |
| 2698 | Block size in bytes. Default: 4096. |
| 2699 | .TP |
| 2700 | .BI numruns\fR=\fPint |
| 2701 | Number of runs. |
| 2702 | .TP |
| 2703 | .BI dir\fR=\fPstr |
| 2704 | Test directory. |
| 2705 | .TP |
| 2706 | .BI threads\fR=\fPint |
| 2707 | Number of threads. |
| 2708 | .SH OUTPUT |
| 2709 | Fio spits out a lot of output. While running, fio will display the status of the |
| 2710 | jobs created. An example of that would be: |
| 2711 | .P |
| 2712 | .nf |
| 2713 | 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] |
| 2714 | .fi |
| 2715 | .P |
| 2716 | The characters inside the first set of square brackets denote the current status of |
| 2717 | each thread. The first character is the first job defined in the job file, and so |
| 2718 | forth. The possible values (in typical life cycle order) are: |
| 2719 | .RS |
| 2720 | .TP |
| 2721 | .PD 0 |
| 2722 | .B P |
| 2723 | Thread setup, but not started. |
| 2724 | .TP |
| 2725 | .B C |
| 2726 | Thread created. |
| 2727 | .TP |
| 2728 | .B I |
| 2729 | Thread initialized, waiting or generating necessary data. |
| 2730 | .TP |
| 2731 | .B P |
| 2732 | Thread running pre\-reading file(s). |
| 2733 | .TP |
| 2734 | .B / |
| 2735 | Thread is in ramp period. |
| 2736 | .TP |
| 2737 | .B R |
| 2738 | Running, doing sequential reads. |
| 2739 | .TP |
| 2740 | .B r |
| 2741 | Running, doing random reads. |
| 2742 | .TP |
| 2743 | .B W |
| 2744 | Running, doing sequential writes. |
| 2745 | .TP |
| 2746 | .B w |
| 2747 | Running, doing random writes. |
| 2748 | .TP |
| 2749 | .B M |
| 2750 | Running, doing mixed sequential reads/writes. |
| 2751 | .TP |
| 2752 | .B m |
| 2753 | Running, doing mixed random reads/writes. |
| 2754 | .TP |
| 2755 | .B D |
| 2756 | Running, doing sequential trims. |
| 2757 | .TP |
| 2758 | .B d |
| 2759 | Running, doing random trims. |
| 2760 | .TP |
| 2761 | .B F |
| 2762 | Running, currently waiting for \fBfsync\fR\|(2). |
| 2763 | .TP |
| 2764 | .B V |
| 2765 | Running, doing verification of written data. |
| 2766 | .TP |
| 2767 | .B f |
| 2768 | Thread finishing. |
| 2769 | .TP |
| 2770 | .B E |
| 2771 | Thread exited, not reaped by main thread yet. |
| 2772 | .TP |
| 2773 | .B \- |
| 2774 | Thread reaped. |
| 2775 | .TP |
| 2776 | .B X |
| 2777 | Thread reaped, exited with an error. |
| 2778 | .TP |
| 2779 | .B K |
| 2780 | Thread reaped, exited due to signal. |
| 2781 | .PD |
| 2782 | .RE |
| 2783 | .P |
| 2784 | Fio will condense the thread string as not to take up more space on the command |
| 2785 | line than needed. For instance, if you have 10 readers and 10 writers running, |
| 2786 | the output would look like this: |
| 2787 | .P |
| 2788 | .nf |
| 2789 | 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] |
| 2790 | .fi |
| 2791 | .P |
| 2792 | Note that the status string is displayed in order, so it's possible to tell which of |
| 2793 | the jobs are currently doing what. In the example above this means that jobs 1\-\-10 |
| 2794 | are readers and 11\-\-20 are writers. |
| 2795 | .P |
| 2796 | The other values are fairly self explanatory \-\- number of threads currently |
| 2797 | running and doing I/O, the number of currently open files (f=), the estimated |
| 2798 | completion percentage, the rate of I/O since last check (read speed listed first, |
| 2799 | then write speed and optionally trim speed) in terms of bandwidth and IOPS, |
| 2800 | and time to completion for the current running group. It's impossible to estimate |
| 2801 | runtime of the following groups (if any). |
| 2802 | .P |
| 2803 | When fio is done (or interrupted by Ctrl\-C), it will show the data for |
| 2804 | each thread, group of threads, and disks in that order. For each overall thread (or |
| 2805 | group) the output looks like: |
| 2806 | .P |
| 2807 | .nf |
| 2808 | Client1: (groupid=0, jobs=1): err= 0: pid=16109: Sat Jun 24 12:07:54 2017 |
| 2809 | write: IOPS=88, BW=623KiB/s (638kB/s)(30.4MiB/50032msec) |
| 2810 | slat (nsec): min=500, max=145500, avg=8318.00, stdev=4781.50 |
| 2811 | clat (usec): min=170, max=78367, avg=4019.02, stdev=8293.31 |
| 2812 | lat (usec): min=174, max=78375, avg=4027.34, stdev=8291.79 |
| 2813 | clat percentiles (usec): |
| 2814 | | 1.00th=[ 302], 5.00th=[ 326], 10.00th=[ 343], 20.00th=[ 363], |
| 2815 | | 30.00th=[ 392], 40.00th=[ 404], 50.00th=[ 416], 60.00th=[ 445], |
| 2816 | | 70.00th=[ 816], 80.00th=[ 6718], 90.00th=[12911], 95.00th=[21627], |
| 2817 | | 99.00th=[43779], 99.50th=[51643], 99.90th=[68682], 99.95th=[72877], |
