| 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 \-\-merge\-blktrace\-only |
| 24 | Merge blktraces only, don't start any I/O. |
| 25 | .TP |
| 26 | .BI \-\-output \fR=\fPfilename |
| 27 | Write output to \fIfilename\fR. |
| 28 | .TP |
| 29 | .BI \-\-output\-format \fR=\fPformat |
| 30 | Set the reporting \fIformat\fR to `normal', `terse', `json', or |
| 31 | `json+'. Multiple formats can be selected, separate by a comma. `terse' |
| 32 | is a CSV based format. `json+' is like `json', except it adds a full |
| 33 | dump of the latency buckets. |
| 34 | .TP |
| 35 | .BI \-\-bandwidth\-log |
| 36 | Generate aggregate bandwidth logs. |
| 37 | .TP |
| 38 | .BI \-\-minimal |
| 39 | Print statistics in a terse, semicolon\-delimited format. |
| 40 | .TP |
| 41 | .BI \-\-append\-terse |
| 42 | Print statistics in selected mode AND terse, semicolon\-delimited format. |
| 43 | \fBDeprecated\fR, use \fB\-\-output\-format\fR instead to select multiple formats. |
| 44 | .TP |
| 45 | .BI \-\-terse\-version \fR=\fPversion |
| 46 | Set terse \fIversion\fR output format (default `3', or `2', `4', `5'). |
| 47 | .TP |
| 48 | .BI \-\-version |
| 49 | Print version information and exit. |
| 50 | .TP |
| 51 | .BI \-\-help |
| 52 | Print a summary of the command line options and exit. |
| 53 | .TP |
| 54 | .BI \-\-cpuclock\-test |
| 55 | Perform test and validation of internal CPU clock. |
| 56 | .TP |
| 57 | .BI \-\-crctest \fR=\fP[test] |
| 58 | Test the speed of the built\-in checksumming functions. If no argument is given, |
| 59 | all of them are tested. Alternatively, a comma separated list can be passed, in which |
| 60 | case the given ones are tested. |
| 61 | .TP |
| 62 | .BI \-\-cmdhelp \fR=\fPcommand |
| 63 | Print help information for \fIcommand\fR. May be `all' for all commands. |
| 64 | .TP |
| 65 | .BI \-\-enghelp \fR=\fP[ioengine[,command]] |
| 66 | List all commands defined by \fIioengine\fR, or print help for \fIcommand\fR |
| 67 | defined by \fIioengine\fR. If no \fIioengine\fR is given, list all |
| 68 | available ioengines. |
| 69 | .TP |
| 70 | .BI \-\-showcmd |
| 71 | Convert given \fIjobfile\fRs to a set of command\-line options. |
| 72 | .TP |
| 73 | .BI \-\-readonly |
| 74 | Turn on safety read\-only checks, preventing writes and trims. The \fB\-\-readonly\fR |
| 75 | option is an extra safety guard to prevent users from accidentally starting |
| 76 | a write or trim workload when that is not desired. Fio will only modify the |
| 77 | device under test if `rw=write/randwrite/rw/randrw/trim/randtrim/trimwrite' |
| 78 | is given. This safety net can be used as an extra precaution. |
| 79 | .TP |
| 80 | .BI \-\-eta \fR=\fPwhen |
| 81 | Specifies when real\-time ETA estimate should be printed. \fIwhen\fR may |
| 82 | be `always', `never' or `auto'. `auto' is the default, it prints ETA when |
| 83 | requested if the output is a TTY. `always' disregards the output type, and |
| 84 | prints ETA when requested. `never' never prints ETA. |
| 85 | .TP |
| 86 | .BI \-\-eta\-interval \fR=\fPtime |
| 87 | By default, fio requests client ETA status roughly every second. With this |
| 88 | option, the interval is configurable. Fio imposes a minimum allowed time to |
| 89 | avoid flooding the console, less than 250 msec is not supported. |
| 90 | .TP |
| 91 | .BI \-\-eta\-newline \fR=\fPtime |
| 92 | Force a new line for every \fItime\fR period passed. When the unit is omitted, |
| 93 | the value is interpreted in seconds. |
| 94 | .TP |
| 95 | .BI \-\-status\-interval \fR=\fPtime |
| 96 | Force a full status dump of cumulative (from job start) values at \fItime\fR |
| 97 | intervals. This option does *not* provide per-period measurements. So |
| 98 | values such as bandwidth are running averages. When the time unit is omitted, |
| 99 | \fItime\fR is interpreted in seconds. Note that using this option with |
| 100 | `\-\-output-format=json' will yield output that technically isn't valid json, |
| 101 | since the output will be collated sets of valid json. It will need to be split |
| 102 | into valid sets of json after the run. |
| 103 | .TP |
| 104 | .BI \-\-section \fR=\fPname |
| 105 | Only run specified section \fIname\fR in job file. Multiple sections can be specified. |
| 106 | The \fB\-\-section\fR option allows one to combine related jobs into one file. |
| 107 | E.g. one job file could define light, moderate, and heavy sections. Tell |
| 108 | fio to run only the "heavy" section by giving `\-\-section=heavy' |
| 109 | command line option. One can also specify the "write" operations in one |
| 110 | section and "verify" operation in another section. The \fB\-\-section\fR option |
| 111 | only applies to job sections. The reserved *global* section is always |
| 112 | parsed and used. |
| 113 | .TP |
| 114 | .BI \-\-alloc\-size \fR=\fPkb |
| 115 | Allocate additional internal smalloc pools of size \fIkb\fR in KiB. The |
| 116 | \fB\-\-alloc\-size\fR option increases shared memory set aside for use by fio. |
| 117 | If running large jobs with randommap enabled, fio can run out of memory. |
| 118 | Smalloc is an internal allocator for shared structures from a fixed size |
| 119 | memory pool and can grow to 16 pools. The pool size defaults to 16MiB. |
| 120 | NOTE: While running `.fio_smalloc.*' backing store files are visible |
| 121 | in `/tmp'. |
| 122 | .TP |
| 123 | .BI \-\-warnings\-fatal |
| 124 | All fio parser warnings are fatal, causing fio to exit with an error. |
| 125 | .TP |
| 126 | .BI \-\-max\-jobs \fR=\fPnr |
| 127 | Set the maximum number of threads/processes to support to \fInr\fR. |
| 128 | NOTE: On Linux, it may be necessary to increase the shared-memory limit |
| 129 | (`/proc/sys/kernel/shmmax') if fio runs into errors while creating jobs. |
| 130 | .TP |
| 131 | .BI \-\-server \fR=\fPargs |
| 132 | Start a backend server, with \fIargs\fR specifying what to listen to. |
| 133 | See \fBCLIENT/SERVER\fR section. |
| 134 | .TP |
| 135 | .BI \-\-daemonize \fR=\fPpidfile |
| 136 | Background a fio server, writing the pid to the given \fIpidfile\fR file. |
| 137 | .TP |
| 138 | .BI \-\-client \fR=\fPhostname |
| 139 | Instead of running the jobs locally, send and run them on the given \fIhostname\fR |
| 140 | or set of \fIhostname\fRs. See \fBCLIENT/SERVER\fR section. |
| 141 | .TP |
| 142 | .BI \-\-remote\-config \fR=\fPfile |
| 143 | Tell fio server to load this local \fIfile\fR. |
| 144 | .TP |
| 145 | .BI \-\-idle\-prof \fR=\fPoption |
| 146 | Report CPU idleness. \fIoption\fR is one of the following: |
| 147 | .RS |
| 148 | .RS |
| 149 | .TP |
| 150 | .B calibrate |
| 151 | Run unit work calibration only and exit. |
| 152 | .TP |
| 153 | .B system |
| 154 | Show aggregate system idleness and unit work. |
| 155 | .TP |
| 156 | .B percpu |
| 157 | As \fBsystem\fR but also show per CPU idleness. |
| 158 | .RE |
| 159 | .RE |
| 160 | .TP |
| 161 | .BI \-\-inflate\-log \fR=\fPlog |
| 162 | Inflate and output compressed \fIlog\fR. |
| 163 | .TP |
| 164 | .BI \-\-trigger\-file \fR=\fPfile |
| 165 | Execute trigger command when \fIfile\fR exists. |
| 166 | .TP |
| 167 | .BI \-\-trigger\-timeout \fR=\fPtime |
| 168 | Execute trigger at this \fItime\fR. |
| 169 | .TP |
| 170 | .BI \-\-trigger \fR=\fPcommand |
| 171 | Set this \fIcommand\fR as local trigger. |
| 172 | .TP |
| 173 | .BI \-\-trigger\-remote \fR=\fPcommand |
| 174 | Set this \fIcommand\fR as remote trigger. |
| 175 | .TP |
| 176 | .BI \-\-aux\-path \fR=\fPpath |
| 177 | Use the directory specified by \fIpath\fP for generated state files instead |
| 178 | of the current working directory. |
| 179 | .SH "JOB FILE FORMAT" |
| 180 | Any parameters following the options will be assumed to be job files, unless |
| 181 | they match a job file parameter. Multiple job files can be listed and each job |
| 182 | file will be regarded as a separate group. Fio will \fBstonewall\fR execution |
| 183 | between each group. |
| 184 | |
| 185 | Fio accepts one or more job files describing what it is |
| 186 | supposed to do. The job file format is the classic ini file, where the names |
| 187 | enclosed in [] brackets define the job name. You are free to use any ASCII name |
| 188 | you want, except *global* which has special meaning. Following the job name is |
| 189 | a sequence of zero or more parameters, one per line, that define the behavior of |
| 190 | the job. If the first character in a line is a ';' or a '#', the entire line is |
| 191 | discarded as a comment. |
| 192 | |
| 193 | A *global* section sets defaults for the jobs described in that file. A job may |
| 194 | override a *global* section parameter, and a job file may even have several |
| 195 | *global* sections if so desired. A job is only affected by a *global* section |
| 196 | residing above it. |
| 197 | |
| 198 | The \fB\-\-cmdhelp\fR option also lists all options. If used with an \fIcommand\fR |
| 199 | argument, \fB\-\-cmdhelp\fR will detail the given \fIcommand\fR. |
| 200 | |
| 201 | See the `examples/' directory for inspiration on how to write job files. Note |
| 202 | the copyright and license requirements currently apply to |
| 203 | `examples/' files. |
| 204 | |
| 205 | Note that the maximum length of a line in the job file is 8192 bytes. |
| 206 | .SH "JOB FILE PARAMETERS" |
| 207 | Some parameters take an option of a given type, such as an integer or a |
| 208 | string. Anywhere a numeric value is required, an arithmetic expression may be |
| 209 | used, provided it is surrounded by parentheses. Supported operators are: |
| 210 | .RS |
| 211 | .P |
| 212 | .B addition (+) |
| 213 | .P |
| 214 | .B subtraction (\-) |
| 215 | .P |
| 216 | .B multiplication (*) |
| 217 | .P |
| 218 | .B division (/) |
| 219 | .P |
| 220 | .B modulus (%) |
| 221 | .P |
| 222 | .B exponentiation (^) |
| 223 | .RE |
| 224 | .P |
| 225 | For time values in expressions, units are microseconds by default. This is |
| 226 | different than for time values not in expressions (not enclosed in |
| 227 | parentheses). |
| 228 | .SH "PARAMETER TYPES" |
| 229 | The following parameter types are used. |
| 230 | .TP |
| 231 | .I str |
| 232 | String. A sequence of alphanumeric characters. |
| 233 | .TP |
| 234 | .I time |
| 235 | Integer with possible time suffix. Without a unit value is interpreted as |
| 236 | seconds unless otherwise specified. Accepts a suffix of 'd' for days, 'h' for |
| 237 | hours, 'm' for minutes, 's' for seconds, 'ms' (or 'msec') for milliseconds and 'us' |
| 238 | (or 'usec') for microseconds. For example, use 10m for 10 minutes. |
| 239 | .TP |
| 240 | .I int |
| 241 | Integer. A whole number value, which may contain an integer prefix |
| 242 | and an integer suffix. |
| 243 | .RS |
| 244 | .RS |
| 245 | .P |
| 246 | [*integer prefix*] **number** [*integer suffix*] |
| 247 | .RE |
| 248 | .P |
| 249 | The optional *integer prefix* specifies the number's base. The default |
| 250 | is decimal. *0x* specifies hexadecimal. |
| 251 | .P |
| 252 | The optional *integer suffix* specifies the number's units, and includes an |
| 253 | optional unit prefix and an optional unit. For quantities of data, the |
| 254 | default unit is bytes. For quantities of time, the default unit is seconds |
| 255 | unless otherwise specified. |
| 256 | .P |
| 257 | With `kb_base=1000', fio follows international standards for unit |
| 258 | prefixes. To specify power-of-10 decimal values defined in the |
| 259 | International System of Units (SI): |
| 260 | .RS |
| 261 | .P |
| 262 | .PD 0 |
| 263 | K means kilo (K) or 1000 |
| 264 | .P |
| 265 | M means mega (M) or 1000**2 |
| 266 | .P |
| 267 | G means giga (G) or 1000**3 |
| 268 | .P |
| 269 | T means tera (T) or 1000**4 |
| 270 | .P |
| 271 | P means peta (P) or 1000**5 |
| 272 | .PD |
| 273 | .RE |
| 274 | .P |
| 275 | To specify power-of-2 binary values defined in IEC 80000-13: |
| 276 | .RS |
| 277 | .P |
| 278 | .PD 0 |
| 279 | Ki means kibi (Ki) or 1024 |
| 280 | .P |
| 281 | Mi means mebi (Mi) or 1024**2 |
| 282 | .P |
| 283 | Gi means gibi (Gi) or 1024**3 |
| 284 | .P |
| 285 | Ti means tebi (Ti) or 1024**4 |
| 286 | .P |
| 287 | Pi means pebi (Pi) or 1024**5 |
| 288 | .PD |
| 289 | .RE |
| 290 | .P |
| 291 | For Zone Block Device Mode: |
| 292 | .RS |
| 293 | .P |
| 294 | .PD 0 |
| 295 | z means Zone |
| 296 | .P |
| 297 | .PD |
| 298 | .RE |
| 299 | .P |
| 300 | With `kb_base=1024' (the default), the unit prefixes are opposite |
| 301 | from those specified in the SI and IEC 80000-13 standards to provide |
| 302 | compatibility with old scripts. For example, 4k means 4096. |
| 303 | .P |
| 304 | For quantities of data, an optional unit of 'B' may be included |
| 305 | (e.g., 'kB' is the same as 'k'). |
| 306 | .P |
| 307 | The *integer suffix* is not case sensitive (e.g., m/mi mean mebi/mega, |
| 308 | not milli). 'b' and 'B' both mean byte, not bit. |
| 309 | .P |
| 310 | Examples with `kb_base=1000': |
| 311 | .RS |
| 312 | .P |
| 313 | .PD 0 |
| 314 | 4 KiB: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB |
| 315 | .P |
| 316 | 1 MiB: 1048576, 1m, 1024k |
| 317 | .P |
| 318 | 1 MB: 1000000, 1mi, 1000ki |
| 319 | .P |
| 320 | 1 TiB: 1073741824, 1t, 1024m, 1048576k |
| 321 | .P |
| 322 | 1 TB: 1000000000, 1ti, 1000mi, 1000000ki |
| 323 | .PD |
| 324 | .RE |
| 325 | .P |
| 326 | Examples with `kb_base=1024' (default): |
| 327 | .RS |
| 328 | .P |
| 329 | .PD 0 |
| 330 | 4 KiB: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB |
| 331 | .P |
| 332 | 1 MiB: 1048576, 1m, 1024k |
| 333 | .P |
| 334 | 1 MB: 1000000, 1mi, 1000ki |
| 335 | .P |
| 336 | 1 TiB: 1073741824, 1t, 1024m, 1048576k |
| 337 | .P |
| 338 | 1 TB: 1000000000, 1ti, 1000mi, 1000000ki |
| 339 | .PD |
| 340 | .RE |
| 341 | .P |
| 342 | To specify times (units are not case sensitive): |
| 343 | .RS |
| 344 | .P |
| 345 | .PD 0 |
| 346 | D means days |
| 347 | .P |
| 348 | H means hours |
| 349 | .P |
| 350 | M mean minutes |
| 351 | .P |
| 352 | s or sec means seconds (default) |
| 353 | .P |
| 354 | ms or msec means milliseconds |
| 355 | .P |
| 356 | us or usec means microseconds |
| 357 | .PD |
| 358 | .RE |
| 359 | .P |
| 360 | `z' suffix specifies that the value is measured in zones. |
| 361 | Value is recalculated once block device's zone size becomes known. |
| 362 | .P |
| 363 | If the option accepts an upper and lower range, use a colon ':' or |
| 364 | minus '\-' to separate such values. See \fIirange\fR parameter type. |
| 365 | If the lower value specified happens to be larger than the upper value |
| 366 | the two values are swapped. |
| 367 | .RE |
| 368 | .TP |
| 369 | .I bool |
| 370 | Boolean. Usually parsed as an integer, however only defined for |
| 371 | true and false (1 and 0). |
| 372 | .TP |
| 373 | .I irange |
| 374 | Integer range with suffix. Allows value range to be given, such as |
| 375 | 1024\-4096. A colon may also be used as the separator, e.g. 1k:4k. If the |
| 376 | option allows two sets of ranges, they can be specified with a ',' or '/' |
| 377 | delimiter: 1k\-4k/8k\-32k. Also see \fIint\fR parameter type. |
| 378 | .TP |
| 379 | .I float_list |
| 380 | A list of floating point numbers, separated by a ':' character. |
| 381 | .SH "JOB PARAMETERS" |
| 382 | With the above in mind, here follows the complete list of fio job parameters. |
| 383 | .SS "Units" |
| 384 | .TP |
| 385 | .BI kb_base \fR=\fPint |
| 386 | Select the interpretation of unit prefixes in input parameters. |
| 387 | .RS |
| 388 | .RS |
| 389 | .TP |
| 390 | .B 1000 |
| 391 | Inputs comply with IEC 80000-13 and the International |
| 392 | System of Units (SI). Use: |
| 393 | .RS |
| 394 | .P |
| 395 | .PD 0 |
| 396 | \- power-of-2 values with IEC prefixes (e.g., KiB) |
| 397 | .P |
| 398 | \- power-of-10 values with SI prefixes (e.g., kB) |
| 399 | .PD |
| 400 | .RE |
| 401 | .TP |
| 402 | .B 1024 |
| 403 | Compatibility mode (default). To avoid breaking old scripts: |
| 404 | .P |
| 405 | .RS |
| 406 | .PD 0 |
| 407 | \- power-of-2 values with SI prefixes |
| 408 | .P |
| 409 | \- power-of-10 values with IEC prefixes |
| 410 | .PD |
| 411 | .RE |
| 412 | .RE |
| 413 | .P |
| 414 | See \fBbs\fR for more details on input parameters. |
| 415 | .P |
| 416 | Outputs always use correct prefixes. Most outputs include both |
| 417 | side-by-side, like: |
| 418 | .P |
| 419 | .RS |
| 420 | bw=2383.3kB/s (2327.4KiB/s) |
| 421 | .RE |
| 422 | .P |
| 423 | If only one value is reported, then kb_base selects the one to use: |