| 2818 | | 99.99th=[78119] |
| 2819 | bw ( KiB/s): min= 532, max= 686, per=0.10%, avg=622.87, stdev=24.82, samples= 100 |
| 2820 | iops : min= 76, max= 98, avg=88.98, stdev= 3.54, samples= 100 |
| 2821 | lat (usec) : 250=0.04%, 500=64.11%, 750=4.81%, 1000=2.79% |
| 2822 | lat (msec) : 2=4.16%, 4=1.84%, 10=4.90%, 20=11.33%, 50=5.37% |
| 2823 | lat (msec) : 100=0.65% |
| 2824 | cpu : usr=0.27%, sys=0.18%, ctx=12072, majf=0, minf=21 |
| 2825 | IO depths : 1=85.0%, 2=13.1%, 4=1.8%, 8=0.1%, 16=0.0%, 32=0.0%, >=64=0.0% |
| 2826 | submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% |
| 2827 | complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% |
| 2828 | issued rwt: total=0,4450,0, short=0,0,0, dropped=0,0,0 |
| 2829 | latency : target=0, window=0, percentile=100.00%, depth=8 |
| 2830 | .fi |
| 2831 | .P |
| 2832 | The job name (or first job's name when using \fBgroup_reporting\fR) is printed, |
| 2833 | along with the group id, count of jobs being aggregated, last error id seen (which |
| 2834 | is 0 when there are no errors), pid/tid of that thread and the time the job/group |
| 2835 | completed. Below are the I/O statistics for each data direction performed (showing |
| 2836 | writes in the example above). In the order listed, they denote: |
| 2837 | .RS |
| 2838 | .TP |
| 2839 | .B read/write/trim |
| 2840 | The string before the colon shows the I/O direction the statistics |
| 2841 | are for. \fIIOPS\fR is the average I/Os performed per second. \fIBW\fR |
| 2842 | is the average bandwidth rate shown as: value in power of 2 format |
| 2843 | (value in power of 10 format). The last two values show: (total |
| 2844 | I/O performed in power of 2 format / \fIruntime\fR of that thread). |
| 2845 | .TP |
| 2846 | .B slat |
| 2847 | Submission latency (\fImin\fR being the minimum, \fImax\fR being the |
| 2848 | maximum, \fIavg\fR being the average, \fIstdev\fR being the standard |
| 2849 | deviation). This is the time it took to submit the I/O. For |
| 2850 | sync I/O this row is not displayed as the slat is really the |
| 2851 | completion latency (since queue/complete is one operation there). |
| 2852 | This value can be in nanoseconds, microseconds or milliseconds \-\-\- |
| 2853 | fio will choose the most appropriate base and print that (in the |
| 2854 | example above nanoseconds was the best scale). Note: in \fB\-\-minimal\fR mode |
| 2855 | latencies are always expressed in microseconds. |
| 2856 | .TP |
| 2857 | .B clat |
| 2858 | Completion latency. Same names as slat, this denotes the time from |
| 2859 | submission to completion of the I/O pieces. For sync I/O, clat will |
| 2860 | usually be equal (or very close) to 0, as the time from submit to |
| 2861 | complete is basically just CPU time (I/O has already been done, see slat |
| 2862 | explanation). |
| 2863 | .TP |
| 2864 | .B lat |
| 2865 | Total latency. Same names as slat and clat, this denotes the time from |
| 2866 | when fio created the I/O unit to completion of the I/O operation. |
| 2867 | .TP |
| 2868 | .B bw |
| 2869 | Bandwidth statistics based on samples. Same names as the xlat stats, |
| 2870 | but also includes the number of samples taken (\fIsamples\fR) and an |
| 2871 | approximate percentage of total aggregate bandwidth this thread |
| 2872 | received in its group (\fIper\fR). This last value is only really |
| 2873 | useful if the threads in this group are on the same disk, since they |
| 2874 | are then competing for disk access. |
| 2875 | .TP |
| 2876 | .B iops |
| 2877 | IOPS statistics based on samples. Same names as \fBbw\fR. |
| 2878 | .TP |
| 2879 | .B lat (nsec/usec/msec) |
| 2880 | The distribution of I/O completion latencies. This is the time from when |
| 2881 | I/O leaves fio and when it gets completed. Unlike the separate |
| 2882 | read/write/trim sections above, the data here and in the remaining |
| 2883 | sections apply to all I/Os for the reporting group. 250=0.04% means that |
| 2884 | 0.04% of the I/Os completed in under 250us. 500=64.11% means that 64.11% |
| 2885 | of the I/Os required 250 to 499us for completion. |
| 2886 | .TP |
| 2887 | .B cpu |
| 2888 | CPU usage. User and system time, along with the number of context |
| 2889 | switches this thread went through, usage of system and user time, and |
| 2890 | finally the number of major and minor page faults. The CPU utilization |
| 2891 | numbers are averages for the jobs in that reporting group, while the |
| 2892 | context and fault counters are summed. |