| 424 | .P |
| 425 | .RS |
| 426 | .PD 0 |
| 427 | 1000 \-\- SI prefixes |
| 428 | .P |
| 429 | 1024 \-\- IEC prefixes |
| 430 | .PD |
| 431 | .RE |
| 432 | .RE |
| 433 | .TP |
| 434 | .BI unit_base \fR=\fPint |
| 435 | Base unit for reporting. Allowed values are: |
| 436 | .RS |
| 437 | .RS |
| 438 | .TP |
| 439 | .B 0 |
| 440 | Use auto-detection (default). |
| 441 | .TP |
| 442 | .B 8 |
| 443 | Byte based. |
| 444 | .TP |
| 445 | .B 1 |
| 446 | Bit based. |
| 447 | .RE |
| 448 | .RE |
| 449 | .SS "Job description" |
| 450 | .TP |
| 451 | .BI name \fR=\fPstr |
| 452 | ASCII name of the job. This may be used to override the name printed by fio |
| 453 | for this job. Otherwise the job name is used. On the command line this |
| 454 | parameter has the special purpose of also signaling the start of a new job. |
| 455 | .TP |
| 456 | .BI description \fR=\fPstr |
| 457 | Text description of the job. Doesn't do anything except dump this text |
| 458 | description when this job is run. It's not parsed. |
| 459 | .TP |
| 460 | .BI loops \fR=\fPint |
| 461 | Run the specified number of iterations of this job. Used to repeat the same |
| 462 | workload a given number of times. Defaults to 1. |
| 463 | .TP |
| 464 | .BI numjobs \fR=\fPint |
| 465 | Create the specified number of clones of this job. Each clone of job |
| 466 | is spawned as an independent thread or process. May be used to setup a |
| 467 | larger number of threads/processes doing the same thing. Each thread is |
| 468 | reported separately; to see statistics for all clones as a whole, use |
| 469 | \fBgroup_reporting\fR in conjunction with \fBnew_group\fR. |
| 470 | See \fB\-\-max\-jobs\fR. Default: 1. |
| 471 | .SS "Time related parameters" |
| 472 | .TP |
| 473 | .BI runtime \fR=\fPtime |
| 474 | Tell fio to terminate processing after the specified period of time. It |
| 475 | can be quite hard to determine for how long a specified job will run, so |
| 476 | this parameter is handy to cap the total runtime to a given time. When |
| 477 | the unit is omitted, the value is interpreted in seconds. |
| 478 | .TP |
| 479 | .BI time_based |
| 480 | If set, fio will run for the duration of the \fBruntime\fR specified |
| 481 | even if the file(s) are completely read or written. It will simply loop over |
| 482 | the same workload as many times as the \fBruntime\fR allows. |
| 483 | .TP |
| 484 | .BI startdelay \fR=\fPirange(int) |
| 485 | Delay the start of job for the specified amount of time. Can be a single |
| 486 | value or a range. When given as a range, each thread will choose a value |
| 487 | randomly from within the range. Value is in seconds if a unit is omitted. |
| 488 | .TP |
| 489 | .BI ramp_time \fR=\fPtime |
| 490 | If set, fio will run the specified workload for this amount of time before |
| 491 | logging any performance numbers. Useful for letting performance settle |
| 492 | before logging results, thus minimizing the runtime required for stable |
| 493 | results. Note that the \fBramp_time\fR is considered lead in time for a job, |
| 494 | thus it will increase the total runtime if a special timeout or |
| 495 | \fBruntime\fR is specified. When the unit is omitted, the value is |
| 496 | given in seconds. |
| 497 | .TP |
| 498 | .BI clocksource \fR=\fPstr |
| 499 | Use the given clocksource as the base of timing. The supported options are: |
| 500 | .RS |
| 501 | .RS |
| 502 | .TP |
| 503 | .B gettimeofday |
| 504 | \fBgettimeofday\fR\|(2) |
| 505 | .TP |
| 506 | .B clock_gettime |
| 507 | \fBclock_gettime\fR\|(2) |
| 508 | .TP |
| 509 | .B cpu |
| 510 | Internal CPU clock source |
| 511 | .RE |
| 512 | .P |
| 513 | \fBcpu\fR is the preferred clocksource if it is reliable, as it is very fast (and |
| 514 | fio is heavy on time calls). Fio will automatically use this clocksource if |
| 515 | it's supported and considered reliable on the system it is running on, |
| 516 | unless another clocksource is specifically set. For x86/x86\-64 CPUs, this |
| 517 | means supporting TSC Invariant. |
| 518 | .RE |
| 519 | .TP |
| 520 | .BI gtod_reduce \fR=\fPbool |
| 521 | Enable all of the \fBgettimeofday\fR\|(2) reducing options |
| 522 | (\fBdisable_clat\fR, \fBdisable_slat\fR, \fBdisable_bw_measurement\fR) plus |
| 523 | reduce precision of the timeout somewhat to really shrink the |
| 524 | \fBgettimeofday\fR\|(2) call count. With this option enabled, we only do |
| 525 | about 0.4% of the \fBgettimeofday\fR\|(2) calls we would have done if all |
| 526 | time keeping was enabled. |
| 527 | .TP |
| 528 | .BI gtod_cpu \fR=\fPint |
| 529 | Sometimes it's cheaper to dedicate a single thread of execution to just |
| 530 | getting the current time. Fio (and databases, for instance) are very |
| 531 | intensive on \fBgettimeofday\fR\|(2) calls. With this option, you can set |
| 532 | one CPU aside for doing nothing but logging current time to a shared memory |
| 533 | location. Then the other threads/processes that run I/O workloads need only |
| 534 | copy that segment, instead of entering the kernel with a |
| 535 | \fBgettimeofday\fR\|(2) call. The CPU set aside for doing these time |
| 536 | calls will be excluded from other uses. Fio will manually clear it from the |
| 537 | CPU mask of other jobs. |
| 538 | .SS "Target file/device" |
| 539 | .TP |
| 540 | .BI directory \fR=\fPstr |
| 541 | Prefix \fBfilename\fRs with this directory. Used to place files in a different |
| 542 | location than `./'. You can specify a number of directories by |
| 543 | separating the names with a ':' character. These directories will be |
| 544 | assigned equally distributed to job clones created by \fBnumjobs\fR as |
| 545 | long as they are using generated filenames. If specific \fBfilename\fR(s) are |
| 546 | set fio will use the first listed directory, and thereby matching the |
| 547 | \fBfilename\fR semantic (which generates a file for each clone if not |
| 548 | specified, but lets all clones use the same file if set). |
| 549 | .RS |
| 550 | .P |
| 551 | See the \fBfilename\fR option for information on how to escape ':' |
| 552 | characters within the directory path itself. |
| 553 | .P |
| 554 | Note: To control the directory fio will use for internal state files |
| 555 | use \fB\-\-aux\-path\fR. |
| 556 | .RE |
| 557 | .TP |
| 558 | .BI filename \fR=\fPstr |
| 559 | Fio normally makes up a \fBfilename\fR based on the job name, thread number, and |
| 560 | file number (see \fBfilename_format\fR). If you want to share files |
| 561 | between threads in a job or several |
| 562 | jobs with fixed file paths, specify a \fBfilename\fR for each of them to override |
| 563 | the default. If the ioengine is file based, you can specify a number of files |
| 564 | by separating the names with a ':' colon. So if you wanted a job to open |
| 565 | `/dev/sda' and `/dev/sdb' as the two working files, you would use |
| 566 | `filename=/dev/sda:/dev/sdb'. This also means that whenever this option is |
| 567 | specified, \fBnrfiles\fR is ignored. The size of regular files specified |
| 568 | by this option will be \fBsize\fR divided by number of files unless an |
| 569 | explicit size is specified by \fBfilesize\fR. |
| 570 | .RS |
| 571 | .P |
| 572 | Each colon in the wanted path must be escaped with a '\e' |
| 573 | character. For instance, if the path is `/dev/dsk/foo@3,0:c' then you |
| 574 | would use `filename=/dev/dsk/foo@3,0\\:c' and if the path is |
| 575 | `F:\\filename' then you would use `filename=F\\:\\filename'. |
| 576 | .P |
| 577 | On Windows, disk devices are accessed as `\\\\.\\PhysicalDrive0' for |
| 578 | the first device, `\\\\.\\PhysicalDrive1' for the second etc. |
| 579 | Note: Windows and FreeBSD prevent write access to areas |
| 580 | of the disk containing in-use data (e.g. filesystems). |
| 581 | .P |
| 582 | The filename `\-' is a reserved name, meaning *stdin* or *stdout*. Which |
| 583 | of the two depends on the read/write direction set. |
| 584 | .RE |
| 585 | .TP |
| 586 | .BI filename_format \fR=\fPstr |
| 587 | If sharing multiple files between jobs, it is usually necessary to have fio |
| 588 | generate the exact names that you want. By default, fio will name a file |
| 589 | based on the default file format specification of |
| 590 | `jobname.jobnumber.filenumber'. With this option, that can be |
| 591 | customized. Fio will recognize and replace the following keywords in this |
| 592 | string: |
| 593 | .RS |
| 594 | .RS |
| 595 | .TP |
| 596 | .B $jobname |
| 597 | The name of the worker thread or process. |
| 598 | .TP |
| 599 | .B $clientuid |
| 600 | IP of the fio process when using client/server mode. |
| 601 | .TP |
| 602 | .B $jobnum |
| 603 | The incremental number of the worker thread or process. |
| 604 | .TP |
| 605 | .B $filenum |
| 606 | The incremental number of the file for that worker thread or process. |
| 607 | .RE |
| 608 | .P |
| 609 | To have dependent jobs share a set of files, this option can be set to have |
| 610 | fio generate filenames that are shared between the two. For instance, if |
| 611 | `testfiles.$filenum' is specified, file number 4 for any job will be |
| 612 | named `testfiles.4'. The default of `$jobname.$jobnum.$filenum' |
| 613 | will be used if no other format specifier is given. |
| 614 | .P |
| 615 | If you specify a path then the directories will be created up to the main |
| 616 | directory for the file. So for example if you specify `a/b/c/$jobnum` then the |
| 617 | directories a/b/c will be created before the file setup part of the job. If you |
| 618 | specify \fBdirectory\fR then the path will be relative that directory, otherwise |
| 619 | it is treated as the absolute path. |
| 620 | .RE |
| 621 | .TP |
| 622 | .BI unique_filename \fR=\fPbool |
| 623 | To avoid collisions between networked clients, fio defaults to prefixing any |
| 624 | generated filenames (with a directory specified) with the source of the |
| 625 | client connecting. To disable this behavior, set this option to 0. |
| 626 | .TP |
| 627 | .BI opendir \fR=\fPstr |
| 628 | Recursively open any files below directory \fIstr\fR. |
| 629 | .TP |
| 630 | .BI lockfile \fR=\fPstr |
| 631 | Fio defaults to not locking any files before it does I/O to them. If a file |
| 632 | or file descriptor is shared, fio can serialize I/O to that file to make the |
| 633 | end result consistent. This is usual for emulating real workloads that share |
| 634 | files. The lock modes are: |
| 635 | .RS |
| 636 | .RS |
| 637 | .TP |
| 638 | .B none |
| 639 | No locking. The default. |
| 640 | .TP |
| 641 | .B exclusive |
| 642 | Only one thread or process may do I/O at a time, excluding all others. |
| 643 | .TP |
| 644 | .B readwrite |
| 645 | Read\-write locking on the file. Many readers may |
| 646 | access the file at the same time, but writes get exclusive access. |
| 647 | .RE |
| 648 | .RE |
| 649 | .TP |
| 650 | .BI nrfiles \fR=\fPint |
| 651 | Number of files to use for this job. Defaults to 1. The size of files |
| 652 | will be \fBsize\fR divided by this unless explicit size is specified by |
| 653 | \fBfilesize\fR. Files are created for each thread separately, and each |
| 654 | file will have a file number within its name by default, as explained in |
| 655 | \fBfilename\fR section. |
| 656 | .TP |
| 657 | .BI openfiles \fR=\fPint |
| 658 | Number of files to keep open at the same time. Defaults to the same as |
| 659 | \fBnrfiles\fR, can be set smaller to limit the number simultaneous |
| 660 | opens. |
| 661 | .TP |
| 662 | .BI file_service_type \fR=\fPstr |
| 663 | Defines how fio decides which file from a job to service next. The following |
| 664 | types are defined: |
| 665 | .RS |
| 666 | .RS |
| 667 | .TP |
| 668 | .B random |
| 669 | Choose a file at random. |
| 670 | .TP |
| 671 | .B roundrobin |
| 672 | Round robin over opened files. This is the default. |
| 673 | .TP |
| 674 | .B sequential |
| 675 | Finish one file before moving on to the next. Multiple files can |
| 676 | still be open depending on \fBopenfiles\fR. |
| 677 | .TP |
| 678 | .B zipf |
| 679 | Use a Zipf distribution to decide what file to access. |
| 680 | .TP |
| 681 | .B pareto |
| 682 | Use a Pareto distribution to decide what file to access. |
| 683 | .TP |
| 684 | .B normal |
| 685 | Use a Gaussian (normal) distribution to decide what file to access. |
| 686 | .TP |
| 687 | .B gauss |
| 688 | Alias for normal. |
| 689 | .RE |
| 690 | .P |
| 691 | For \fBrandom\fR, \fBroundrobin\fR, and \fBsequential\fR, a postfix can be appended to |
| 692 | tell fio how many I/Os to issue before switching to a new file. For example, |
| 693 | specifying `file_service_type=random:8' would cause fio to issue |
| 694 | 8 I/Os before selecting a new file at random. For the non-uniform |
| 695 | distributions, a floating point postfix can be given to influence how the |
| 696 | distribution is skewed. See \fBrandom_distribution\fR for a description |
| 697 | of how that would work. |
| 698 | .RE |
| 699 | .TP |
| 700 | .BI ioscheduler \fR=\fPstr |
| 701 | Attempt to switch the device hosting the file to the specified I/O scheduler |
| 702 | before running. If the file is a pipe, a character device file or if device |
| 703 | hosting the file could not be determined, this option is ignored. |
| 704 | .TP |
| 705 | .BI create_serialize \fR=\fPbool |
| 706 | If true, serialize the file creation for the jobs. This may be handy to |
| 707 | avoid interleaving of data files, which may greatly depend on the filesystem |
| 708 | used and even the number of processors in the system. Default: true. |
| 709 | .TP |
| 710 | .BI create_fsync \fR=\fPbool |
| 711 | \fBfsync\fR\|(2) the data file after creation. This is the default. |
| 712 | .TP |
| 713 | .BI create_on_open \fR=\fPbool |
| 714 | If true, don't pre-create files but allow the job's open() to create a file |
| 715 | when it's time to do I/O. Default: false \-\- pre-create all necessary files |
| 716 | when the job starts. |
| 717 | .TP |
| 718 | .BI create_only \fR=\fPbool |
| 719 | If true, fio will only run the setup phase of the job. If files need to be |
| 720 | laid out or updated on disk, only that will be done \-\- the actual job contents |
| 721 | are not executed. Default: false. |
| 722 | .TP |
| 723 | .BI allow_file_create \fR=\fPbool |
| 724 | If true, fio is permitted to create files as part of its workload. If this |
| 725 | option is false, then fio will error out if |
| 726 | the files it needs to use don't already exist. Default: true. |
| 727 | .TP |
| 728 | .BI allow_mounted_write \fR=\fPbool |
| 729 | If this isn't set, fio will abort jobs that are destructive (e.g. that write) |
| 730 | to what appears to be a mounted device or partition. This should help catch |
| 731 | creating inadvertently destructive tests, not realizing that the test will |
| 732 | destroy data on the mounted file system. Note that some platforms don't allow |
| 733 | writing against a mounted device regardless of this option. Default: false. |
| 734 | .TP |
| 735 | .BI pre_read \fR=\fPbool |
| 736 | If this is given, files will be pre-read into memory before starting the |
| 737 | given I/O operation. This will also clear the \fBinvalidate\fR flag, |
| 738 | since it is pointless to pre-read and then drop the cache. This will only |
| 739 | work for I/O engines that are seek-able, since they allow you to read the |
| 740 | same data multiple times. Thus it will not work on non-seekable I/O engines |
| 741 | (e.g. network, splice). Default: false. |
| 742 | .TP |
| 743 | .BI unlink \fR=\fPbool |
| 744 | Unlink the job files when done. Not the default, as repeated runs of that |
| 745 | job would then waste time recreating the file set again and again. Default: |
| 746 | false. |
| 747 | .TP |
| 748 | .BI unlink_each_loop \fR=\fPbool |
| 749 | Unlink job files after each iteration or loop. Default: false. |
| 750 | .TP |
| 751 | .BI zonemode \fR=\fPstr |
| 752 | Accepted values are: |
| 753 | .RS |
| 754 | .RS |
| 755 | .TP |
| 756 | .B none |
| 757 | The \fBzonerange\fR, \fBzonesize\fR \fBzonecapacity\fR and \fBzoneskip\fR |
| 758 | parameters are ignored. |
| 759 | .TP |
| 760 | .B strided |
| 761 | I/O happens in a single zone until \fBzonesize\fR bytes have been transferred. |
| 762 | After that number of bytes has been transferred processing of the next zone |
| 763 | starts. The \fBzonecapacity\fR parameter is ignored. |
| 764 | .TP |
| 765 | .B zbd |