| 2893 | .TP |
| 2894 | .B IO depths |
| 2895 | The distribution of I/O depths over the job lifetime. The numbers are |
| 2896 | divided into powers of 2 and each entry covers depths from that value |
| 2897 | up to those that are lower than the next entry \-\- e.g., 16= covers |
| 2898 | depths from 16 to 31. Note that the range covered by a depth |
| 2899 | distribution entry can be different to the range covered by the |
| 2900 | equivalent \fBsubmit\fR/\fBcomplete\fR distribution entry. |
| 2901 | .TP |
| 2902 | .B IO submit |
| 2903 | How many pieces of I/O were submitting in a single submit call. Each |
| 2904 | entry denotes that amount and below, until the previous entry \-\- e.g., |
| 2905 | 16=100% means that we submitted anywhere between 9 to 16 I/Os per submit |
| 2906 | call. Note that the range covered by a \fBsubmit\fR distribution entry can |
| 2907 | be different to the range covered by the equivalent depth distribution |
| 2908 | entry. |
| 2909 | .TP |
| 2910 | .B IO complete |
| 2911 | Like the above \fBsubmit\fR number, but for completions instead. |
| 2912 | .TP |
| 2913 | .B IO issued rwt |
| 2914 | The number of \fBread/write/trim\fR requests issued, and how many of them were |
| 2915 | short or dropped. |
| 2916 | .TP |
| 2917 | .B IO latency |
| 2918 | These values are for \fBlatency-target\fR and related options. When |
| 2919 | these options are engaged, this section describes the I/O depth required |
| 2920 | to meet the specified latency target. |
| 2921 | .RE |
| 2922 | .P |
| 2923 | After each client has been listed, the group statistics are printed. They |
| 2924 | will look like this: |
| 2925 | .P |
| 2926 | .nf |
| 2927 | Run status group 0 (all jobs): |
| 2928 | 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 |
| 2929 | WRITE: bw=1231KiB/s (1261kB/s), 616KiB/s\-621KiB/s (630kB/s\-636kB/s), io=64.0MiB (67.1MB), run=52747\-53223msec |
| 2930 | .fi |
| 2931 | .P |
| 2932 | For each data direction it prints: |
| 2933 | .RS |
| 2934 | .TP |
| 2935 | .B bw |
| 2936 | Aggregate bandwidth of threads in this group followed by the |
| 2937 | minimum and maximum bandwidth of all the threads in this group. |
| 2938 | Values outside of brackets are power\-of\-2 format and those |
| 2939 | within are the equivalent value in a power\-of\-10 format. |
| 2940 | .TP |
| 2941 | .B io |
| 2942 | Aggregate I/O performed of all threads in this group. The |
| 2943 | format is the same as \fBbw\fR. |
| 2944 | .TP |
| 2945 | .B run |
| 2946 | The smallest and longest runtimes of the threads in this group. |
| 2947 | .RE |
| 2948 | .P |
| 2949 | And finally, the disk statistics are printed. This is Linux specific. |
| 2950 | They will look like this: |
| 2951 | .P |
| 2952 | .nf |
| 2953 | Disk stats (read/write): |
| 2954 | sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00% |
| 2955 | .fi |
| 2956 | .P |
| 2957 | Each value is printed for both reads and writes, with reads first. The |
| 2958 | numbers denote: |
| 2959 | .RS |
| 2960 | .TP |
| 2961 | .B ios |
| 2962 | Number of I/Os performed by all groups. |
| 2963 | .TP |
| 2964 | .B merge |
| 2965 | Number of merges performed by the I/O scheduler. |
| 2966 | .TP |
| 2967 | .B ticks |
| 2968 | Number of ticks we kept the disk busy. |
| 2969 | .TP |
| 2970 | .B in_queue |
| 2971 | Total time spent in the disk queue. |
| 2972 | .TP |
| 2973 | .B util |
| 2974 | The disk utilization. A value of 100% means we kept the disk |
| 2975 | busy constantly, 50% would be a disk idling half of the time. |
| 2976 | .RE |
| 2977 | .P |
| 2978 | It is also possible to get fio to dump the current output while it is running, |
| 2979 | without terminating the job. To do that, send fio the USR1 signal. You can |
| 2980 | also get regularly timed dumps by using the \fB\-\-status\-interval\fR |
| 2981 | parameter, or by creating a file in `/tmp' named |
| 2982 | `fio\-dump\-status'. If fio sees this file, it will unlink it and dump the |
| 2983 | current output status. |
| 2984 | .SH TERSE OUTPUT |
| 2985 | For scripted usage where you typically want to generate tables or graphs of the |
| 2986 | results, fio can output the results in a semicolon separated format. The format |
| 2987 | is one long line of values, such as: |
| 2988 | .P |
| 2989 | .nf |
| 2990 | 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% |
| 2991 | A description of this job goes here. |
| 2992 | .fi |
| 2993 | .P |
| 2994 | The job description (if provided) follows on a second line. |
| 2995 | .P |
| 2996 | To enable terse output, use the \fB\-\-minimal\fR or |
| 2997 | `\-\-output\-format=terse' command line options. The |
| 2998 | first value is the version of the terse output format. If the output has to be |
| 2999 | changed for some reason, this number will be incremented by 1 to signify that |
| 3000 | change. |
| 3001 | .P |
| 3002 | Split up, the format is as follows (comments in brackets denote when a |
| 3003 | field was introduced or whether it's specific to some terse version): |
| 3004 | .P |
| 3005 | .nf |
| 3006 | terse version, fio version [v3], jobname, groupid, error |
| 3007 | .fi |
| 3008 | .RS |
| 3009 | .P |
| 3010 | .B |
| 3011 | READ status: |
| 3012 | .RE |
| 3013 | .P |
| 3014 | .nf |
| 3015 | Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec) |
| 3016 | Submission latency: min, max, mean, stdev (usec) |
| 3017 | Completion latency: min, max, mean, stdev (usec) |
| 3018 | Completion latency percentiles: 20 fields (see below) |
| 3019 | Total latency: min, max, mean, stdev (usec) |
| 3020 | Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5] |
| 3021 | IOPS [v5]: min, max, mean, stdev, number of samples |
| 3022 | .fi |
| 3023 | .RS |
| 3024 | .P |
| 3025 | .B |
| 3026 | WRITE status: |
| 3027 | .RE |
| 3028 | .P |
| 3029 | .nf |
| 3030 | Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec) |
| 3031 | Submission latency: min, max, mean, stdev (usec) |
| 3032 | Completion latency: min, max, mean, stdev (usec) |
| 3033 | Completion latency percentiles: 20 fields (see below) |
| 3034 | Total latency: min, max, mean, stdev (usec) |
| 3035 | Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5] |
| 3036 | IOPS [v5]: min, max, mean, stdev, number of samples |
| 3037 | .fi |
| 3038 | .RS |
| 3039 | .P |
| 3040 | .B |
| 3041 | TRIM status [all but version 3]: |
| 3042 | .RE |
| 3043 | .P |
| 3044 | .nf |
| 3045 | Fields are similar to \fBREAD/WRITE\fR status. |
| 3046 | .fi |
| 3047 | .RS |
| 3048 | .P |
| 3049 | .B |
| 3050 | CPU usage: |
| 3051 | .RE |
| 3052 | .P |
| 3053 | .nf |
| 3054 | user, system, context switches, major faults, minor faults |
| 3055 | .fi |
| 3056 | .RS |
| 3057 | .P |
| 3058 | .B |
| 3059 | I/O depths: |
| 3060 | .RE |
| 3061 | .P |
| 3062 | .nf |
| 3063 | <=1, 2, 4, 8, 16, 32, >=64 |
| 3064 | .fi |
| 3065 | .RS |
| 3066 | .P |
| 3067 | .B |
| 3068 | I/O latencies microseconds: |
| 3069 | .RE |
| 3070 | .P |
| 3071 | .nf |
| 3072 | <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000 |
| 3073 | .fi |
| 3074 | .RS |
| 3075 | .P |
| 3076 | .B |
| 3077 | I/O latencies milliseconds: |
| 3078 | .RE |
| 3079 | .P |
| 3080 | .nf |
| 3081 | <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000 |
| 3082 | .fi |
| 3083 | .RS |
| 3084 | .P |
| 3085 | .B |
| 3086 | Disk utilization [v3]: |
| 3087 | .RE |
| 3088 | .P |
| 3089 | .nf |
| 3090 | disk name, read ios, write ios, read merges, write merges, read ticks, write ticks, time spent in queue, disk utilization percentage |
| 3091 | .fi |
| 3092 | .RS |
| 3093 | .P |
| 3094 | .B |
| 3095 | Additional Info (dependent on continue_on_error, default off): |
| 3096 | .RE |
| 3097 | .P |
| 3098 | .nf |
| 3099 | total # errors, first error code |
| 3100 | .fi |
| 3101 | .RS |
| 3102 | .P |
| 3103 | .B |
| 3104 | Additional Info (dependent on description being set): |
| 3105 | .RE |
| 3106 | .P |
| 3107 | .nf |
| 3108 | Text description |
| 3109 | .fi |
| 3110 | .P |
| 3111 | Completion latency percentiles can be a grouping of up to 20 sets, so for the |
| 3112 | terse output fio writes all of them. Each field will look like this: |
| 3113 | .P |
| 3114 | .nf |
| 3115 | 1.00%=6112 |
| 3116 | .fi |
| 3117 | .P |
| 3118 | which is the Xth percentile, and the `usec' latency associated with it. |
| 3119 | .P |
| 3120 | For \fBDisk utilization\fR, all disks used by fio are shown. So for each disk there |
| 3121 | will be a disk utilization section. |
| 3122 | .P |
| 3123 | Below is a single line containing short names for each of the fields in the |
| 3124 | minimal output v3, separated by semicolons: |
| 3125 | .P |
| 3126 | .nf |