| 766 | Zoned block device mode. I/O happens sequentially in each zone, even if random |
| 767 | I/O has been selected. Random I/O happens across all zones instead of being |
| 768 | restricted to a single zone. |
| 769 | Trim is handled using a zone reset operation. Trim only considers non-empty |
| 770 | sequential write required and sequential write preferred zones. |
| 771 | .RE |
| 772 | .RE |
| 773 | .TP |
| 774 | .BI zonerange \fR=\fPint |
| 775 | For \fBzonemode\fR=strided, this is the size of a single zone. See also |
| 776 | \fBzonesize\fR and \fBzoneskip\fR. |
| 777 | |
| 778 | For \fBzonemode\fR=zbd, this parameter is ignored. |
| 779 | .TP |
| 780 | .BI zonesize \fR=\fPint |
| 781 | For \fBzonemode\fR=strided, this is the number of bytes to transfer before |
| 782 | skipping \fBzoneskip\fR bytes. If this parameter is smaller than |
| 783 | \fBzonerange\fR then only a fraction of each zone with \fBzonerange\fR bytes |
| 784 | will be accessed. If this parameter is larger than \fBzonerange\fR then each |
| 785 | zone will be accessed multiple times before skipping to the next zone. |
| 786 | |
| 787 | For \fBzonemode\fR=zbd, this is the size of a single zone. The |
| 788 | \fBzonerange\fR parameter is ignored in this mode. For a job accessing a |
| 789 | zoned block device, the specified \fBzonesize\fR must be 0 or equal to the |
| 790 | device zone size. For a regular block device or file, the specified |
| 791 | \fBzonesize\fR must be at least 512B. |
| 792 | .TP |
| 793 | .BI zonecapacity \fR=\fPint |
| 794 | For \fBzonemode\fR=zbd, this defines the capacity of a single zone, which is |
| 795 | the accessible area starting from the zone start address. This parameter only |
| 796 | applies when using \fBzonemode\fR=zbd in combination with regular block devices. |
| 797 | If not specified it defaults to the zone size. If the target device is a zoned |
| 798 | block device, the zone capacity is obtained from the device information and this |
| 799 | option is ignored. |
| 800 | .TP |
| 801 | .BI zoneskip \fR=\fPint[z] |
| 802 | For \fBzonemode\fR=strided, the number of bytes to skip after \fBzonesize\fR |
| 803 | bytes of data have been transferred. |
| 804 | |
| 805 | For \fBzonemode\fR=zbd, the \fBzonesize\fR aligned number of bytes to skip |
| 806 | once a zone is fully written (write workloads) or all written data in the |
| 807 | zone have been read (read workloads). This parameter is valid only for |
| 808 | sequential workloads and ignored for random workloads. For read workloads, |
| 809 | see also \fBread_beyond_wp\fR. |
| 810 | |
| 811 | .TP |
| 812 | .BI read_beyond_wp \fR=\fPbool |
| 813 | This parameter applies to \fBzonemode=zbd\fR only. |
| 814 | |
| 815 | Zoned block devices are block devices that consist of multiple zones. Each |
| 816 | zone has a type, e.g. conventional or sequential. A conventional zone can be |
| 817 | written at any offset that is a multiple of the block size. Sequential zones |
| 818 | must be written sequentially. The position at which a write must occur is |
| 819 | called the write pointer. A zoned block device can be either host managed or |
| 820 | host aware. For host managed devices the host must ensure that writes happen |
| 821 | sequentially. Fio recognizes host managed devices and serializes writes to |
| 822 | sequential zones for these devices. |
| 823 | |
| 824 | If a read occurs in a sequential zone beyond the write pointer then the zoned |
| 825 | block device will complete the read without reading any data from the storage |
| 826 | medium. Since such reads lead to unrealistically high bandwidth and IOPS |
| 827 | numbers fio only reads beyond the write pointer if explicitly told to do |
| 828 | so. Default: false. |
| 829 | .TP |
| 830 | .BI max_open_zones \fR=\fPint |
| 831 | When running a random write test across an entire drive many more zones will be |
| 832 | open than in a typical application workload. Hence this command line option |
| 833 | that allows one to limit the number of open zones. The number of open zones is |
| 834 | defined as the number of zones to which write commands are issued by all |
| 835 | threads/processes. |
| 836 | .TP |
| 837 | .BI job_max_open_zones \fR=\fPint |
| 838 | Limit on the number of simultaneously opened zones per single thread/process. |
| 839 | .TP |
| 840 | .BI ignore_zone_limits \fR=\fPbool |
| 841 | If this option is used, fio will ignore the maximum number of open zones limit |
| 842 | of the zoned block device in use, thus allowing the option \fBmax_open_zones\fR |
| 843 | value to be larger than the device reported limit. Default: false. |
| 844 | .TP |
| 845 | .BI zone_reset_threshold \fR=\fPfloat |
| 846 | A number between zero and one that indicates the ratio of logical blocks with |
| 847 | data to the total number of logical blocks in the test above which zones |
| 848 | should be reset periodically. |
| 849 | .TP |
| 850 | .BI zone_reset_frequency \fR=\fPfloat |
| 851 | A number between zero and one that indicates how often a zone reset should be |
| 852 | issued if the zone reset threshold has been exceeded. A zone reset is |
| 853 | submitted after each (1 / zone_reset_frequency) write requests. This and the |
| 854 | previous parameter can be used to simulate garbage collection activity. |
| 855 | |
| 856 | .SS "I/O type" |
| 857 | .TP |
| 858 | .BI direct \fR=\fPbool |
| 859 | If value is true, use non-buffered I/O. This is usually O_DIRECT. Note that |
| 860 | OpenBSD and ZFS on Solaris don't support direct I/O. On Windows the synchronous |
| 861 | ioengines don't support direct I/O. Default: false. |
| 862 | .TP |
| 863 | .BI atomic \fR=\fPbool |
| 864 | If value is true, attempt to use atomic direct I/O. Atomic writes are |
| 865 | guaranteed to be stable once acknowledged by the operating system. Only |
| 866 | Linux supports O_ATOMIC right now. |
| 867 | .TP |
| 868 | .BI buffered \fR=\fPbool |
| 869 | If value is true, use buffered I/O. This is the opposite of the |
| 870 | \fBdirect\fR option. Defaults to true. |
| 871 | .TP |
| 872 | .BI readwrite \fR=\fPstr "\fR,\fP rw" \fR=\fPstr |
| 873 | Type of I/O pattern. Accepted values are: |
| 874 | .RS |
| 875 | .RS |
| 876 | .TP |
| 877 | .B read |
| 878 | Sequential reads. |
| 879 | .TP |
| 880 | .B write |
| 881 | Sequential writes. |
| 882 | .TP |
| 883 | .B trim |
| 884 | Sequential trims (Linux block devices and SCSI character devices only). |
| 885 | .TP |
| 886 | .B randread |
| 887 | Random reads. |
| 888 | .TP |
| 889 | .B randwrite |
| 890 | Random writes. |
| 891 | .TP |
| 892 | .B randtrim |
| 893 | Random trims (Linux block devices and SCSI character devices only). |
| 894 | .TP |
| 895 | .B rw,readwrite |
| 896 | Sequential mixed reads and writes. |
| 897 | .TP |
| 898 | .B randrw |
| 899 | Random mixed reads and writes. |
| 900 | .TP |
| 901 | .B trimwrite |
| 902 | Sequential trim+write sequences. Blocks will be trimmed first, |
| 903 | then the same blocks will be written to. So if `io_size=64K' is specified, |
| 904 | Fio will trim a total of 64K bytes and also write 64K bytes on the same |
| 905 | trimmed blocks. This behaviour will be consistent with `number_ios' or |
| 906 | other Fio options limiting the total bytes or number of I/O's. |
| 907 | .TP |
| 908 | .B randtrimwrite |
| 909 | Like |
| 910 | .B trimwrite , |
| 911 | but uses random offsets rather than sequential writes. |
| 912 | .RE |
| 913 | .P |
| 914 | Fio defaults to read if the option is not specified. For the mixed I/O |
| 915 | types, the default is to split them 50/50. For certain types of I/O the |
| 916 | result may still be skewed a bit, since the speed may be different. |
| 917 | .P |
| 918 | It is possible to specify the number of I/Os to do before getting a new |
| 919 | offset by appending `:<nr>' to the end of the string given. For a |
| 920 | random read, it would look like `rw=randread:8' for passing in an offset |
| 921 | modifier with a value of 8. If the suffix is used with a sequential I/O |
| 922 | pattern, then the `<nr>' value specified will be added to the generated |
| 923 | offset for each I/O turning sequential I/O into sequential I/O with holes. |
| 924 | For instance, using `rw=write:4k' will skip 4k for every write. Also see |
| 925 | the \fBrw_sequencer\fR option. |
| 926 | .RE |
| 927 | .TP |
| 928 | .BI rw_sequencer \fR=\fPstr |
| 929 | If an offset modifier is given by appending a number to the `rw=\fIstr\fR' |
| 930 | line, then this option controls how that number modifies the I/O offset |
| 931 | being generated. Accepted values are: |
| 932 | .RS |
| 933 | .RS |
| 934 | .TP |
| 935 | .B sequential |
| 936 | Generate sequential offset. |
| 937 | .TP |
| 938 | .B identical |
| 939 | Generate the same offset. |
| 940 | .RE |
| 941 | .P |
| 942 | \fBsequential\fR is only useful for random I/O, where fio would normally |
| 943 | generate a new random offset for every I/O. If you append e.g. 8 to randread, |
| 944 | you would get a new random offset for every 8 I/Os. The result would be a |
| 945 | seek for only every 8 I/Os, instead of for every I/O. Use `rw=randread:8' |
| 946 | to specify that. As sequential I/O is already sequential, setting |
| 947 | \fBsequential\fR for that would not result in any differences. \fBidentical\fR |
| 948 | behaves in a similar fashion, except it sends the same offset 8 number of |
| 949 | times before generating a new offset. |
| 950 | .RE |
| 951 | .TP |
| 952 | .BI unified_rw_reporting \fR=\fPstr |
| 953 | Fio normally reports statistics on a per data direction basis, meaning that |
| 954 | reads, writes, and trims are accounted and reported separately. This option |
| 955 | determines whether fio reports the results normally, summed together, or as |
| 956 | both options. |
| 957 | Accepted values are: |
| 958 | .RS |
| 959 | .TP |
| 960 | .B none |
| 961 | Normal statistics reporting. |
| 962 | .TP |
| 963 | .B mixed |
| 964 | Statistics are summed per data direction and reported together. |
| 965 | .TP |
| 966 | .B both |
| 967 | Statistics are reported normally, followed by the mixed statistics. |
| 968 | .TP |
| 969 | .B 0 |
| 970 | Backward-compatible alias for \fBnone\fR. |
| 971 | .TP |
| 972 | .B 1 |
| 973 | Backward-compatible alias for \fBmixed\fR. |
| 974 | .TP |
| 975 | .B 2 |
| 976 | Alias for \fBboth\fR. |
| 977 | .RE |
| 978 | .TP |
| 979 | .BI randrepeat \fR=\fPbool |
| 980 | Seed the random number generator used for random I/O patterns in a |
| 981 | predictable way so the pattern is repeatable across runs. Default: true. |
| 982 | .TP |
| 983 | .BI allrandrepeat \fR=\fPbool |
| 984 | Seed all random number generators in a predictable way so results are |
| 985 | repeatable across runs. Default: false. |
| 986 | .TP |
| 987 | .BI randseed \fR=\fPint |
| 988 | Seed the random number generators based on this seed value, to be able to |
| 989 | control what sequence of output is being generated. If not set, the random |
| 990 | sequence depends on the \fBrandrepeat\fR setting. |
| 991 | .TP |
| 992 | .BI fallocate \fR=\fPstr |
| 993 | Whether pre-allocation is performed when laying down files. |
| 994 | Accepted values are: |
| 995 | .RS |
| 996 | .RS |
| 997 | .TP |
| 998 | .B none |
| 999 | Do not pre-allocate space. |
| 1000 | .TP |
| 1001 | .B native |
| 1002 | Use a platform's native pre-allocation call but fall back to |
| 1003 | \fBnone\fR behavior if it fails/is not implemented. |
| 1004 | .TP |
| 1005 | .B posix |
| 1006 | Pre-allocate via \fBposix_fallocate\fR\|(3). |
| 1007 | .TP |
| 1008 | .B keep |
| 1009 | Pre-allocate via \fBfallocate\fR\|(2) with |
| 1010 | FALLOC_FL_KEEP_SIZE set. |
| 1011 | .TP |
| 1012 | .B truncate |
| 1013 | Extend file to final size using \fBftruncate\fR|(2) |
| 1014 | instead of allocating. |
| 1015 | .TP |
| 1016 | .B 0 |
| 1017 | Backward-compatible alias for \fBnone\fR. |
| 1018 | .TP |
| 1019 | .B 1 |
| 1020 | Backward-compatible alias for \fBposix\fR. |
| 1021 | .RE |
| 1022 | .P |
| 1023 | May not be available on all supported platforms. \fBkeep\fR is only available |
| 1024 | on Linux. If using ZFS on Solaris this cannot be set to \fBposix\fR |
| 1025 | because ZFS doesn't support pre-allocation. Default: \fBnative\fR if any |
| 1026 | pre-allocation methods except \fBtruncate\fR are available, \fBnone\fR if not. |
| 1027 | .P |
| 1028 | Note that using \fBtruncate\fR on Windows will interact surprisingly |
| 1029 | with non-sequential write patterns. When writing to a file that has |
| 1030 | been extended by setting the end-of-file information, Windows will |
| 1031 | backfill the unwritten portion of the file up to that offset with |
| 1032 | zeroes before issuing the new write. This means that a single small |
| 1033 | write to the end of an extended file will stall until the entire |
| 1034 | file has been filled with zeroes. |
| 1035 | .RE |
| 1036 | .TP |
| 1037 | .BI fadvise_hint \fR=\fPstr |
| 1038 | Use \fBposix_fadvise\fR\|(2) or \fBposix_madvise\fR\|(2) to advise the kernel |
| 1039 | what I/O patterns are likely to be issued. Accepted values are: |
| 1040 | .RS |
| 1041 | .RS |
| 1042 | .TP |
| 1043 | .B 0 |
| 1044 | Backwards compatible hint for "no hint". |
| 1045 | .TP |
| 1046 | .B 1 |
| 1047 | Backwards compatible hint for "advise with fio workload type". This |
| 1048 | uses FADV_RANDOM for a random workload, and FADV_SEQUENTIAL |
| 1049 | for a sequential workload. |
| 1050 | .TP |
| 1051 | .B sequential |
| 1052 | Advise using FADV_SEQUENTIAL. |
| 1053 | .TP |
| 1054 | .B random |
| 1055 | Advise using FADV_RANDOM. |
| 1056 | .RE |
| 1057 | .RE |
| 1058 | .TP |
| 1059 | .BI write_hint \fR=\fPstr |
| 1060 | Use \fBfcntl\fR\|(2) to advise the kernel what life time to expect |
| 1061 | from a write. Only supported on Linux, as of version 4.13. Accepted |
| 1062 | values are: |
| 1063 | .RS |
| 1064 | .RS |
| 1065 | .TP |
| 1066 | .B none |
| 1067 | No particular life time associated with this file. |
| 1068 | .TP |
| 1069 | .B short |
| 1070 | Data written to this file has a short life time. |
| 1071 | .TP |
| 1072 | .B medium |
| 1073 | Data written to this file has a medium life time. |
| 1074 | .TP |
| 1075 | .B long |
| 1076 | Data written to this file has a long life time. |
| 1077 | .TP |
| 1078 | .B extreme |
| 1079 | Data written to this file has a very long life time. |
| 1080 | .RE |
| 1081 | .P |
| 1082 | The values are all relative to each other, and no absolute meaning |
| 1083 | should be associated with them. |
| 1084 | .RE |
| 1085 | .TP |
| 1086 | .BI offset \fR=\fPint[%|z] |
| 1087 | Start I/O at the provided offset in the file, given as either a fixed size in |
| 1088 | bytes, zones or a percentage. If a percentage is given, the generated offset will be |
| 1089 | aligned to the minimum \fBblocksize\fR or to the value of \fBoffset_align\fR if |
| 1090 | provided. Data before the given offset will not be touched. This |
| 1091 | effectively caps the file size at `real_size \- offset'. Can be combined with |
| 1092 | \fBsize\fR to constrain the start and end range of the I/O workload. |
| 1093 | A percentage can be specified by a number between 1 and 100 followed by '%', |
| 1094 | for example, `offset=20%' to specify 20%. In ZBD mode, value can be set as |
| 1095 | number of zones using 'z'. |
| 1096 | .TP |
| 1097 | .BI offset_align \fR=\fPint |
| 1098 | If set to non-zero value, the byte offset generated by a percentage \fBoffset\fR |
| 1099 | is aligned upwards to this value. Defaults to 0 meaning that a percentage |
| 1100 | offset is aligned to the minimum block size. |
| 1101 | .TP |
| 1102 | .BI offset_increment \fR=\fPint[%|z] |
| 1103 | If this is provided, then the real offset becomes `\fBoffset\fR + \fBoffset_increment\fR |
| 1104 | * thread_number', where the thread number is a counter that starts at 0 and |
| 1105 | is incremented for each sub-job (i.e. when \fBnumjobs\fR option is |
| 1106 | specified). This option is useful if there are several jobs which are |
| 1107 | intended to operate on a file in parallel disjoint segments, with even |
| 1108 | spacing between the starting points. Percentages can be used for this option. |
| 1109 | If a percentage is given, the generated offset will be aligned to the minimum |
| 1110 | \fBblocksize\fR or to the value of \fBoffset_align\fR if provided.In ZBD mode, value |
| 1111 | can be set as number of zones using 'z'. |
| 1112 | .TP |
| 1113 | .BI number_ios \fR=\fPint |