| 3127 | terse_version_3;fio_version;jobname;groupid;error;read_kb;read_bandwidth;read_iops;read_runtime_ms;read_slat_min;read_slat_max;read_slat_mean;read_slat_dev;read_clat_min;read_clat_max;read_clat_mean;read_clat_dev;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;read_lat_max;read_lat_mean;read_lat_dev;read_bw_min;read_bw_max;read_bw_agg_pct;read_bw_mean;read_bw_dev;write_kb;write_bandwidth;write_iops;write_runtime_ms;write_slat_min;write_slat_max;write_slat_mean;write_slat_dev;write_clat_min;write_clat_max;write_clat_mean;write_clat_dev;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;write_lat_max;write_lat_mean;write_lat_dev;write_bw_min;write_bw_max;write_bw_agg_pct;write_bw_mean;write_bw_dev;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 |
| 3128 | .fi |
| 3129 | .SH JSON OUTPUT |
| 3130 | The \fBjson\fR output format is intended to be both human readable and convenient |
| 3131 | for automated parsing. For the most part its sections mirror those of the |
| 3132 | \fBnormal\fR output. The \fBruntime\fR value is reported in msec and the \fBbw\fR value is |
| 3133 | reported in 1024 bytes per second units. |
| 3134 | .fi |
| 3135 | .SH JSON+ OUTPUT |
| 3136 | The \fBjson+\fR output format is identical to the \fBjson\fR output format except that it |
| 3137 | adds a full dump of the completion latency bins. Each \fBbins\fR object contains a |
| 3138 | set of (key, value) pairs where keys are latency durations and values count how |
| 3139 | many I/Os had completion latencies of the corresponding duration. For example, |
| 3140 | consider: |
| 3141 | .RS |
| 3142 | .P |
| 3143 | "bins" : { "87552" : 1, "89600" : 1, "94720" : 1, "96768" : 1, "97792" : 1, "99840" : 1, "100864" : 2, "103936" : 6, "104960" : 534, "105984" : 5995, "107008" : 7529, ... } |
| 3144 | .RE |
| 3145 | .P |
| 3146 | This data indicates that one I/O required 87,552ns to complete, two I/Os required |
| 3147 | 100,864ns to complete, and 7529 I/Os required 107,008ns to complete. |
| 3148 | .P |
| 3149 | Also included with fio is a Python script \fBfio_jsonplus_clat2csv\fR that takes |
| 3150 | json+ output and generates CSV\-formatted latency data suitable for plotting. |
| 3151 | .P |
| 3152 | The latency durations actually represent the midpoints of latency intervals. |
| 3153 | For details refer to `stat.h' in the fio source. |
| 3154 | .SH TRACE FILE FORMAT |
| 3155 | There are two trace file format that you can encounter. The older (v1) format is |
| 3156 | unsupported since version 1.20\-rc3 (March 2008). It will still be described |
| 3157 | below in case that you get an old trace and want to understand it. |
| 3158 | .P |
| 3159 | In any case the trace is a simple text file with a single action per line. |
| 3160 | .TP |
| 3161 | .B Trace file format v1 |
| 3162 | Each line represents a single I/O action in the following format: |
| 3163 | .RS |
| 3164 | .RS |
| 3165 | .P |
| 3166 | rw, offset, length |
| 3167 | .RE |
| 3168 | .P |
| 3169 | where `rw=0/1' for read/write, and the `offset' and `length' entries being in bytes. |
| 3170 | .P |
| 3171 | This format is not supported in fio versions >= 1.20\-rc3. |
| 3172 | .RE |
| 3173 | .TP |
| 3174 | .B Trace file format v2 |
| 3175 | The second version of the trace file format was added in fio version 1.17. It |
| 3176 | allows to access more then one file per trace and has a bigger set of possible |
| 3177 | file actions. |
| 3178 | .RS |
| 3179 | .P |
| 3180 | The first line of the trace file has to be: |
| 3181 | .RS |
| 3182 | .P |
| 3183 | "fio version 2 iolog" |
| 3184 | .RE |
| 3185 | .P |
| 3186 | Following this can be lines in two different formats, which are described below. |
| 3187 | .P |
| 3188 | .B |
| 3189 | The file management format: |
| 3190 | .RS |
| 3191 | filename action |
| 3192 | .P |
| 3193 | The `filename' is given as an absolute path. The `action' can be one of these: |
| 3194 | .RS |
| 3195 | .TP |
| 3196 | .B add |
| 3197 | Add the given `filename' to the trace. |
| 3198 | .TP |
| 3199 | .B open |
| 3200 | Open the file with the given `filename'. The `filename' has to have |
| 3201 | been added with the \fBadd\fR action before. |
| 3202 | .TP |
| 3203 | .B close |
| 3204 | Close the file with the given `filename'. The file has to have been |
| 3205 | \fBopen\fRed before. |
| 3206 | .RE |
| 3207 | .RE |
| 3208 | .P |
| 3209 | .B |
| 3210 | The file I/O action format: |
| 3211 | .RS |
| 3212 | filename action offset length |
| 3213 | .P |
| 3214 | The `filename' is given as an absolute path, and has to have been \fBadd\fRed and |
| 3215 | \fBopen\fRed before it can be used with this format. The `offset' and `length' are |
| 3216 | given in bytes. The `action' can be one of these: |
| 3217 | .RS |
| 3218 | .TP |
| 3219 | .B wait |
| 3220 | Wait for `offset' microseconds. Everything below 100 is discarded. |
| 3221 | The time is relative to the previous `wait' statement. |
| 3222 | .TP |
| 3223 | .B read |
| 3224 | Read `length' bytes beginning from `offset'. |
| 3225 | .TP |
| 3226 | .B write |
| 3227 | Write `length' bytes beginning from `offset'. |
| 3228 | .TP |
| 3229 | .B sync |
| 3230 | \fBfsync\fR\|(2) the file. |
| 3231 | .TP |
| 3232 | .B datasync |
| 3233 | \fBfdatasync\fR\|(2) the file. |
| 3234 | .TP |
| 3235 | .B trim |
| 3236 | Trim the given file from the given `offset' for `length' bytes. |
| 3237 | .RE |
| 3238 | .RE |
| 3239 | .SH CPU IDLENESS PROFILING |
| 3240 | In some cases, we want to understand CPU overhead in a test. For example, we |
| 3241 | test patches for the specific goodness of whether they reduce CPU usage. |
| 3242 | Fio implements a balloon approach to create a thread per CPU that runs at idle |
| 3243 | priority, meaning that it only runs when nobody else needs the cpu. |
| 3244 | By measuring the amount of work completed by the thread, idleness of each CPU |
| 3245 | can be derived accordingly. |
| 3246 | .P |
| 3247 | An unit work is defined as touching a full page of unsigned characters. Mean and |
| 3248 | standard deviation of time to complete an unit work is reported in "unit work" |
| 3249 | section. Options can be chosen to report detailed percpu idleness or overall |
| 3250 | system idleness by aggregating percpu stats. |
| 3251 | .SH VERIFICATION AND TRIGGERS |
| 3252 | Fio is usually run in one of two ways, when data verification is done. The first |
| 3253 | is a normal write job of some sort with verify enabled. When the write phase has |
| 3254 | completed, fio switches to reads and verifies everything it wrote. The second |
| 3255 | model is running just the write phase, and then later on running the same job |
| 3256 | (but with reads instead of writes) to repeat the same I/O patterns and verify |
| 3257 | the contents. Both of these methods depend on the write phase being completed, |
| 3258 | as fio otherwise has no idea how much data was written. |
| 3259 | .P |
| 3260 | With verification triggers, fio supports dumping the current write state to |
| 3261 | local files. Then a subsequent read verify workload can load this state and know |
| 3262 | exactly where to stop. This is useful for testing cases where power is cut to a |
| 3263 | server in a managed fashion, for instance. |
| 3264 | .P |
| 3265 | A verification trigger consists of two things: |
| 3266 | .RS |
| 3267 | .P |
| 3268 | 1) Storing the write state of each job. |
| 3269 | .P |
| 3270 | 2) Executing a trigger command. |
| 3271 | .RE |
| 3272 | .P |
| 3273 | The write state is relatively small, on the order of hundreds of bytes to single |
| 3274 | kilobytes. It contains information on the number of completions done, the last X |
| 3275 | completions, etc. |
| 3276 | .P |
| 3277 | A trigger is invoked either through creation ('touch') of a specified file in |
| 3278 | the system, or through a timeout setting. If fio is run with |
| 3279 | `\-\-trigger\-file=/tmp/trigger\-file', then it will continually |
| 3280 | check for the existence of `/tmp/trigger\-file'. When it sees this file, it |
| 3281 | will fire off the trigger (thus saving state, and executing the trigger |
| 3282 | command). |
| 3283 | .P |
| 3284 | For client/server runs, there's both a local and remote trigger. If fio is |
| 3285 | running as a server backend, it will send the job states back to the client for |
| 3286 | safe storage, then execute the remote trigger, if specified. If a local trigger |
| 3287 | is specified, the server will still send back the write state, but the client |
| 3288 | will then execute the trigger. |
| 3289 | .RE |
| 3290 | .P |
| 3291 | .B Verification trigger example |
| 3292 | .RS |
| 3293 | Let's say we want to run a powercut test on the remote Linux machine 'server'. |
| 3294 | Our write workload is in `write\-test.fio'. We want to cut power to 'server' at |
| 3295 | some point during the run, and we'll run this test from the safety or our local |
| 3296 | machine, 'localbox'. On the server, we'll start the fio backend normally: |
| 3297 | .RS |
| 3298 | .P |
| 3299 | server# fio \-\-server |
| 3300 | .RE |
| 3301 | .P |
| 3302 | and on the client, we'll fire off the workload: |
| 3303 | .RS |
| 3304 | .P |
| 3305 | localbox$ fio \-\-client=server \-\-trigger\-file=/tmp/my\-trigger \-\-trigger\-remote="bash \-c "echo b > /proc/sysrq\-triger"" |
| 3306 | .RE |
| 3307 | .P |