| 1114 | Fio will normally perform I/Os until it has exhausted the size of the region |
| 1115 | set by \fBsize\fR, or if it exhaust the allocated time (or hits an error |
| 1116 | condition). With this setting, the range/size can be set independently of |
| 1117 | the number of I/Os to perform. When fio reaches this number, it will exit |
| 1118 | normally and report status. Note that this does not extend the amount of I/O |
| 1119 | that will be done, it will only stop fio if this condition is met before |
| 1120 | other end-of-job criteria. |
| 1121 | .TP |
| 1122 | .BI fsync \fR=\fPint |
| 1123 | If writing to a file, issue an \fBfsync\fR\|(2) (or its equivalent) of |
| 1124 | the dirty data for every number of blocks given. For example, if you give 32 |
| 1125 | as a parameter, fio will sync the file after every 32 writes issued. If fio is |
| 1126 | using non-buffered I/O, we may not sync the file. The exception is the sg |
| 1127 | I/O engine, which synchronizes the disk cache anyway. Defaults to 0, which |
| 1128 | means fio does not periodically issue and wait for a sync to complete. Also |
| 1129 | see \fBend_fsync\fR and \fBfsync_on_close\fR. |
| 1130 | .TP |
| 1131 | .BI fdatasync \fR=\fPint |
| 1132 | Like \fBfsync\fR but uses \fBfdatasync\fR\|(2) to only sync data and |
| 1133 | not metadata blocks. In Windows, DragonFlyBSD or OSX there is no |
| 1134 | \fBfdatasync\fR\|(2) so this falls back to using \fBfsync\fR\|(2). |
| 1135 | Defaults to 0, which means fio does not periodically issue and wait for a |
| 1136 | data-only sync to complete. |
| 1137 | .TP |
| 1138 | .BI write_barrier \fR=\fPint |
| 1139 | Make every N\-th write a barrier write. |
| 1140 | .TP |
| 1141 | .BI sync_file_range \fR=\fPstr:int |
| 1142 | Use \fBsync_file_range\fR\|(2) for every \fIint\fR number of write |
| 1143 | operations. Fio will track range of writes that have happened since the last |
| 1144 | \fBsync_file_range\fR\|(2) call. \fIstr\fR can currently be one or more of: |
| 1145 | .RS |
| 1146 | .RS |
| 1147 | .TP |
| 1148 | .B wait_before |
| 1149 | SYNC_FILE_RANGE_WAIT_BEFORE |
| 1150 | .TP |
| 1151 | .B write |
| 1152 | SYNC_FILE_RANGE_WRITE |
| 1153 | .TP |
| 1154 | .B wait_after |
| 1155 | SYNC_FILE_RANGE_WRITE_AFTER |
| 1156 | .RE |
| 1157 | .P |
| 1158 | So if you do `sync_file_range=wait_before,write:8', fio would use |
| 1159 | `SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE' for every 8 |
| 1160 | writes. Also see the \fBsync_file_range\fR\|(2) man page. This option is |
| 1161 | Linux specific. |
| 1162 | .RE |
| 1163 | .TP |
| 1164 | .BI overwrite \fR=\fPbool |
| 1165 | If true, writes to a file will always overwrite existing data. If the file |
| 1166 | doesn't already exist, it will be created before the write phase begins. If |
| 1167 | the file exists and is large enough for the specified write phase, nothing |
| 1168 | will be done. Default: false. |
| 1169 | .TP |
| 1170 | .BI end_fsync \fR=\fPbool |
| 1171 | If true, \fBfsync\fR\|(2) file contents when a write stage has completed. |
| 1172 | Default: false. |
| 1173 | .TP |
| 1174 | .BI fsync_on_close \fR=\fPbool |
| 1175 | If true, fio will \fBfsync\fR\|(2) a dirty file on close. This differs |
| 1176 | from \fBend_fsync\fR in that it will happen on every file close, not |
| 1177 | just at the end of the job. Default: false. |
| 1178 | .TP |
| 1179 | .BI rwmixread \fR=\fPint |
| 1180 | Percentage of a mixed workload that should be reads. Default: 50. |
| 1181 | .TP |
| 1182 | .BI rwmixwrite \fR=\fPint |
| 1183 | Percentage of a mixed workload that should be writes. If both |
| 1184 | \fBrwmixread\fR and \fBrwmixwrite\fR is given and the values do not |
| 1185 | add up to 100%, the latter of the two will be used to override the |
| 1186 | first. This may interfere with a given rate setting, if fio is asked to |
| 1187 | limit reads or writes to a certain rate. If that is the case, then the |
| 1188 | distribution may be skewed. Default: 50. |
| 1189 | .TP |
| 1190 | .BI random_distribution \fR=\fPstr:float[:float][,str:float][,str:float] |
| 1191 | By default, fio will use a completely uniform random distribution when asked |
| 1192 | to perform random I/O. Sometimes it is useful to skew the distribution in |
| 1193 | specific ways, ensuring that some parts of the data is more hot than others. |
| 1194 | fio includes the following distribution models: |
| 1195 | .RS |
| 1196 | .RS |
| 1197 | .TP |
| 1198 | .B random |
| 1199 | Uniform random distribution |
| 1200 | .TP |
| 1201 | .B zipf |
| 1202 | Zipf distribution |
| 1203 | .TP |
| 1204 | .B pareto |
| 1205 | Pareto distribution |
| 1206 | .TP |
| 1207 | .B normal |
| 1208 | Normal (Gaussian) distribution |
| 1209 | .TP |
| 1210 | .B zoned |
| 1211 | Zoned random distribution |
| 1212 | .B zoned_abs |
| 1213 | Zoned absolute random distribution |
| 1214 | .RE |
| 1215 | .P |
| 1216 | When using a \fBzipf\fR or \fBpareto\fR distribution, an input value is also |
| 1217 | needed to define the access pattern. For \fBzipf\fR, this is the `Zipf theta'. |
| 1218 | For \fBpareto\fR, it's the `Pareto power'. Fio includes a test |
| 1219 | program, \fBfio\-genzipf\fR, that can be used visualize what the given input |
| 1220 | values will yield in terms of hit rates. If you wanted to use \fBzipf\fR with |
| 1221 | a `theta' of 1.2, you would use `random_distribution=zipf:1.2' as the |
| 1222 | option. If a non\-uniform model is used, fio will disable use of the random |
| 1223 | map. For the \fBnormal\fR distribution, a normal (Gaussian) deviation is |
| 1224 | supplied as a value between 0 and 100. |
| 1225 | .P |
| 1226 | The second, optional float is allowed for \fBpareto\fR, \fBzipf\fR and \fBnormal\fR |
| 1227 | distributions. It allows one to set base of distribution in non-default place, giving |
| 1228 | more control over most probable outcome. This value is in range [0-1] which maps linearly to |
| 1229 | range of possible random values. |
| 1230 | Defaults are: random for \fBpareto\fR and \fBzipf\fR, and 0.5 for \fBnormal\fR. |
| 1231 | If you wanted to use \fBzipf\fR with a `theta` of 1.2 centered on 1/4 of allowed value range, |
| 1232 | you would use `random_distribution=zipf:1.2:0.25`. |
| 1233 | .P |
| 1234 | For a \fBzoned\fR distribution, fio supports specifying percentages of I/O |
| 1235 | access that should fall within what range of the file or device. For |
| 1236 | example, given a criteria of: |
| 1237 | .RS |
| 1238 | .P |
| 1239 | .PD 0 |
| 1240 | 60% of accesses should be to the first 10% |
| 1241 | .P |
| 1242 | 30% of accesses should be to the next 20% |
| 1243 | .P |
| 1244 | 8% of accesses should be to the next 30% |
| 1245 | .P |
| 1246 | 2% of accesses should be to the next 40% |
| 1247 | .PD |
| 1248 | .RE |
| 1249 | .P |
| 1250 | we can define that through zoning of the random accesses. For the above |
| 1251 | example, the user would do: |
| 1252 | .RS |
| 1253 | .P |
| 1254 | random_distribution=zoned:60/10:30/20:8/30:2/40 |
| 1255 | .RE |
| 1256 | .P |
| 1257 | A \fBzoned_abs\fR distribution works exactly like the\fBzoned\fR, except that |
| 1258 | it takes absolute sizes. For example, let's say you wanted to define access |
| 1259 | according to the following criteria: |
| 1260 | .RS |
| 1261 | .P |
| 1262 | .PD 0 |
| 1263 | 60% of accesses should be to the first 20G |
| 1264 | .P |
| 1265 | 30% of accesses should be to the next 100G |
| 1266 | .P |
| 1267 | 10% of accesses should be to the next 500G |
| 1268 | .PD |
| 1269 | .RE |
| 1270 | .P |
| 1271 | we can define an absolute zoning distribution with: |
| 1272 | .RS |
| 1273 | .P |
| 1274 | random_distribution=zoned:60/10:30/20:8/30:2/40 |
| 1275 | .RE |
| 1276 | .P |
| 1277 | For both \fBzoned\fR and \fBzoned_abs\fR, fio supports defining up to 256 |
| 1278 | separate zones. |
| 1279 | .P |
| 1280 | Similarly to how \fBbssplit\fR works for setting ranges and percentages |
| 1281 | of block sizes. Like \fBbssplit\fR, it's possible to specify separate |
| 1282 | zones for reads, writes, and trims. If just one set is given, it'll apply to |
| 1283 | all of them. |
| 1284 | .RE |
| 1285 | .TP |
| 1286 | .BI percentage_random \fR=\fPint[,int][,int] |
| 1287 | For a random workload, set how big a percentage should be random. This |
| 1288 | defaults to 100%, in which case the workload is fully random. It can be set |
| 1289 | from anywhere from 0 to 100. Setting it to 0 would make the workload fully |
| 1290 | sequential. Any setting in between will result in a random mix of sequential |
| 1291 | and random I/O, at the given percentages. Comma-separated values may be |
| 1292 | specified for reads, writes, and trims as described in \fBblocksize\fR. |
| 1293 | .TP |
| 1294 | .BI norandommap |
| 1295 | Normally fio will cover every block of the file when doing random I/O. If |
| 1296 | this option is given, fio will just get a new random offset without looking |
| 1297 | at past I/O history. This means that some blocks may not be read or written, |
| 1298 | and that some blocks may be read/written more than once. If this option is |
| 1299 | used with \fBverify\fR and multiple blocksizes (via \fBbsrange\fR), |
| 1300 | only intact blocks are verified, i.e., partially-overwritten blocks are |
| 1301 | ignored. With an async I/O engine and an I/O depth > 1, it is possible for |
| 1302 | the same block to be overwritten, which can cause verification errors. Either |
| 1303 | do not use norandommap in this case, or also use the lfsr random generator. |
| 1304 | .TP |
| 1305 | .BI softrandommap \fR=\fPbool |
| 1306 | See \fBnorandommap\fR. If fio runs with the random block map enabled and |
| 1307 | it fails to allocate the map, if this option is set it will continue without |
| 1308 | a random block map. As coverage will not be as complete as with random maps, |
| 1309 | this option is disabled by default. |
| 1310 | .TP |
| 1311 | .BI random_generator \fR=\fPstr |
| 1312 | Fio supports the following engines for generating I/O offsets for random I/O: |
| 1313 | .RS |
| 1314 | .RS |
| 1315 | .TP |
| 1316 | .B tausworthe |
| 1317 | Strong 2^88 cycle random number generator. |
| 1318 | .TP |
| 1319 | .B lfsr |
| 1320 | Linear feedback shift register generator. |
| 1321 | .TP |
| 1322 | .B tausworthe64 |
| 1323 | Strong 64\-bit 2^258 cycle random number generator. |
| 1324 | .RE |
| 1325 | .P |
| 1326 | \fBtausworthe\fR is a strong random number generator, but it requires tracking |
| 1327 | on the side if we want to ensure that blocks are only read or written |
| 1328 | once. \fBlfsr\fR guarantees that we never generate the same offset twice, and |
| 1329 | it's also less computationally expensive. It's not a true random generator, |
| 1330 | however, though for I/O purposes it's typically good enough. \fBlfsr\fR only |
| 1331 | works with single block sizes, not with workloads that use multiple block |
| 1332 | sizes. If used with such a workload, fio may read or write some blocks |
| 1333 | multiple times. The default value is \fBtausworthe\fR, unless the required |
| 1334 | space exceeds 2^32 blocks. If it does, then \fBtausworthe64\fR is |
| 1335 | selected automatically. |
| 1336 | .RE |
| 1337 | .SS "Block size" |
| 1338 | .TP |
| 1339 | .BI blocksize \fR=\fPint[,int][,int] "\fR,\fB bs" \fR=\fPint[,int][,int] |
| 1340 | The block size in bytes used for I/O units. Default: 4096. A single value |
| 1341 | applies to reads, writes, and trims. Comma-separated values may be |
| 1342 | specified for reads, writes, and trims. A value not terminated in a comma |
| 1343 | applies to subsequent types. Examples: |
| 1344 | .RS |
| 1345 | .RS |
| 1346 | .P |
| 1347 | .PD 0 |
| 1348 | bs=256k means 256k for reads, writes and trims. |
| 1349 | .P |
| 1350 | bs=8k,32k means 8k for reads, 32k for writes and trims. |
| 1351 | .P |
| 1352 | bs=8k,32k, means 8k for reads, 32k for writes, and default for trims. |
| 1353 | .P |
| 1354 | bs=,8k means default for reads, 8k for writes and trims. |
| 1355 | .P |
| 1356 | bs=,8k, means default for reads, 8k for writes, and default for trims. |
| 1357 | .PD |
| 1358 | .RE |
| 1359 | .RE |
| 1360 | .TP |
| 1361 | .BI blocksize_range \fR=\fPirange[,irange][,irange] "\fR,\fB bsrange" \fR=\fPirange[,irange][,irange] |
| 1362 | A range of block sizes in bytes for I/O units. The issued I/O unit will |
| 1363 | always be a multiple of the minimum size, unless |
| 1364 | \fBblocksize_unaligned\fR is set. |
| 1365 | Comma-separated ranges may be specified for reads, writes, and trims as |
| 1366 | described in \fBblocksize\fR. Example: |
| 1367 | .RS |
| 1368 | .RS |
| 1369 | .P |
| 1370 | bsrange=1k\-4k,2k\-8k |
| 1371 | .RE |
| 1372 | .RE |
| 1373 | .TP |
| 1374 | .BI bssplit \fR=\fPstr[,str][,str] |
| 1375 | Sometimes you want even finer grained control of the block sizes issued, not |
| 1376 | just an even split between them. This option allows you to weight various |
| 1377 | block sizes, so that you are able to define a specific amount of block sizes |
| 1378 | issued. The format for this option is: |
| 1379 | .RS |
| 1380 | .RS |
| 1381 | .P |
| 1382 | bssplit=blocksize/percentage:blocksize/percentage |
| 1383 | .RE |
| 1384 | .P |
| 1385 | for as many block sizes as needed. So if you want to define a workload that |
| 1386 | has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would write: |
| 1387 | .RS |
| 1388 | .P |
| 1389 | bssplit=4k/10:64k/50:32k/40 |
| 1390 | .RE |
| 1391 | .P |
| 1392 | Ordering does not matter. If the percentage is left blank, fio will fill in |
| 1393 | the remaining values evenly. So a bssplit option like this one: |
| 1394 | .RS |
| 1395 | .P |
| 1396 | bssplit=4k/50:1k/:32k/ |
| 1397 | .RE |
| 1398 | .P |
| 1399 | would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always add up |
| 1400 | to 100, if bssplit is given a range that adds up to more, it will error out. |
| 1401 | .P |
| 1402 | Comma-separated values may be specified for reads, writes, and trims as |
| 1403 | described in \fBblocksize\fR. |
| 1404 | .P |
| 1405 | If you want a workload that has 50% 2k reads and 50% 4k reads, while having |
| 1406 | 90% 4k writes and 10% 8k writes, you would specify: |
| 1407 | .RS |
| 1408 | .P |
| 1409 | bssplit=2k/50:4k/50,4k/90:8k/10 |
| 1410 | .RE |
| 1411 | .P |
| 1412 | Fio supports defining up to 64 different weights for each data direction. |
| 1413 | .RE |
| 1414 | .TP |
| 1415 | .BI blocksize_unaligned "\fR,\fB bs_unaligned" |
| 1416 | If set, fio will issue I/O units with any size within |
| 1417 | \fBblocksize_range\fR, not just multiples of the minimum size. This |
| 1418 | typically won't work with direct I/O, as that normally requires sector |
| 1419 | alignment. |
| 1420 | .TP |
| 1421 | .BI bs_is_seq_rand \fR=\fPbool |
| 1422 | If this option is set, fio will use the normal read,write blocksize settings |
| 1423 | as sequential,random blocksize settings instead. Any random read or write |
| 1424 | will use the WRITE blocksize settings, and any sequential read or write will |
| 1425 | use the READ blocksize settings. |
| 1426 | .TP |
| 1427 | .BI blockalign \fR=\fPint[,int][,int] "\fR,\fB ba" \fR=\fPint[,int][,int] |
| 1428 | Boundary to which fio will align random I/O units. Default: |
| 1429 | \fBblocksize\fR. Minimum alignment is typically 512b for using direct |
| 1430 | I/O, though it usually depends on the hardware block size. This option is |
| 1431 | mutually exclusive with using a random map for files, so it will turn off |
| 1432 | that option. Comma-separated values may be specified for reads, writes, and |
| 1433 | trims as described in \fBblocksize\fR. |
| 1434 | .SS "Buffers and memory" |
| 1435 | .TP |
| 1436 | .BI zero_buffers |
| 1437 | Initialize buffers with all zeros. Default: fill buffers with random data. |
| 1438 | .TP |
| 1439 | .BI refill_buffers |
| 1440 | If this option is given, fio will refill the I/O buffers on every |
| 1441 | submit. The default is to only fill it at init time and reuse that |
| 1442 | data. Only makes sense if zero_buffers isn't specified, naturally. If data |
| 1443 | verification is enabled, \fBrefill_buffers\fR is also automatically enabled. |
| 1444 | .TP |
| 1445 | .BI scramble_buffers \fR=\fPbool |
| 1446 | If \fBrefill_buffers\fR is too costly and the target is using data |
| 1447 | deduplication, then setting this option will slightly modify the I/O buffer |
| 1448 | contents to defeat normal de-dupe attempts. This is not enough to defeat |
| 1449 | more clever block compression attempts, but it will stop naive dedupe of |
| 1450 | blocks. Default: true. |
| 1451 | .TP |
| 1452 | .BI buffer_compress_percentage \fR=\fPint |
| 1453 | If this is set, then fio will attempt to provide I/O buffer content |
| 1454 | (on WRITEs) that compresses to the specified level. Fio does this by |
| 1455 | providing a mix of random data followed by fixed pattern data. The |
| 1456 | fixed pattern is either zeros, or the pattern specified by |
| 1457 | \fBbuffer_pattern\fR. If the \fBbuffer_pattern\fR option is used, it |
| 1458 | might skew the compression ratio slightly. Setting |
| 1459 | \fBbuffer_compress_percentage\fR to a value other than 100 will also |
| 1460 | enable \fBrefill_buffers\fR in order to reduce the likelihood that |
| 1461 | adjacent blocks are so similar that they over compress when seen |
| 1462 | together. See \fBbuffer_compress_chunk\fR for how to set a finer or |
| 1463 | coarser granularity of the random/fixed data regions. Defaults to unset |
| 1464 | i.e., buffer data will not adhere to any compression level. |
| 1465 | .TP |
| 1466 | .BI buffer_compress_chunk \fR=\fPint |
| 1467 | This setting allows fio to manage how big the random/fixed data region |
| 1468 | is when using \fBbuffer_compress_percentage\fR. When |
| 1469 | \fBbuffer_compress_chunk\fR is set to some non-zero value smaller than the |
| 1470 | block size, fio can repeat the random/fixed region throughout the I/O |
| 1471 | buffer at the specified interval (which particularly useful when |
| 1472 | bigger block sizes are used for a job). When set to 0, fio will use a |
| 1473 | chunk size that matches the block size resulting in a single |
| 1474 | random/fixed region within the I/O buffer. Defaults to 512. When the |
| 1475 | unit is omitted, the value is interpreted in bytes. |
| 1476 | .TP |
| 1477 | .BI buffer_pattern \fR=\fPstr |
| 1478 | If set, fio will fill the I/O buffers with this pattern or with the contents |
| 1479 | of a file. If not set, the contents of I/O buffers are defined by the other |
| 1480 | options related to buffer contents. The setting can be any pattern of bytes, |
| 1481 | and can be prefixed with 0x for hex values. It may also be a string, where |
| 1482 | the string must then be wrapped with "". Or it may also be a filename, |
| 1483 | where the filename must be wrapped with '' in which case the file is |
| 1484 | opened and read. Note that not all the file contents will be read if that |
| 1485 | would cause the buffers to overflow. So, for example: |
| 1486 | .RS |
| 1487 | .RS |
| 1488 | .P |
| 1489 | .PD 0 |
| 1490 | buffer_pattern='filename' |
| 1491 | .P |
| 1492 | or: |
| 1493 | .P |
| 1494 | buffer_pattern="abcd" |
| 1495 | .P |
| 1496 | or: |
| 1497 | .P |
| 1498 | buffer_pattern=\-12 |
| 1499 | .P |
| 1500 | or: |
| 1501 | .P |
| 1502 | buffer_pattern=0xdeadface |
| 1503 | .PD |
| 1504 | .RE |
| 1505 | .P |
| 1506 | Also you can combine everything together in any order: |
| 1507 | .RS |
| 1508 | .P |
| 1509 | buffer_pattern=0xdeadface"abcd"\-12'filename' |
| 1510 | .RE |
| 1511 | .RE |
| 1512 | .TP |
| 1513 | .BI dedupe_percentage \fR=\fPint |
| 1514 | If set, fio will generate this percentage of identical buffers when |
| 1515 | writing. These buffers will be naturally dedupable. The contents of the |
| 1516 | buffers depend on what other buffer compression settings have been set. It's |
| 1517 | possible to have the individual buffers either fully compressible, or not at |
| 1518 | all \-\- this option only controls the distribution of unique buffers. Setting |
| 1519 | this option will also enable \fBrefill_buffers\fR to prevent every buffer |
| 1520 | being identical. |
| 1521 | .TP |
| 1522 | .BI dedupe_mode \fR=\fPstr |
| 1523 | If \fBdedupe_percentage\fR is given, then this option controls how fio |
| 1524 | generates the dedupe buffers. |
| 1525 | .RS |
| 1526 | .RS |
| 1527 | .TP |
| 1528 | .B repeat |
| 1529 | .P |
| 1530 | .RS |
| 1531 | Generate dedupe buffers by repeating previous writes |
| 1532 | .RE |
| 1533 | .TP |
| 1534 | .B working_set |
| 1535 | .P |
| 1536 | .RS |
| 1537 | Generate dedupe buffers from working set |
| 1538 | .RE |
| 1539 | .RE |
| 1540 | .P |
| 1541 | \fBrepeat\fR is the default option for fio. Dedupe buffers are generated |
| 1542 | by repeating previous unique write. |
| 1543 | |
| 1544 | \fBworking_set\fR is a more realistic workload. |
| 1545 | With \fBworking_set\fR, \fBdedupe_working_set_percentage\fR should be provided. |
| 1546 | Given that, fio will use the initial unique write buffers as its working set. |
| 1547 | Upon deciding to dedupe, fio will randomly choose a buffer from the working set. |
| 1548 | Note that by using \fBworking_set\fR the dedupe percentage will converge |
| 1549 | to the desired over time while \fBrepeat\fR maintains the desired percentage |
| 1550 | throughout the job. |
| 1551 | .RE |
| 1552 | .RE |
| 1553 | .TP |
| 1554 | .BI dedupe_working_set_percentage \fR=\fPint |
| 1555 | If \fBdedupe_mode\fR is set to \fBworking_set\fR, then this controls |
| 1556 | the percentage of size of the file or device used as the buffers |
| 1557 | fio will choose to generate the dedupe buffers from |
| 1558 | .P |
| 1559 | .RS |
| 1560 | Note that \fBsize\fR needs to be explicitly provided and only 1 file |
| 1561 | per job is supported |
| 1562 | .RE |
| 1563 | .TP |
| 1564 | .BI dedupe_global \fR=\fPbool |
| 1565 | This controls whether the deduplication buffers will be shared amongst |
| 1566 | all jobs that have this option set. The buffers are spread evenly between |
| 1567 | participating jobs. |
| 1568 | .P |
| 1569 | .RS |
| 1570 | Note that \fBdedupe_mode\fR must be set to \fBworking_set\fR for this to work. |
| 1571 | Can be used in combination with compression |
| 1572 | .TP |
| 1573 | .BI invalidate \fR=\fPbool |
| 1574 | Invalidate the buffer/page cache parts of the files to be used prior to |
| 1575 | starting I/O if the platform and file type support it. Defaults to true. |
| 1576 | This will be ignored if \fBpre_read\fR is also specified for the |
| 1577 | same job. |
| 1578 | .TP |
| 1579 | .BI sync \fR=\fPstr |
| 1580 | Whether, and what type, of synchronous I/O to use for writes. The allowed |
| 1581 | values are: |
| 1582 | .RS |
| 1583 | .RS |
| 1584 | .TP |
| 1585 | .B none |
| 1586 | Do not use synchronous IO, the default. |
| 1587 | .TP |
| 1588 | .B 0 |
| 1589 | Same as \fBnone\fR. |
| 1590 | .TP |
| 1591 | .B sync |
| 1592 | Use synchronous file IO. For the majority of I/O engines, |
| 1593 | this means using O_SYNC. |
| 1594 | .TP |
| 1595 | .B 1 |
| 1596 | Same as \fBsync\fR. |
| 1597 | .TP |
| 1598 | .B dsync |
| 1599 | Use synchronous data IO. For the majority of I/O engines, |
| 1600 | this means using O_DSYNC. |
| 1601 | .PD |
| 1602 | .RE |
| 1603 | .RE |
| 1604 | .TP |
| 1605 | .BI iomem \fR=\fPstr "\fR,\fP mem" \fR=\fPstr |
| 1606 | Fio can use various types of memory as the I/O unit buffer. The allowed |
| 1607 | values are: |
| 1608 | .RS |
| 1609 | .RS |
| 1610 | .TP |
| 1611 | .B malloc |
| 1612 | Use memory from \fBmalloc\fR\|(3) as the buffers. Default memory type. |
| 1613 | .TP |
| 1614 | .B shm |
| 1615 | Use shared memory as the buffers. Allocated through \fBshmget\fR\|(2). |
| 1616 | .TP |
| 1617 | .B shmhuge |
| 1618 | Same as \fBshm\fR, but use huge pages as backing. |
| 1619 | .TP |
| 1620 | .B mmap |
| 1621 | Use \fBmmap\fR\|(2) to allocate buffers. May either be anonymous memory, or can |
| 1622 | be file backed if a filename is given after the option. The format |
| 1623 | is `mem=mmap:/path/to/file'. |
| 1624 | .TP |
| 1625 | .B mmaphuge |
| 1626 | Use a memory mapped huge file as the buffer backing. Append filename |
| 1627 | after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file'. |
| 1628 | .TP |
| 1629 | .B mmapshared |
| 1630 | Same as \fBmmap\fR, but use a MMAP_SHARED mapping. |
| 1631 | .TP |
| 1632 | .B cudamalloc |
| 1633 | Use GPU memory as the buffers for GPUDirect RDMA benchmark. |
| 1634 | The \fBioengine\fR must be \fBrdma\fR. |
| 1635 | .RE |
| 1636 | .P |
| 1637 | The area allocated is a function of the maximum allowed bs size for the job, |
| 1638 | multiplied by the I/O depth given. Note that for \fBshmhuge\fR and |
| 1639 | \fBmmaphuge\fR to work, the system must have free huge pages allocated. This |
| 1640 | can normally be checked and set by reading/writing |
| 1641 | `/proc/sys/vm/nr_hugepages' on a Linux system. Fio assumes a huge page |
| 1642 | is 2 or 4MiB in size depending on the platform. So to calculate the number of |
| 1643 | huge pages you need for a given job file, add up the I/O depth of all jobs |
| 1644 | (normally one unless \fBiodepth\fR is used) and multiply by the maximum bs set. |
| 1645 | Then divide that number by the huge page size. You can see the size of the huge |
| 1646 | pages in `/proc/meminfo'. If no huge pages are allocated by having a non-zero |
| 1647 | number in `nr_hugepages', using \fBmmaphuge\fR or \fBshmhuge\fR will fail. Also |
| 1648 | see \fBhugepage\-size\fR. |
| 1649 | .P |
| 1650 | \fBmmaphuge\fR also needs to have hugetlbfs mounted and the file location |
| 1651 | should point there. So if it's mounted in `/huge', you would use |
| 1652 | `mem=mmaphuge:/huge/somefile'. |
| 1653 | .RE |
| 1654 | .TP |
| 1655 | .BI iomem_align \fR=\fPint "\fR,\fP mem_align" \fR=\fPint |
| 1656 | This indicates the memory alignment of the I/O memory buffers. Note that |
| 1657 | the given alignment is applied to the first I/O unit buffer, if using |
| 1658 | \fBiodepth\fR the alignment of the following buffers are given by the |
| 1659 | \fBbs\fR used. In other words, if using a \fBbs\fR that is a |
| 1660 | multiple of the page sized in the system, all buffers will be aligned to |
| 1661 | this value. If using a \fBbs\fR that is not page aligned, the alignment |
| 1662 | of subsequent I/O memory buffers is the sum of the \fBiomem_align\fR and |
| 1663 | \fBbs\fR used. |
| 1664 | .TP |
| 1665 | .BI hugepage\-size \fR=\fPint |
| 1666 | Defines the size of a huge page. Must at least be equal to the system setting, |
| 1667 | see `/proc/meminfo' and `/sys/kernel/mm/hugepages/'. Defaults to 2 or 4MiB |
| 1668 | depending on the platform. Should probably always be a multiple of megabytes, |
| 1669 | so using `hugepage\-size=Xm' is the preferred way to set this to avoid setting |
| 1670 | a non-pow-2 bad value. |
| 1671 | .TP |
| 1672 | .BI lockmem \fR=\fPint |
| 1673 | Pin the specified amount of memory with \fBmlock\fR\|(2). Can be used to |
| 1674 | simulate a smaller amount of memory. The amount specified is per worker. |
| 1675 | .SS "I/O size" |
| 1676 | .TP |
| 1677 | .BI size \fR=\fPint[%|z] |
| 1678 | The total size of file I/O for each thread of this job. Fio will run until |
| 1679 | this many bytes has been transferred, unless runtime is altered by other means |
| 1680 | such as (1) \fBruntime\fR, (2) \fBio_size\fR, (3) \fBnumber_ios\fR, (4) |
| 1681 | gaps/holes while doing I/O's such as `rw=read:16K', or (5) sequential I/O |
| 1682 | reaching end of the file which is possible when \fBpercentage_random\fR is |
| 1683 | less than 100. |
| 1684 | Fio will divide this size between the available files determined by options |
| 1685 | such as \fBnrfiles\fR, \fBfilename\fR, unless \fBfilesize\fR is |
| 1686 | specified by the job. If the result of division happens to be 0, the size is |
| 1687 | set to the physical size of the given files or devices if they exist. |
| 1688 | If this option is not specified, fio will use the full size of the given |
| 1689 | files or devices. If the files do not exist, size must be given. It is also |
| 1690 | possible to give size as a percentage between 1 and 100. If `size=20%' is |
| 1691 | given, fio will use 20% of the full size of the given files or devices. In ZBD mode, |
| 1692 | size can be given in units of number of zones using 'z'. Can be combined with \fBoffset\fR to |
| 1693 | constrain the start and end range that I/O will be done within. |
| 1694 | .TP |
| 1695 | .BI io_size \fR=\fPint[%|z] "\fR,\fB io_limit" \fR=\fPint[%|z] |
| 1696 | Normally fio operates within the region set by \fBsize\fR, which means |
| 1697 | that the \fBsize\fR option sets both the region and size of I/O to be |
| 1698 | performed. Sometimes that is not what you want. With this option, it is |
| 1699 | possible to define just the amount of I/O that fio should do. For instance, |
| 1700 | if \fBsize\fR is set to 20GiB and \fBio_size\fR is set to 5GiB, fio |
| 1701 | will perform I/O within the first 20GiB but exit when 5GiB have been |
| 1702 | done. The opposite is also possible \-\- if \fBsize\fR is set to 20GiB, |
| 1703 | and \fBio_size\fR is set to 40GiB, then fio will do 40GiB of I/O within |
| 1704 | the 0..20GiB region. Value can be set as percentage: \fBio_size\fR=N%. |
| 1705 | In this case \fBio_size\fR multiplies \fBsize\fR= value. In ZBD mode, value can |
| 1706 | also be set as number of zones using 'z'. |
| 1707 | .TP |
| 1708 | .BI filesize \fR=\fPirange(int) |
| 1709 | Individual file sizes. May be a range, in which case fio will select sizes |
| 1710 | for files at random within the given range. If not given, each created file |
| 1711 | is the same size. This option overrides \fBsize\fR in terms of file size, |
| 1712 | i.e. \fBsize\fR becomes merely the default for \fBio_size\fR (and |
| 1713 | has no effect it all if \fBio_size\fR is set explicitly). |
| 1714 | .TP |
| 1715 | .BI file_append \fR=\fPbool |
| 1716 | Perform I/O after the end of the file. Normally fio will operate within the |
| 1717 | size of a file. If this option is set, then fio will append to the file |
| 1718 | instead. This has identical behavior to setting \fBoffset\fR to the size |
| 1719 | of a file. This option is ignored on non-regular files. |
| 1720 | .TP |
| 1721 | .BI fill_device \fR=\fPbool "\fR,\fB fill_fs" \fR=\fPbool |
| 1722 | Sets size to something really large and waits for ENOSPC (no space left on |
| 1723 | device) or EDQUOT (disk quota exceeded) |
| 1724 | as the terminating condition. Only makes sense with sequential |
| 1725 | write. For a read workload, the mount point will be filled first then I/O |
| 1726 | started on the result. |
| 1727 | .SS "I/O engine" |
| 1728 | .TP |
| 1729 | .BI ioengine \fR=\fPstr |
| 1730 | Defines how the job issues I/O to the file. The following types are defined: |
| 1731 | .RS |
| 1732 | .RS |
| 1733 | .TP |
| 1734 | .B sync |
| 1735 | Basic \fBread\fR\|(2) or \fBwrite\fR\|(2) |
| 1736 | I/O. \fBlseek\fR\|(2) is used to position the I/O location. |
| 1737 | See \fBfsync\fR and \fBfdatasync\fR for syncing write I/Os. |
| 1738 | .TP |
| 1739 | .B psync |
| 1740 | Basic \fBpread\fR\|(2) or \fBpwrite\fR\|(2) I/O. Default on |
| 1741 | all supported operating systems except for Windows. |
| 1742 | .TP |
| 1743 | .B vsync |
| 1744 | Basic \fBreadv\fR\|(2) or \fBwritev\fR\|(2) I/O. Will emulate |
| 1745 | queuing by coalescing adjacent I/Os into a single submission. |
| 1746 | .TP |
| 1747 | .B pvsync |
| 1748 | Basic \fBpreadv\fR\|(2) or \fBpwritev\fR\|(2) I/O. |
| 1749 | .TP |
| 1750 | .B pvsync2 |
| 1751 | Basic \fBpreadv2\fR\|(2) or \fBpwritev2\fR\|(2) I/O. |
| 1752 | .TP |
| 1753 | .B io_uring |
| 1754 | Fast Linux native asynchronous I/O. Supports async IO |
| 1755 | for both direct and buffered IO. |
| 1756 | This engine defines engine specific options. |
| 1757 | .TP |
| 1758 | .B io_uring_cmd |
| 1759 | Fast Linux native asynchronous I/O for passthrough commands. |
| 1760 | This engine defines engine specific options. |
| 1761 | .TP |
| 1762 | .B libaio |
| 1763 | Linux native asynchronous I/O. Note that Linux may only support |
| 1764 | queued behavior with non-buffered I/O (set `direct=1' or |
| 1765 | `buffered=0'). |
| 1766 | This engine defines engine specific options. |
| 1767 | .TP |
| 1768 | .B posixaio |
| 1769 | POSIX asynchronous I/O using \fBaio_read\fR\|(3) and |
| 1770 | \fBaio_write\fR\|(3). |
| 1771 | .TP |
| 1772 | .B solarisaio |
| 1773 | Solaris native asynchronous I/O. |
| 1774 | .TP |
| 1775 | .B windowsaio |
| 1776 | Windows native asynchronous I/O. Default on Windows. |
| 1777 | .TP |
| 1778 | .B mmap |
| 1779 | File is memory mapped with \fBmmap\fR\|(2) and data copied |
| 1780 | to/from using \fBmemcpy\fR\|(3). |
| 1781 | .TP |
| 1782 | .B splice |
| 1783 | \fBsplice\fR\|(2) is used to transfer the data and |