| 3308 | We set `/tmp/my\-trigger' as the trigger file, and we tell fio to execute: |
| 3309 | .RS |
| 3310 | .P |
| 3311 | echo b > /proc/sysrq\-trigger |
| 3312 | .RE |
| 3313 | .P |
| 3314 | on the server once it has received the trigger and sent us the write state. This |
| 3315 | will work, but it's not really cutting power to the server, it's merely |
| 3316 | abruptly rebooting it. If we have a remote way of cutting power to the server |
| 3317 | through IPMI or similar, we could do that through a local trigger command |
| 3318 | instead. Let's assume we have a script that does IPMI reboot of a given hostname, |
| 3319 | ipmi\-reboot. On localbox, we could then have run fio with a local trigger |
| 3320 | instead: |
| 3321 | .RS |
| 3322 | .P |
| 3323 | localbox$ fio \-\-client=server \-\-trigger\-file=/tmp/my\-trigger \-\-trigger="ipmi\-reboot server" |
| 3324 | .RE |
| 3325 | .P |
| 3326 | For this case, fio would wait for the server to send us the write state, then |
| 3327 | execute `ipmi\-reboot server' when that happened. |
| 3328 | .RE |
| 3329 | .P |
| 3330 | .B Loading verify state |
| 3331 | .RS |
| 3332 | To load stored write state, a read verification job file must contain the |
| 3333 | \fBverify_state_load\fR option. If that is set, fio will load the previously |
| 3334 | stored state. For a local fio run this is done by loading the files directly, |
| 3335 | and on a client/server run, the server backend will ask the client to send the |
| 3336 | files over and load them from there. |
| 3337 | .RE |
| 3338 | .SH LOG FILE FORMATS |
| 3339 | Fio supports a variety of log file formats, for logging latencies, bandwidth, |
| 3340 | and IOPS. The logs share a common format, which looks like this: |
| 3341 | .RS |
| 3342 | .P |
| 3343 | time (msec), value, data direction, block size (bytes), offset (bytes) |
| 3344 | .RE |
| 3345 | .P |
| 3346 | `Time' for the log entry is always in milliseconds. The `value' logged depends |
| 3347 | on the type of log, it will be one of the following: |
| 3348 | .RS |
| 3349 | .TP |
| 3350 | .B Latency log |
| 3351 | Value is latency in nsecs |
| 3352 | .TP |
| 3353 | .B Bandwidth log |
| 3354 | Value is in KiB/sec |
| 3355 | .TP |
| 3356 | .B IOPS log |
| 3357 | Value is IOPS |
| 3358 | .RE |
| 3359 | .P |
| 3360 | `Data direction' is one of the following: |
| 3361 | .RS |
| 3362 | .TP |
| 3363 | .B 0 |
| 3364 | I/O is a READ |
| 3365 | .TP |
| 3366 | .B 1 |
| 3367 | I/O is a WRITE |
| 3368 | .TP |
| 3369 | .B 2 |
| 3370 | I/O is a TRIM |
| 3371 | .RE |
| 3372 | .P |
| 3373 | The entry's `block size' is always in bytes. The `offset' is the offset, in bytes, |
| 3374 | from the start of the file, for that particular I/O. The logging of the offset can be |
| 3375 | toggled with \fBlog_offset\fR. |
| 3376 | .P |
| 3377 | Fio defaults to logging every individual I/O. When IOPS are logged for individual |
| 3378 | I/Os the `value' entry will always be 1. If windowed logging is enabled through |
| 3379 | \fBlog_avg_msec\fR, fio logs the average values over the specified period of time. |
| 3380 | If windowed logging is enabled and \fBlog_max_value\fR is set, then fio logs |
| 3381 | maximum values in that window instead of averages. Since `data direction', `block size' |
| 3382 | and `offset' are per\-I/O values, if windowed logging is enabled they |
| 3383 | aren't applicable and will be 0. |
| 3384 | .SH CLIENT / SERVER |
| 3385 | Normally fio is invoked as a stand\-alone application on the machine where the |
| 3386 | I/O workload should be generated. However, the backend and frontend of fio can |
| 3387 | be run separately i.e., the fio server can generate an I/O workload on the "Device |
| 3388 | Under Test" while being controlled by a client on another machine. |
| 3389 | .P |
| 3390 | Start the server on the machine which has access to the storage DUT: |
| 3391 | .RS |
| 3392 | .P |
| 3393 | $ fio \-\-server=args |
| 3394 | .RE |
| 3395 | .P |
| 3396 | where `args' defines what fio listens to. The arguments are of the form |
| 3397 | `type,hostname' or `IP,port'. `type' is either `ip' (or ip4) for TCP/IP |
| 3398 | v4, `ip6' for TCP/IP v6, or `sock' for a local unix domain socket. |
| 3399 | `hostname' is either a hostname or IP address, and `port' is the port to listen |
| 3400 | to (only valid for TCP/IP, not a local socket). Some examples: |
| 3401 | .RS |
| 3402 | .TP |
| 3403 | 1) \fBfio \-\-server\fR |
| 3404 | Start a fio server, listening on all interfaces on the default port (8765). |
| 3405 | .TP |