| 1784 | \fBvmsplice\fR\|(2) to transfer data from user space to the |
| 1785 | kernel. |
| 1786 | .TP |
| 1787 | .B sg |
| 1788 | SCSI generic sg v3 I/O. May either be synchronous using the SG_IO |
| 1789 | ioctl, or if the target is an sg character device we use |
| 1790 | \fBread\fR\|(2) and \fBwrite\fR\|(2) for asynchronous |
| 1791 | I/O. Requires \fBfilename\fR option to specify either block or |
| 1792 | character devices. This engine supports trim operations. The |
| 1793 | sg engine includes engine specific options. |
| 1794 | .TP |
| 1795 | .B libzbc |
| 1796 | Read, write, trim and ZBC/ZAC operations to a zoned block device using |
| 1797 | \fBlibzbc\fR library. The target can be either an SG character device or |
| 1798 | a block device file. |
| 1799 | .TP |
| 1800 | .B null |
| 1801 | Doesn't transfer any data, just pretends to. This is mainly used to |
| 1802 | exercise fio itself and for debugging/testing purposes. |
| 1803 | .TP |
| 1804 | .B net |
| 1805 | Transfer over the network to given `host:port'. Depending on the |
| 1806 | \fBprotocol\fR used, the \fBhostname\fR, \fBport\fR, |
| 1807 | \fBlisten\fR and \fBfilename\fR options are used to specify |
| 1808 | what sort of connection to make, while the \fBprotocol\fR option |
| 1809 | determines which protocol will be used. This engine defines engine |
| 1810 | specific options. |
| 1811 | .TP |
| 1812 | .B netsplice |
| 1813 | Like \fBnet\fR, but uses \fBsplice\fR\|(2) and |
| 1814 | \fBvmsplice\fR\|(2) to map data and send/receive. |
| 1815 | This engine defines engine specific options. |
| 1816 | .TP |
| 1817 | .B cpuio |
| 1818 | Doesn't transfer any data, but burns CPU cycles according to the |
| 1819 | \fBcpuload\fR, \fBcpuchunks\fR and \fBcpumode\fR options. |
| 1820 | A job never finishes unless there is at least one non-cpuio job. |
| 1821 | .RS |
| 1822 | .P |
| 1823 | .PD 0 |
| 1824 | \fBcpuload\fR\=85 will cause that job to do nothing but burn 85% of the CPU. |
| 1825 | In case of SMP machines, use \fBnumjobs=<nr_of_cpu>\fR\ to get desired CPU usage, |
| 1826 | as the cpuload only loads a single CPU at the desired rate. |
| 1827 | |
| 1828 | .P |
| 1829 | \fBcpumode\fR\=qsort replace the default noop instructions loop |
| 1830 | by a qsort algorithm to consume more energy. |
| 1831 | |
| 1832 | .P |
| 1833 | .RE |
| 1834 | .TP |
| 1835 | .B rdma |
| 1836 | The RDMA I/O engine supports both RDMA memory semantics |
| 1837 | (RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the |
| 1838 | InfiniBand, RoCE and iWARP protocols. This engine defines engine |
| 1839 | specific options. |
| 1840 | .TP |
| 1841 | .B falloc |
| 1842 | I/O engine that does regular fallocate to simulate data transfer as |
| 1843 | fio ioengine. |
| 1844 | .RS |
| 1845 | .P |
| 1846 | .PD 0 |
| 1847 | DDIR_READ does fallocate(,mode = FALLOC_FL_KEEP_SIZE,). |
| 1848 | .P |
| 1849 | DIR_WRITE does fallocate(,mode = 0). |
| 1850 | .P |
| 1851 | DDIR_TRIM does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE). |
| 1852 | .PD |
| 1853 | .RE |
| 1854 | .TP |
| 1855 | .B ftruncate |
| 1856 | I/O engine that sends \fBftruncate\fR\|(2) operations in response |
| 1857 | to write (DDIR_WRITE) events. Each ftruncate issued sets the file's |
| 1858 | size to the current block offset. \fBblocksize\fR is ignored. |
| 1859 | .TP |
| 1860 | .B e4defrag |
| 1861 | I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate |
| 1862 | defragment activity in request to DDIR_WRITE event. |
| 1863 | .TP |
| 1864 | .B rados |
| 1865 | I/O engine supporting direct access to Ceph Reliable Autonomic Distributed |
| 1866 | Object Store (RADOS) via librados. This ioengine defines engine specific |
| 1867 | options. |
| 1868 | .TP |
| 1869 | .B rbd |
| 1870 | I/O engine supporting direct access to Ceph Rados Block Devices |
| 1871 | (RBD) via librbd without the need to use the kernel rbd driver. This |
| 1872 | ioengine defines engine specific options. |
| 1873 | .TP |
| 1874 | .B http |
| 1875 | I/O engine supporting GET/PUT requests over HTTP(S) with libcurl to |
| 1876 | a WebDAV or S3 endpoint. This ioengine defines engine specific options. |
| 1877 | |
| 1878 | This engine only supports direct IO of iodepth=1; you need to scale this |
| 1879 | via numjobs. blocksize defines the size of the objects to be created. |
| 1880 | |
| 1881 | TRIM is translated to object deletion. |
| 1882 | .TP |
| 1883 | .B gfapi |
| 1884 | Using GlusterFS libgfapi sync interface to direct access to |
| 1885 | GlusterFS volumes without having to go through FUSE. This ioengine |
| 1886 | defines engine specific options. |
| 1887 | .TP |
| 1888 | .B gfapi_async |
| 1889 | Using GlusterFS libgfapi async interface to direct access to |
| 1890 | GlusterFS volumes without having to go through FUSE. This ioengine |
| 1891 | defines engine specific options. |
| 1892 | .TP |
| 1893 | .B libhdfs |
| 1894 | Read and write through Hadoop (HDFS). The \fBfilename\fR option |
| 1895 | is used to specify host,port of the hdfs name\-node to connect. This |
| 1896 | engine interprets offsets a little differently. In HDFS, files once |
| 1897 | created cannot be modified so random writes are not possible. To |
| 1898 | imitate this the libhdfs engine expects a bunch of small files to be |
| 1899 | created over HDFS and will randomly pick a file from them |
| 1900 | based on the offset generated by fio backend (see the example |
| 1901 | job file to create such files, use `rw=write' option). Please |
| 1902 | note, it may be necessary to set environment variables to work |
| 1903 | with HDFS/libhdfs properly. Each job uses its own connection to |
| 1904 | HDFS. |
| 1905 | .TP |
| 1906 | .B mtd |
| 1907 | Read, write and erase an MTD character device (e.g., |
| 1908 | `/dev/mtd0'). Discards are treated as erases. Depending on the |
| 1909 | underlying device type, the I/O may have to go in a certain pattern, |
| 1910 | e.g., on NAND, writing sequentially to erase blocks and discarding |
| 1911 | before overwriting. The \fBtrimwrite\fR mode works well for this |
| 1912 | constraint. |
| 1913 | .TP |
| 1914 | .B pmemblk |
| 1915 | Read and write using filesystem DAX to a file on a filesystem |
| 1916 | mounted with DAX on a persistent memory device through the PMDK |
| 1917 | libpmemblk library. |
| 1918 | .TP |
| 1919 | .B dev\-dax |
| 1920 | Read and write using device DAX to a persistent memory device (e.g., |
| 1921 | /dev/dax0.0) through the PMDK libpmem library. |
| 1922 | .TP |
| 1923 | .B external |
| 1924 | Prefix to specify loading an external I/O engine object file. Append |
| 1925 | the engine filename, e.g. `ioengine=external:/tmp/foo.o' to load |
| 1926 | ioengine `foo.o' in `/tmp'. The path can be either |
| 1927 | absolute or relative. See `engines/skeleton_external.c' in the fio source for |
| 1928 | details of writing an external I/O engine. |
| 1929 | .TP |
| 1930 | .B filecreate |
| 1931 | Simply create the files and do no I/O to them. You still need to set |
| 1932 | \fBfilesize\fR so that all the accounting still occurs, but no actual I/O will be |
| 1933 | done other than creating the file. |
| 1934 | .TP |
| 1935 | .B filestat |
| 1936 | Simply do stat() and do no I/O to the file. You need to set 'filesize' |
| 1937 | and 'nrfiles', so that files will be created. |
| 1938 | This engine is to measure file lookup and meta data access. |
| 1939 | .TP |
| 1940 | .B filedelete |
| 1941 | Simply delete files by unlink() and do no I/O to the file. You need to set 'filesize' |
| 1942 | and 'nrfiles', so that files will be created. |
| 1943 | This engine is to measure file delete. |
| 1944 | .TP |
| 1945 | .B libpmem |
| 1946 | Read and write using mmap I/O to a file on a filesystem |
| 1947 | mounted with DAX on a persistent memory device through the PMDK |
| 1948 | libpmem library. |
| 1949 | .TP |
| 1950 | .B ime_psync |
| 1951 | Synchronous read and write using DDN's Infinite Memory Engine (IME). This |
| 1952 | engine is very basic and issues calls to IME whenever an IO is queued. |
| 1953 | .TP |
| 1954 | .B ime_psyncv |
| 1955 | Synchronous read and write using DDN's Infinite Memory Engine (IME). This |
| 1956 | engine uses iovecs and will try to stack as much IOs as possible (if the IOs |
| 1957 | are "contiguous" and the IO depth is not exceeded) before issuing a call to IME. |
| 1958 | .TP |
| 1959 | .B ime_aio |
| 1960 | Asynchronous read and write using DDN's Infinite Memory Engine (IME). This |
| 1961 | engine will try to stack as much IOs as possible by creating requests for IME. |
| 1962 | FIO will then decide when to commit these requests. |
| 1963 | .TP |
| 1964 | .B libiscsi |
| 1965 | Read and write iscsi lun with libiscsi. |
| 1966 | .TP |
| 1967 | .B nbd |
| 1968 | Synchronous read and write a Network Block Device (NBD). |
| 1969 | .TP |
| 1970 | .B libcufile |
| 1971 | I/O engine supporting libcufile synchronous access to nvidia-fs and a |
| 1972 | GPUDirect Storage-supported filesystem. This engine performs |
| 1973 | I/O without transferring buffers between user-space and the kernel, |
| 1974 | unless \fBverify\fR is set or \fBcuda_io\fR is \fBposix\fR. \fBiomem\fR must |
| 1975 | not be \fBcudamalloc\fR. This ioengine defines engine specific options. |
| 1976 | .TP |
| 1977 | .B dfs |
| 1978 | I/O engine supporting asynchronous read and write operations to the DAOS File |
| 1979 | System (DFS) via libdfs. |
| 1980 | .TP |
| 1981 | .B nfs |
| 1982 | I/O engine supporting asynchronous read and write operations to |
| 1983 | NFS filesystems from userspace via libnfs. This is useful for |
| 1984 | achieving higher concurrency and thus throughput than is possible |
| 1985 | via kernel NFS. |
| 1986 | .TP |
| 1987 | .B exec |
| 1988 | Execute 3rd party tools. Could be used to perform monitoring during jobs runtime. |
| 1989 | .TP |
| 1990 | .B xnvme |
| 1991 | I/O engine using the xNVMe C API, for NVMe devices. The xnvme engine provides |
| 1992 | flexibility to access GNU/Linux Kernel NVMe driver via libaio, IOCTLs, io_uring, |
| 1993 | the SPDK NVMe driver, or your own custom NVMe driver. The xnvme engine includes |
| 1994 | engine specific options. (See \fIhttps://xnvme.io/\fR). |
| 1995 | .TP |
| 1996 | .B libblkio |
| 1997 | Use the libblkio library (\fIhttps://gitlab.com/libblkio/libblkio\fR). The |
| 1998 | specific driver to use must be set using \fBlibblkio_driver\fR. If |
| 1999 | \fBmem\fR/\fBiomem\fR is not specified, memory allocation is delegated to |
| 2000 | libblkio (and so is guaranteed to work with the selected driver). |
| 2001 | .SS "I/O engine specific parameters" |
| 2002 | In addition, there are some parameters which are only valid when a specific |
| 2003 | \fBioengine\fR is in use. These are used identically to normal parameters, |
| 2004 | with the caveat that when used on the command line, they must come after the |
| 2005 | \fBioengine\fR that defines them is selected. |
| 2006 | .TP |
| 2007 | .BI (io_uring,libaio)cmdprio_percentage \fR=\fPint[,int] |
| 2008 | Set the percentage of I/O that will be issued with the highest priority. |
| 2009 | Default: 0. A single value applies to reads and writes. Comma-separated |
| 2010 | values may be specified for reads and writes. For this option to be effective, |
| 2011 | NCQ priority must be supported and enabled, and `direct=1' option must be |
| 2012 | used. fio must also be run as the root user. Unlike slat/clat/lat stats, which |
| 2013 | can be tracked and reported independently, per priority stats only track and |
| 2014 | report a single type of latency. By default, completion latency (clat) will be |
| 2015 | reported, if \fBlat_percentiles\fR is set, total latency (lat) will be reported. |
| 2016 | .TP |
| 2017 | .BI (io_uring,libaio)cmdprio_class \fR=\fPint[,int] |
| 2018 | Set the I/O priority class to use for I/Os that must be issued with a |
| 2019 | priority when \fBcmdprio_percentage\fR or \fBcmdprio_bssplit\fR is set. |
| 2020 | If not specified when \fBcmdprio_percentage\fR or \fBcmdprio_bssplit\fR |
| 2021 | is set, this defaults to the highest priority class. A single value applies |
| 2022 | to reads and writes. Comma-separated values may be specified for reads and |
| 2023 | writes. See man \fBionice\fR\|(1). See also the \fBprioclass\fR option. |
| 2024 | .TP |
| 2025 | .BI (io_uring,libaio)cmdprio \fR=\fPint[,int] |
| 2026 | Set the I/O priority value to use for I/Os that must be issued with a |
| 2027 | priority when \fBcmdprio_percentage\fR or \fBcmdprio_bssplit\fR is set. |
| 2028 | If not specified when \fBcmdprio_percentage\fR or \fBcmdprio_bssplit\fR |
| 2029 | is set, this defaults to 0. Linux limits us to a positive value between |
| 2030 | 0 and 7, with 0 being the highest. A single value applies to reads and writes. |
| 2031 | Comma-separated values may be specified for reads and writes. See man |
| 2032 | \fBionice\fR\|(1). Refer to an appropriate manpage for other operating systems |
| 2033 | since the meaning of priority may differ. See also the \fBprio\fR option. |
| 2034 | .TP |
| 2035 | .BI (io_uring,libaio)cmdprio_bssplit \fR=\fPstr[,str] |
| 2036 | To get a finer control over I/O priority, this option allows specifying |
| 2037 | the percentage of IOs that must have a priority set depending on the block |
| 2038 | size of the IO. This option is useful only when used together with the option |
| 2039 | \fBbssplit\fR, that is, multiple different block sizes are used for reads and |
| 2040 | writes. |
| 2041 | .RS |
| 2042 | .P |
| 2043 | The first accepted format for this option is the same as the format of the |
| 2044 | \fBbssplit\fR option: |
| 2045 | .RS |
| 2046 | .P |
| 2047 | cmdprio_bssplit=blocksize/percentage:blocksize/percentage |
| 2048 | .RE |
| 2049 | .P |
| 2050 | In this case, each entry will use the priority class and priority level defined |
| 2051 | by the options \fBcmdprio_class\fR and \fBcmdprio\fR respectively. |
| 2052 | .P |
| 2053 | The second accepted format for this option is: |
| 2054 | .RS |
| 2055 | .P |
| 2056 | cmdprio_bssplit=blocksize/percentage/class/level:blocksize/percentage/class/level |
| 2057 | .RE |
| 2058 | .P |
| 2059 | In this case, the priority class and priority level is defined inside each |
| 2060 | entry. In comparison with the first accepted format, the second accepted format |
| 2061 | does not restrict all entries to have the same priority class and priority |
| 2062 | level. |
| 2063 | .P |
| 2064 | For both formats, only the read and write data directions are supported, values |
| 2065 | for trim IOs are ignored. This option is mutually exclusive with the |
| 2066 | \fBcmdprio_percentage\fR option. |
| 2067 | .RE |
| 2068 | .TP |
| 2069 | .BI (io_uring,io_uring_cmd)fixedbufs |
| 2070 | If fio is asked to do direct IO, then Linux will map pages for each IO call, and |
| 2071 | release them when IO is done. If this option is set, the pages are pre-mapped |
| 2072 | before IO is started. This eliminates the need to map and release for each IO. |
| 2073 | This is more efficient, and reduces the IO latency as well. |
| 2074 | .TP |
| 2075 | .BI (io_uring,io_uring_cmd)nonvectored \fR=\fPint |
| 2076 | With this option, fio will use non-vectored read/write commands, where address |
| 2077 | must contain the address directly. Default is -1. |
| 2078 | .TP |
| 2079 | .BI (io_uring,io_uring_cmd)force_async |
| 2080 | Normal operation for io_uring is to try and issue an sqe as non-blocking first, |
| 2081 | and if that fails, execute it in an async manner. With this option set to N, |
| 2082 | then every N request fio will ask sqe to be issued in an async manner. Default |
| 2083 | is 0. |
| 2084 | .TP |
| 2085 | .BI (io_uring,io_uring_cmd,xnvme)hipri |
| 2086 | If this option is set, fio will attempt to use polled IO completions. Normal IO |
| 2087 | completions generate interrupts to signal the completion of IO, polled |
| 2088 | completions do not. Hence they are require active reaping by the application. |
| 2089 | The benefits are more efficient IO for high IOPS scenarios, and lower latencies |
| 2090 | for low queue depth IO. |
| 2091 | .TP |
| 2092 | .BI (io_uring,io_uring_cmd)registerfiles |
| 2093 | With this option, fio registers the set of files being used with the kernel. |
| 2094 | This avoids the overhead of managing file counts in the kernel, making the |
| 2095 | submission and completion part more lightweight. Required for the below |
| 2096 | sqthread_poll option. |
| 2097 | .TP |
| 2098 | .BI (io_uring,io_uring_cmd,xnvme)sqthread_poll |
| 2099 | Normally fio will submit IO by issuing a system call to notify the kernel of |
| 2100 | available items in the SQ ring. If this option is set, the act of submitting IO |
| 2101 | will be done by a polling thread in the kernel. This frees up cycles for fio, at |
| 2102 | the cost of using more CPU in the system. As submission is just the time it |
| 2103 | takes to fill in the sqe entries and any syscall required to wake up the idle |
| 2104 | kernel thread, fio will not report submission latencies. |
| 2105 | .TP |
| 2106 | .BI (io_uring,io_uring_cmd)sqthread_poll_cpu \fR=\fPint |
| 2107 | When `sqthread_poll` is set, this option provides a way to define which CPU |
| 2108 | should be used for the polling thread. |
| 2109 | .TP |
| 2110 | .BI (io_uring_cmd)cmd_type \fR=\fPstr |
| 2111 | Specifies the type of uring passthrough command to be used. Supported |
| 2112 | value is nvme. Default is nvme. |
| 2113 | .TP |
| 2114 | .BI (libaio)userspace_reap |
| 2115 | Normally, with the libaio engine in use, fio will use the |
| 2116 | \fBio_getevents\fR\|(3) system call to reap newly returned events. With |
| 2117 | this flag turned on, the AIO ring will be read directly from user-space to |
| 2118 | reap events. The reaping mode is only enabled when polling for a minimum of |
| 2119 | 0 events (e.g. when `iodepth_batch_complete=0'). |
| 2120 | .TP |
| 2121 | .BI (pvsync2)hipri |
| 2122 | Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority |
| 2123 | than normal. |
| 2124 | .TP |
| 2125 | .BI (pvsync2)hipri_percentage |
| 2126 | When hipri is set this determines the probability of a pvsync2 I/O being high |
| 2127 | priority. The default is 100%. |
| 2128 | .TP |
| 2129 | .BI (pvsync2,libaio,io_uring,io_uring_cmd)nowait \fR=\fPbool |
| 2130 | By default if a request cannot be executed immediately (e.g. resource starvation, |
| 2131 | waiting on locks) it is queued and the initiating process will be blocked until |
| 2132 | the required resource becomes free. |
| 2133 | This option sets the RWF_NOWAIT flag (supported from the 4.14 Linux kernel) and |
| 2134 | the call will return instantly with EAGAIN or a partial result rather than waiting. |
| 2135 | |
| 2136 | It is useful to also use \fBignore_error\fR=EAGAIN when using this option. |
| 2137 | Note: glibc 2.27, 2.28 have a bug in syscall wrappers preadv2, pwritev2. |
| 2138 | They return EOPNOTSUP instead of EAGAIN. |
| 2139 | |
| 2140 | For cached I/O, using this option usually means a request operates only with |
| 2141 | cached data. Currently the RWF_NOWAIT flag does not supported for cached write. |
| 2142 | For direct I/O, requests will only succeed if cache invalidation isn't required, |
| 2143 | file blocks are fully allocated and the disk request could be issued immediately. |
| 2144 | .TP |
| 2145 | .BI (cpuio)cpuload \fR=\fPint |
| 2146 | Attempt to use the specified percentage of CPU cycles. This is a mandatory |
| 2147 | option when using cpuio I/O engine. |
| 2148 | .TP |
| 2149 | .BI (cpuio)cpuchunks \fR=\fPint |
| 2150 | Split the load into cycles of the given time. In microseconds. |
| 2151 | .TP |
| 2152 | .BI (cpuio)cpumode \fR=\fPstr |
| 2153 | Specify how to stress the CPU. It can take these two values: |
| 2154 | .RS |
| 2155 | .RS |
| 2156 | .TP |
| 2157 | .B noop |
| 2158 | This is the default and directs the CPU to execute noop instructions. |
| 2159 | .TP |
| 2160 | .B qsort |
| 2161 | Replace the default noop instructions with a qsort algorithm to consume more energy. |
| 2162 | .RE |
| 2163 | .RE |
| 2164 | .TP |
| 2165 | .BI (cpuio)exit_on_io_done \fR=\fPbool |
| 2166 | Detect when I/O threads are done, then exit. |
| 2167 | .TP |
| 2168 | .BI (libhdfs)namenode \fR=\fPstr |
| 2169 | The hostname or IP address of a HDFS cluster namenode to contact. |
| 2170 | .TP |
| 2171 | .BI (libhdfs)port \fR=\fPint |
| 2172 | The listening port of the HFDS cluster namenode. |
| 2173 | .TP |
| 2174 | .BI (netsplice,net)port \fR=\fPint |
| 2175 | The TCP or UDP port to bind to or connect to. If this is used with |
| 2176 | \fBnumjobs\fR to spawn multiple instances of the same job type, then |
| 2177 | this will be the starting port number since fio will use a range of |
| 2178 | ports. |
| 2179 | .TP |
| 2180 | .BI (rdma,librpma_*)port \fR=\fPint |
| 2181 | The port to use for RDMA-CM communication. This should be the same |
| 2182 | value on the client and the server side. |
| 2183 | .TP |
| 2184 | .BI (netsplice,net,rdma)hostname \fR=\fPstr |
| 2185 | The hostname or IP address to use for TCP, UDP or RDMA-CM based I/O. |
| 2186 | If the job is a TCP listener or UDP reader, the hostname is not used |
| 2187 | and must be omitted unless it is a valid UDP multicast address. |
| 2188 | .TP |
| 2189 | .BI (librpma_*)serverip \fR=\fPstr |
| 2190 | The IP address to be used for RDMA-CM based I/O. |
| 2191 | .TP |
| 2192 | .BI (librpma_*_server)direct_write_to_pmem \fR=\fPbool |
| 2193 | Set to 1 only when Direct Write to PMem from the remote host is possible. Otherwise, set to 0. |
| 2194 | .TP |
| 2195 | .BI (librpma_*_server)busy_wait_polling \fR=\fPbool |
| 2196 | Set to 0 to wait for completion instead of busy-wait polling completion. |
| 2197 | Default: 1. |
| 2198 | .TP |
| 2199 | .BI (netsplice,net)interface \fR=\fPstr |
| 2200 | The IP address of the network interface used to send or receive UDP |
| 2201 | multicast. |
| 2202 | .TP |
| 2203 | .BI (netsplice,net)ttl \fR=\fPint |
| 2204 | Time\-to\-live value for outgoing UDP multicast packets. Default: 1. |
| 2205 | .TP |
| 2206 | .BI (netsplice,net)nodelay \fR=\fPbool |
| 2207 | Set TCP_NODELAY on TCP connections. |
| 2208 | .TP |
| 2209 | .BI (netsplice,net)protocol \fR=\fPstr "\fR,\fP proto" \fR=\fPstr |
| 2210 | The network protocol to use. Accepted values are: |
| 2211 | .RS |
| 2212 | .RS |
| 2213 | .TP |
| 2214 | .B tcp |
| 2215 | Transmission control protocol. |
| 2216 | .TP |
| 2217 | .B tcpv6 |
| 2218 | Transmission control protocol V6. |
| 2219 | .TP |
| 2220 | .B udp |
| 2221 | User datagram protocol. |
| 2222 | .TP |
| 2223 | .B udpv6 |
| 2224 | User datagram protocol V6. |
| 2225 | .TP |
| 2226 | .B unix |
| 2227 | UNIX domain socket. |
| 2228 | .RE |
| 2229 | .P |
| 2230 | When the protocol is TCP or UDP, the port must also be given, as well as the |
| 2231 | hostname if the job is a TCP listener or UDP reader. For unix sockets, the |
| 2232 | normal \fBfilename\fR option should be used and the port is invalid. |
| 2233 | .RE |
| 2234 | .TP |
| 2235 | .BI (netsplice,net)listen |
| 2236 | For TCP network connections, tell fio to listen for incoming connections |
| 2237 | rather than initiating an outgoing connection. The \fBhostname\fR must |
| 2238 | be omitted if this option is used. |
| 2239 | .TP |
| 2240 | .BI (netsplice,net)pingpong |
| 2241 | Normally a network writer will just continue writing data, and a network |
| 2242 | reader will just consume packages. If `pingpong=1' is set, a writer will |
| 2243 | send its normal payload to the reader, then wait for the reader to send the |
| 2244 | same payload back. This allows fio to measure network latencies. The |
| 2245 | submission and completion latencies then measure local time spent sending or |
| 2246 | receiving, and the completion latency measures how long it took for the |
| 2247 | other end to receive and send back. For UDP multicast traffic |
| 2248 | `pingpong=1' should only be set for a single reader when multiple readers |
| 2249 | are listening to the same address. |
| 2250 | .TP |
| 2251 | .BI (netsplice,net)window_size \fR=\fPint |
| 2252 | Set the desired socket buffer size for the connection. |
| 2253 | .TP |
| 2254 | .BI (netsplice,net)mss \fR=\fPint |
| 2255 | Set the TCP maximum segment size (TCP_MAXSEG). |
| 2256 | .TP |
| 2257 | .BI (e4defrag)donorname \fR=\fPstr |
| 2258 | File will be used as a block donor (swap extents between files). |
| 2259 | .TP |
| 2260 | .BI (e4defrag)inplace \fR=\fPint |
| 2261 | Configure donor file blocks allocation strategy: |
| 2262 | .RS |
| 2263 | .RS |
| 2264 | .TP |
| 2265 | .B 0 |
| 2266 | Default. Preallocate donor's file on init. |
| 2267 | .TP |
| 2268 | .B 1 |
| 2269 | Allocate space immediately inside defragment event, and free right |
| 2270 | after event. |
| 2271 | .RE |
| 2272 | .RE |
| 2273 | .TP |
| 2274 | .BI (rbd,rados)clustername \fR=\fPstr |
| 2275 | Specifies the name of the Ceph cluster. |
| 2276 | .TP |
| 2277 | .BI (rbd)rbdname \fR=\fPstr |
| 2278 | Specifies the name of the RBD. |
| 2279 | .TP |
| 2280 | .BI (rbd,rados)pool \fR=\fPstr |
| 2281 | Specifies the name of the Ceph pool containing RBD or RADOS data. |
| 2282 | .TP |
| 2283 | .BI (rbd,rados)clientname \fR=\fPstr |
| 2284 | Specifies the username (without the 'client.' prefix) used to access the |
| 2285 | Ceph cluster. If the \fBclustername\fR is specified, the \fBclientname\fR shall be |
| 2286 | the full *type.id* string. If no type. prefix is given, fio will add 'client.' |
| 2287 | by default. |
| 2288 | .TP |
| 2289 | .BI (rados)conf \fR=\fPstr |
| 2290 | Specifies the configuration path of ceph cluster, so conf file does not |
| 2291 | have to be /etc/ceph/ceph.conf. |
| 2292 | .TP |
| 2293 | .BI (rbd,rados)busy_poll \fR=\fPbool |
| 2294 | Poll store instead of waiting for completion. Usually this provides better |
| 2295 | throughput at cost of higher(up to 100%) CPU utilization. |
| 2296 | .TP |
| 2297 | .BI (rados)touch_objects \fR=\fPbool |
| 2298 | During initialization, touch (create if do not exist) all objects (files). |
| 2299 | Touching all objects affects ceph caches and likely impacts test results. |
| 2300 | Enabled by default. |
| 2301 | .TP |
| 2302 | .BI (http)http_host \fR=\fPstr |
| 2303 | Hostname to connect to. For S3, this could be the bucket name. Default |
| 2304 | is \fBlocalhost\fR |
| 2305 | .TP |
| 2306 | .BI (http)http_user \fR=\fPstr |
| 2307 | Username for HTTP authentication. |
| 2308 | .TP |
| 2309 | .BI (http)http_pass \fR=\fPstr |
| 2310 | Password for HTTP authentication. |
| 2311 | .TP |
| 2312 | .BI (http)https \fR=\fPstr |
| 2313 | Whether to use HTTPS instead of plain HTTP. \fRon\fP enables HTTPS; |
| 2314 | \fRinsecure\fP will enable HTTPS, but disable SSL peer verification (use |
| 2315 | with caution!). Default is \fBoff\fR. |
| 2316 | .TP |
| 2317 | .BI (http)http_mode \fR=\fPstr |
| 2318 | Which HTTP access mode to use: webdav, swift, or s3. Default is |
| 2319 | \fBwebdav\fR. |
| 2320 | .TP |
| 2321 | .BI (http)http_s3_region \fR=\fPstr |
| 2322 | The S3 region/zone to include in the request. Default is \fBus-east-1\fR. |
| 2323 | .TP |
| 2324 | .BI (http)http_s3_key \fR=\fPstr |
| 2325 | The S3 secret key. |
| 2326 | .TP |
| 2327 | .BI (http)http_s3_keyid \fR=\fPstr |
| 2328 | The S3 key/access id. |
| 2329 | .TP |
| 2330 | .BI (http)http_s3_sse_customer_key \fR=\fPstr |
| 2331 | The encryption customer key in SSE server side. |
| 2332 | .TP |
| 2333 | .BI (http)http_s3_sse_customer_algorithm \fR=\fPstr |
| 2334 | The encryption customer algorithm in SSE server side. Default is \fBAES256\fR |
| 2335 | .TP |
| 2336 | .BI (http)http_s3_storage_class \fR=\fPstr |
| 2337 | Which storage class to access. User-customizable settings. Default is \fBSTANDARD\fR |
| 2338 | .TP |
| 2339 | .BI (http)http_swift_auth_token \fR=\fPstr |
| 2340 | The Swift auth token. See the example configuration file on how to |
| 2341 | retrieve this. |
| 2342 | .TP |
| 2343 | .BI (http)http_verbose \fR=\fPint |
| 2344 | Enable verbose requests from libcurl. Useful for debugging. 1 turns on |
| 2345 | verbose logging from libcurl, 2 additionally enables HTTP IO tracing. |
| 2346 | Default is \fB0\fR |
| 2347 | .TP |
| 2348 | .BI (mtd)skip_bad \fR=\fPbool |
| 2349 | Skip operations against known bad blocks. |
| 2350 | .TP |
| 2351 | .BI (libhdfs)hdfsdirectory |
| 2352 | libhdfs will create chunk in this HDFS directory. |
| 2353 | .TP |
| 2354 | .BI (libhdfs)chunk_size |
| 2355 | The size of the chunk to use for each file. |
| 2356 | .TP |
| 2357 | .BI (rdma)verb \fR=\fPstr |
| 2358 | The RDMA verb to use on this side of the RDMA ioengine |
| 2359 | connection. Valid values are write, read, send and recv. These |
| 2360 | correspond to the equivalent RDMA verbs (e.g. write = rdma_write |
| 2361 | etc.). Note that this only needs to be specified on the client side of |
| 2362 | the connection. See the examples folder. |
| 2363 | .TP |
| 2364 | .BI (rdma)bindname \fR=\fPstr |
| 2365 | The name to use to bind the local RDMA-CM connection to a local RDMA |
| 2366 | device. This could be a hostname or an IPv4 or IPv6 address. On the |
| 2367 | server side this will be passed into the rdma_bind_addr() function and |
| 2368 | on the client site it will be used in the rdma_resolve_add() |
| 2369 | function. This can be useful when multiple paths exist between the |
| 2370 | client and the server or in certain loopback configurations. |
| 2371 | .TP |
| 2372 | .BI (filestat)stat_type \fR=\fPstr |
| 2373 | Specify stat system call type to measure lookup/getattr performance. |
| 2374 | Default is \fBstat\fR for \fBstat\fR\|(2). |
| 2375 | .TP |
| 2376 | .BI (sg)hipri |
| 2377 | If this option is set, fio will attempt to use polled IO completions. This |
| 2378 | will have a similar effect as (io_uring)hipri. Only SCSI READ and WRITE |
| 2379 | commands will have the SGV4_FLAG_HIPRI set (not UNMAP (trim) nor VERIFY). |
| 2380 | Older versions of the Linux sg driver that do not support hipri will simply |
| 2381 | ignore this flag and do normal IO. The Linux SCSI Low Level Driver (LLD) |
| 2382 | that "owns" the device also needs to support hipri (also known as iopoll |
| 2383 | and mq_poll). The MegaRAID driver is an example of a SCSI LLD. |
| 2384 | Default: clear (0) which does normal (interrupted based) IO. |
| 2385 | .TP |
| 2386 | .BI (sg)readfua \fR=\fPbool |
| 2387 | With readfua option set to 1, read operations include the force |
| 2388 | unit access (fua) flag. Default: 0. |
| 2389 | .TP |
| 2390 | .BI (sg)writefua \fR=\fPbool |
| 2391 | With writefua option set to 1, write operations include the force |
| 2392 | unit access (fua) flag. Default: 0. |
| 2393 | .TP |
| 2394 | .BI (sg)sg_write_mode \fR=\fPstr |
| 2395 | Specify the type of write commands to issue. This option can take multiple |
| 2396 | values: |
| 2397 | .RS |
| 2398 | .RS |
| 2399 | .TP |
| 2400 | .B write (default) |
| 2401 | Write opcodes are issued as usual |
| 2402 | .TP |
| 2403 | .B write_and_verify |
| 2404 | Issue WRITE AND VERIFY commands. The BYTCHK bit is set to 00b. This directs the |
| 2405 | device to carry out a medium verification with no data comparison for the data |
| 2406 | that was written. The writefua option is ignored with this selection. |
| 2407 | .TP |
| 2408 | .B verify |
| 2409 | This option is deprecated. Use write_and_verify instead. |
| 2410 | .TP |
| 2411 | .B write_same |
| 2412 | Issue WRITE SAME commands. This transfers a single block to the device |
| 2413 | and writes this same block of data to a contiguous sequence of LBAs |
| 2414 | beginning at the specified offset. fio's block size parameter |
| 2415 | specifies the amount of data written with each command. However, the |
| 2416 | amount of data actually transferred to the device is equal to the |
| 2417 | device's block (sector) size. For a device with 512 byte sectors, |
| 2418 | blocksize=8k will write 16 sectors with each command. fio will still |
| 2419 | generate 8k of data for each command butonly the first 512 bytes will |
| 2420 | be used and transferred to the device. The writefua option is ignored |
| 2421 | with this selection. |
| 2422 | .TP |
| 2423 | .B same |
| 2424 | This option is deprecated. Use write_same instead. |
| 2425 | .TP |
| 2426 | .B write_same_ndob |
| 2427 | Issue WRITE SAME(16) commands as above but with the No Data Output |
| 2428 | Buffer (NDOB) bit set. No data will be transferred to the device with |
| 2429 | this bit set. Data written will be a pre-determined pattern such as |
| 2430 | all zeroes. |
| 2431 | .TP |
| 2432 | .B write_stream |
| 2433 | Issue WRITE STREAM(16) commands. Use the stream_id option to specify |
| 2434 | the stream identifier. |
| 2435 | .TP |
| 2436 | .B verify_bytchk_00 |
| 2437 | Issue VERIFY commands with BYTCHK set to 00. This directs the device to carry |
| 2438 | out a medium verification with no data comparison. |