| 3406 | 2) \fBfio \-\-server=ip:hostname,4444\fR |
| 3407 | Start a fio server, listening on IP belonging to hostname and on port 4444. |
| 3408 | .TP |
| 3409 | 3) \fBfio \-\-server=ip6:::1,4444\fR |
| 3410 | Start a fio server, listening on IPv6 localhost ::1 and on port 4444. |
| 3411 | .TP |
| 3412 | 4) \fBfio \-\-server=,4444\fR |
| 3413 | Start a fio server, listening on all interfaces on port 4444. |
| 3414 | .TP |
| 3415 | 5) \fBfio \-\-server=1.2.3.4\fR |
| 3416 | Start a fio server, listening on IP 1.2.3.4 on the default port. |
| 3417 | .TP |
| 3418 | 6) \fBfio \-\-server=sock:/tmp/fio.sock\fR |
| 3419 | Start a fio server, listening on the local socket `/tmp/fio.sock'. |
| 3420 | .RE |
| 3421 | .P |
| 3422 | Once a server is running, a "client" can connect to the fio server with: |
| 3423 | .RS |
| 3424 | .P |
| 3425 | $ fio <local\-args> \-\-client=<server> <remote\-args> <job file(s)> |
| 3426 | .RE |
| 3427 | .P |
| 3428 | where `local\-args' are arguments for the client where it is running, `server' |
| 3429 | is the connect string, and `remote\-args' and `job file(s)' are sent to the |
| 3430 | server. The `server' string follows the same format as it does on the server |
| 3431 | side, to allow IP/hostname/socket and port strings. |
| 3432 | .P |
| 3433 | Fio can connect to multiple servers this way: |
| 3434 | .RS |
| 3435 | .P |
| 3436 | $ fio \-\-client=<server1> <job file(s)> \-\-client=<server2> <job file(s)> |
| 3437 | .RE |
| 3438 | .P |
| 3439 | If the job file is located on the fio server, then you can tell the server to |
| 3440 | load a local file as well. This is done by using \fB\-\-remote\-config\fR: |
| 3441 | .RS |
| 3442 | .P |
| 3443 | $ fio \-\-client=server \-\-remote\-config /path/to/file.fio |
| 3444 | .RE |
| 3445 | .P |
| 3446 | Then fio will open this local (to the server) job file instead of being passed |
| 3447 | one from the client. |
| 3448 | .P |
| 3449 | If you have many servers (example: 100 VMs/containers), you can input a pathname |
| 3450 | of a file containing host IPs/names as the parameter value for the |
| 3451 | \fB\-\-client\fR option. For example, here is an example `host.list' |
| 3452 | file containing 2 hostnames: |
| 3453 | .RS |
| 3454 | .P |
| 3455 | .PD 0 |
| 3456 | host1.your.dns.domain |
| 3457 | .P |
| 3458 | host2.your.dns.domain |
| 3459 | .PD |
| 3460 | .RE |
| 3461 | .P |
| 3462 | The fio command would then be: |
| 3463 | .RS |
| 3464 | .P |
| 3465 | $ fio \-\-client=host.list <job file(s)> |
| 3466 | .RE |
| 3467 | .P |
| 3468 | In this mode, you cannot input server\-specific parameters or job files \-\- all |
| 3469 | servers receive the same job file. |
| 3470 | .P |
| 3471 | In order to let `fio \-\-client' runs use a shared filesystem from multiple |
| 3472 | hosts, `fio \-\-client' now prepends the IP address of the server to the |
| 3473 | filename. For example, if fio is using the directory `/mnt/nfs/fio' and is |
| 3474 | writing filename `fileio.tmp', with a \fB\-\-client\fR `hostfile' |
| 3475 | containing two hostnames `h1' and `h2' with IP addresses 192.168.10.120 and |
| 3476 | 192.168.10.121, then fio will create two files: |
| 3477 | .RS |
| 3478 | .P |
| 3479 | .PD 0 |
| 3480 | /mnt/nfs/fio/192.168.10.120.fileio.tmp |
| 3481 | .P |
| 3482 | /mnt/nfs/fio/192.168.10.121.fileio.tmp |
| 3483 | .PD |
| 3484 | .RE |
| 3485 | .SH AUTHORS |
| 3486 | .B fio |
| 3487 | was written by Jens Axboe <jens.axboe@oracle.com>, |
| 3488 | now Jens Axboe <axboe@fb.com>. |
| 3489 | .br |
| 3490 | This man page was written by Aaron Carroll <aaronc@cse.unsw.edu.au> based |
| 3491 | on documentation by Jens Axboe. |
| 3492 | .br |
| 3493 | This man page was rewritten by Tomohiro Kusumi <tkusumi@tuxera.com> based |
| 3494 | on documentation by Jens Axboe. |
| 3495 | .SH "REPORTING BUGS" |
| 3496 | Report bugs to the \fBfio\fR mailing list <fio@vger.kernel.org>. |
| 3497 | .br |
| 3498 | See \fBREPORTING\-BUGS\fR. |
| 3499 | .P |
| 3500 | \fBREPORTING\-BUGS\fR: \fIhttp://git.kernel.dk/cgit/fio/plain/REPORTING\-BUGS\fR |
| 3501 | .SH "SEE ALSO" |
| 3502 | For further documentation see \fBHOWTO\fR and \fBREADME\fR. |
| 3503 | .br |
| 3504 | Sample jobfiles are available in the `examples/' directory. |
| 3505 | .br |
| 3506 | These are typically located under `/usr/share/doc/fio'. |
| 3507 | .P |
| 3508 | \fBHOWTO\fR: \fIhttp://git.kernel.dk/cgit/fio/plain/HOWTO\fR |
| 3509 | .br |
| 3510 | \fBREADME\fR: \fIhttp://git.kernel.dk/cgit/fio/plain/README\fR |