| 2439 | .TP |
| 2440 | .B verify_bytchk_01 |
| 2441 | Issue VERIFY commands with BYTCHK set to 01. This directs the device to |
| 2442 | compare the data on the device with the data transferred to the device. |
| 2443 | .TP |
| 2444 | .B verify_bytchk_11 |
| 2445 | Issue VERIFY commands with BYTCHK set to 11. This transfers a single block to |
| 2446 | the device and compares the contents of this block with the data on the device |
| 2447 | beginning at the specified offset. fio's block size parameter specifies the |
| 2448 | total amount of data compared with this command. However, only one block |
| 2449 | (sector) worth of data is transferred to the device. This is similar to the |
| 2450 | WRITE SAME command except that data is compared instead of written. |
| 2451 | .RE |
| 2452 | .RE |
| 2453 | .TP |
| 2454 | .BI (sg)stream_id \fR=\fPint |
| 2455 | Set the stream identifier for WRITE STREAM commands. If this is set to 0 (which is not |
| 2456 | a valid stream identifier) fio will open a stream and then close it when done. Default |
| 2457 | is 0. |
| 2458 | .TP |
| 2459 | .BI (nbd)uri \fR=\fPstr |
| 2460 | Specify the NBD URI of the server to test. |
| 2461 | The string is a standard NBD URI (see |
| 2462 | \fIhttps://github.com/NetworkBlockDevice/nbd/tree/master/doc\fR). |
| 2463 | Example URIs: |
| 2464 | .RS |
| 2465 | .RS |
| 2466 | .TP |
| 2467 | \fInbd://localhost:10809\fR |
| 2468 | .TP |
| 2469 | \fInbd+unix:///?socket=/tmp/socket\fR |
| 2470 | .TP |
| 2471 | \fInbds://tlshost/exportname\fR |
| 2472 | .RE |
| 2473 | .RE |
| 2474 | .TP |
| 2475 | .BI (libcufile)gpu_dev_ids\fR=\fPstr |
| 2476 | Specify the GPU IDs to use with CUDA. This is a colon-separated list of int. |
| 2477 | GPUs are assigned to workers roundrobin. Default is 0. |
| 2478 | .TP |
| 2479 | .BI (libcufile)cuda_io\fR=\fPstr |
| 2480 | Specify the type of I/O to use with CUDA. This option |
| 2481 | takes the following values: |
| 2482 | .RS |
| 2483 | .RS |
| 2484 | .TP |
| 2485 | .B cufile (default) |
| 2486 | Use libcufile and nvidia-fs. This option performs I/O directly |
| 2487 | between a GPUDirect Storage filesystem and GPU buffers, |
| 2488 | avoiding use of a bounce buffer. If \fBverify\fR is set, |
| 2489 | cudaMemcpy is used to copy verification data between RAM and GPU(s). |
| 2490 | Verification data is copied from RAM to GPU before a write |
| 2491 | and from GPU to RAM after a read. |
| 2492 | \fBdirect\fR must be 1. |
| 2493 | .TP |
| 2494 | .BI posix |
| 2495 | Use POSIX to perform I/O with a RAM buffer, and use |
| 2496 | cudaMemcpy to transfer data between RAM and the GPU(s). |
| 2497 | Data is copied from GPU to RAM before a write and copied |
| 2498 | from RAM to GPU after a read. \fBverify\fR does not affect |
| 2499 | the use of cudaMemcpy. |
| 2500 | .RE |
| 2501 | .RE |
| 2502 | .TP |
| 2503 | .BI (dfs)pool |
| 2504 | Specify the label or UUID of the DAOS pool to connect to. |
| 2505 | .TP |
| 2506 | .BI (dfs)cont |
| 2507 | Specify the label or UUID of the DAOS container to open. |
| 2508 | .TP |
| 2509 | .BI (dfs)chunk_size |
| 2510 | Specify a different chunk size (in bytes) for the dfs file. |
| 2511 | Use DAOS container's chunk size by default. |
| 2512 | .TP |
| 2513 | .BI (dfs)object_class |
| 2514 | Specify a different object class for the dfs file. |
| 2515 | Use DAOS container's object class by default. |
| 2516 | .TP |
| 2517 | .BI (nfs)nfs_url |
| 2518 | URL in libnfs format, eg nfs://<server|ipv4|ipv6>/path[?arg=val[&arg=val]*] |
| 2519 | Refer to the libnfs README for more details. |
| 2520 | .TP |
| 2521 | .BI (exec)program\fR=\fPstr |
| 2522 | Specify the program to execute. |
| 2523 | Note the program will receive a SIGTERM when the job is reaching the time limit. |
| 2524 | A SIGKILL is sent once the job is over. The delay between the two signals is defined by \fBgrace_time\fR option. |
| 2525 | .TP |
| 2526 | .BI (exec)arguments\fR=\fPstr |
| 2527 | Specify arguments to pass to program. |
| 2528 | Some special variables can be expanded to pass fio's job details to the program : |
| 2529 | .RS |
| 2530 | .RS |
| 2531 | .TP |
| 2532 | .B %r |
| 2533 | replaced by the duration of the job in seconds |
| 2534 | .TP |
| 2535 | .BI %n |
| 2536 | replaced by the name of the job |
| 2537 | .RE |
| 2538 | .RE |
| 2539 | .TP |
| 2540 | .BI (exec)grace_time\fR=\fPint |
| 2541 | Defines the time between the SIGTERM and SIGKILL signals. Default is 1 second. |
| 2542 | .TP |
| 2543 | .BI (exec)std_redirect\fR=\fbool |
| 2544 | If set, stdout and stderr streams are redirected to files named from the job name. Default is true. |
| 2545 | .TP |
| 2546 | .BI (xnvme)xnvme_async\fR=\fPstr |
| 2547 | Select the xnvme async command interface. This can take these values. |
| 2548 | .RS |
| 2549 | .RS |
| 2550 | .TP |
| 2551 | .B emu |
| 2552 | This is default and use to emulate asynchronous I/O by using a single thread to |
| 2553 | create a queue pair on top of a synchronous I/O interface using the NVMe driver |
| 2554 | IOCTL. |
| 2555 | .TP |
| 2556 | .BI thrpool |
| 2557 | Emulate an asynchronous I/O interface with a pool of userspace threads on top |
| 2558 | of a synchronous I/O interface using the NVMe driver IOCTL. By default four |
| 2559 | threads are used. |
| 2560 | .TP |
| 2561 | .BI io_uring |
| 2562 | Linux native asynchronous I/O interface which supports both direct and buffered |
| 2563 | I/O. |
| 2564 | .TP |
| 2565 | .BI libaio |
| 2566 | Use Linux aio for Asynchronous I/O |
| 2567 | .TP |
| 2568 | .BI posix |
| 2569 | Use the posix asynchronous I/O interface to perform one or more I/O operations |
| 2570 | asynchronously. |
| 2571 | .TP |
| 2572 | .BI nil |
| 2573 | Do not transfer any data; just pretend to. This is mainly used for |
| 2574 | introspective performance evaluation. |
| 2575 | .RE |
| 2576 | .RE |
| 2577 | .TP |
| 2578 | .BI (xnvme)xnvme_sync\fR=\fPstr |
| 2579 | Select the xnvme synchronous command interface. This can take these values. |
| 2580 | .RS |
| 2581 | .RS |
| 2582 | .TP |
| 2583 | .B nvme |
| 2584 | This is default and uses Linux NVMe Driver ioctl() for synchronous I/O. |
| 2585 | .TP |
| 2586 | .BI psync |
| 2587 | This supports regular as well as vectored pread() and pwrite() commands. |
| 2588 | .TP |
| 2589 | .BI block |
| 2590 | This is the same as psync except that it also supports zone management |
| 2591 | commands using Linux block layer IOCTLs. |
| 2592 | .RE |
| 2593 | .RE |
| 2594 | .TP |
| 2595 | .BI (xnvme)xnvme_admin\fR=\fPstr |
| 2596 | Select the xnvme admin command interface. This can take these values. |
| 2597 | .RS |
| 2598 | .RS |
| 2599 | .TP |
| 2600 | .B nvme |
| 2601 | This is default and uses Linux NVMe Driver ioctl() for admin commands. |
| 2602 | .TP |
| 2603 | .BI block |
| 2604 | Use Linux Block Layer ioctl() and sysfs for admin commands. |
| 2605 | .RE |
| 2606 | .RE |
| 2607 | .TP |
| 2608 | .BI (xnvme)xnvme_dev_nsid\fR=\fPint |
| 2609 | xnvme namespace identifier for userspace NVMe driver such as SPDK. |
| 2610 | .TP |
| 2611 | .BI (xnvme)xnvme_iovec |
| 2612 | If this option is set, xnvme will use vectored read/write commands. |
| 2613 | .TP |
| 2614 | .BI (libblkio)libblkio_driver \fR=\fPstr |
| 2615 | The libblkio driver to use. Different drivers access devices through different |
| 2616 | underlying interfaces. Available drivers depend on the libblkio version in use |
| 2617 | and are listed at \fIhttps://libblkio.gitlab.io/libblkio/blkio.html#drivers\fR |
| 2618 | .TP |
| 2619 | .BI (libblkio)libblkio_pre_connect_props \fR=\fPstr |
| 2620 | A colon-separated list of libblkio properties to be set after creating but |
| 2621 | before connecting the libblkio instance. Each property must have the format |
| 2622 | \fB<name>=<value>\fR. Colons can be escaped as \fB\\:\fR. These are set after |
| 2623 | the engine sets any other properties, so those can be overriden. Available |
| 2624 | properties depend on the libblkio version in use and are listed at |
| 2625 | \fIhttps://libblkio.gitlab.io/libblkio/blkio.html#properties\fR |
| 2626 | .TP |
| 2627 | .BI (libblkio)libblkio_pre_start_props \fR=\fPstr |
| 2628 | A colon-separated list of libblkio properties to be set after connecting but |
| 2629 | before starting the libblkio instance. Each property must have the format |
| 2630 | \fB<name>=<value>\fR. Colons can be escaped as \fB\\:\fR. These are set after |
| 2631 | the engine sets any other properties, so those can be overriden. Available |
| 2632 | properties depend on the libblkio version in use and are listed at |
| 2633 | \fIhttps://libblkio.gitlab.io/libblkio/blkio.html#properties\fR |
| 2634 | .TP |
| 2635 | .BI (libblkio)hipri |
| 2636 | Use poll queues. This is incompatible with \fBlibblkio_wait_mode=eventfd\fR and |
| 2637 | \fBlibblkio_force_enable_completion_eventfd\fR. |
| 2638 | .TP |
| 2639 | .BI (libblkio)libblkio_vectored |
| 2640 | Submit vectored read and write requests. |
| 2641 | .TP |
| 2642 | .BI (libblkio)libblkio_write_zeroes_on_trim |
| 2643 | Submit trims as "write zeroes" requests instead of discard requests. |
| 2644 | .TP |
| 2645 | .BI (libblkio)libblkio_wait_mode \fR=\fPstr |
| 2646 | How to wait for completions: |
| 2647 | .RS |
| 2648 | .RS |
| 2649 | .TP |
| 2650 | .B block \fR(default) |
| 2651 | Use a blocking call to \fBblkioq_do_io()\fR. |
| 2652 | .TP |
| 2653 | .B eventfd |
| 2654 | Use a blocking call to \fBread()\fR on the completion eventfd. |
| 2655 | .TP |
| 2656 | .B loop |
| 2657 | Use a busy loop with a non-blocking call to \fBblkioq_do_io()\fR. |
| 2658 | .RE |
| 2659 | .RE |
| 2660 | .TP |
| 2661 | .BI (libblkio)libblkio_force_enable_completion_eventfd |
| 2662 | Enable the queue's completion eventfd even when unused. This may impact |
| 2663 | performance. The default is to enable it only if |
| 2664 | \fBlibblkio_wait_mode=eventfd\fR. |
| 2665 | .SS "I/O depth" |
| 2666 | .TP |
| 2667 | .BI iodepth \fR=\fPint |
| 2668 | Number of I/O units to keep in flight against the file. Note that |
| 2669 | increasing \fBiodepth\fR beyond 1 will not affect synchronous ioengines (except |
| 2670 | for small degrees when \fBverify_async\fR is in use). Even async |
| 2671 | engines may impose OS restrictions causing the desired depth not to be |
| 2672 | achieved. This may happen on Linux when using libaio and not setting |
| 2673 | `direct=1', since buffered I/O is not async on that OS. Keep an |
| 2674 | eye on the I/O depth distribution in the fio output to verify that the |
| 2675 | achieved depth is as expected. Default: 1. |
| 2676 | .TP |
| 2677 | .BI iodepth_batch_submit \fR=\fPint "\fR,\fP iodepth_batch" \fR=\fPint |
| 2678 | This defines how many pieces of I/O to submit at once. It defaults to 1 |
| 2679 | which means that we submit each I/O as soon as it is available, but can be |
| 2680 | raised to submit bigger batches of I/O at the time. If it is set to 0 the |
| 2681 | \fBiodepth\fR value will be used. |
| 2682 | .TP |
| 2683 | .BI iodepth_batch_complete_min \fR=\fPint "\fR,\fP iodepth_batch_complete" \fR=\fPint |
| 2684 | This defines how many pieces of I/O to retrieve at once. It defaults to 1 |
| 2685 | which means that we'll ask for a minimum of 1 I/O in the retrieval process |
| 2686 | from the kernel. The I/O retrieval will go on until we hit the limit set by |
| 2687 | \fBiodepth_low\fR. If this variable is set to 0, then fio will always |
| 2688 | check for completed events before queuing more I/O. This helps reduce I/O |
| 2689 | latency, at the cost of more retrieval system calls. |
| 2690 | .TP |
| 2691 | .BI iodepth_batch_complete_max \fR=\fPint |
| 2692 | This defines maximum pieces of I/O to retrieve at once. This variable should |
| 2693 | be used along with \fBiodepth_batch_complete_min\fR=\fIint\fR variable, |
| 2694 | specifying the range of min and max amount of I/O which should be |
| 2695 | retrieved. By default it is equal to \fBiodepth_batch_complete_min\fR |
| 2696 | value. Example #1: |
| 2697 | .RS |
| 2698 | .RS |
| 2699 | .P |
| 2700 | .PD 0 |
| 2701 | iodepth_batch_complete_min=1 |
| 2702 | .P |
| 2703 | iodepth_batch_complete_max=<iodepth> |
| 2704 | .PD |
| 2705 | .RE |
| 2706 | .P |
| 2707 | which means that we will retrieve at least 1 I/O and up to the whole |
| 2708 | submitted queue depth. If none of I/O has been completed yet, we will wait. |
| 2709 | Example #2: |
| 2710 | .RS |
| 2711 | .P |
| 2712 | .PD 0 |
| 2713 | iodepth_batch_complete_min=0 |
| 2714 | .P |
| 2715 | iodepth_batch_complete_max=<iodepth> |
| 2716 | .PD |
| 2717 | .RE |
| 2718 | .P |
| 2719 | which means that we can retrieve up to the whole submitted queue depth, but |
| 2720 | if none of I/O has been completed yet, we will NOT wait and immediately exit |
| 2721 | the system call. In this example we simply do polling. |
| 2722 | .RE |
| 2723 | .TP |
| 2724 | .BI iodepth_low \fR=\fPint |
| 2725 | The low water mark indicating when to start filling the queue |
| 2726 | again. Defaults to the same as \fBiodepth\fR, meaning that fio will |
| 2727 | attempt to keep the queue full at all times. If \fBiodepth\fR is set to |
| 2728 | e.g. 16 and \fBiodepth_low\fR is set to 4, then after fio has filled the queue of |
| 2729 | 16 requests, it will let the depth drain down to 4 before starting to fill |
| 2730 | it again. |
| 2731 | .TP |
| 2732 | .BI serialize_overlap \fR=\fPbool |
| 2733 | Serialize in-flight I/Os that might otherwise cause or suffer from data races. |
| 2734 | When two or more I/Os are submitted simultaneously, there is no guarantee that |
| 2735 | the I/Os will be processed or completed in the submitted order. Further, if |
| 2736 | two or more of those I/Os are writes, any overlapping region between them can |
| 2737 | become indeterminate/undefined on certain storage. These issues can cause |
| 2738 | verification to fail erratically when at least one of the racing I/Os is |
| 2739 | changing data and the overlapping region has a non-zero size. Setting |
| 2740 | \fBserialize_overlap\fR tells fio to avoid provoking this behavior by explicitly |
| 2741 | serializing in-flight I/Os that have a non-zero overlap. Note that setting |
| 2742 | this option can reduce both performance and the \fBiodepth\fR achieved. |
| 2743 | .RS |
| 2744 | .P |
| 2745 | This option only applies to I/Os issued for a single job except when it is |
| 2746 | enabled along with \fBio_submit_mode\fR=offload. In offload mode, fio |
| 2747 | will check for overlap among all I/Os submitted by offload jobs with \fBserialize_overlap\fR |
| 2748 | enabled. |
| 2749 | .P |
| 2750 | Default: false. |
| 2751 | .RE |
| 2752 | .TP |
| 2753 | .BI io_submit_mode \fR=\fPstr |
| 2754 | This option controls how fio submits the I/O to the I/O engine. The default |
| 2755 | is `inline', which means that the fio job threads submit and reap I/O |
| 2756 | directly. If set to `offload', the job threads will offload I/O submission |
| 2757 | to a dedicated pool of I/O threads. This requires some coordination and thus |
| 2758 | has a bit of extra overhead, especially for lower queue depth I/O where it |
| 2759 | can increase latencies. The benefit is that fio can manage submission rates |
| 2760 | independently of the device completion rates. This avoids skewed latency |
| 2761 | reporting if I/O gets backed up on the device side (the coordinated omission |
| 2762 | problem). Note that this option cannot reliably be used with async IO engines. |
| 2763 | .SS "I/O rate" |
| 2764 | .TP |
| 2765 | .BI thinktime \fR=\fPtime |
| 2766 | Stall the job for the specified period of time after an I/O has completed before issuing the |
| 2767 | next. May be used to simulate processing being done by an application. |
| 2768 | When the unit is omitted, the value is interpreted in microseconds. See |
| 2769 | \fBthinktime_blocks\fR, \fBthinktime_iotime\fR and \fBthinktime_spin\fR. |
| 2770 | .TP |
| 2771 | .BI thinktime_spin \fR=\fPtime |
| 2772 | Only valid if \fBthinktime\fR is set - pretend to spend CPU time doing |
| 2773 | something with the data received, before falling back to sleeping for the |
| 2774 | rest of the period specified by \fBthinktime\fR. When the unit is |
| 2775 | omitted, the value is interpreted in microseconds. |
| 2776 | .TP |
| 2777 | .BI thinktime_blocks \fR=\fPint |
| 2778 | Only valid if \fBthinktime\fR is set - control how many blocks to issue, |
| 2779 | before waiting \fBthinktime\fR usecs. If not set, defaults to 1 which will make |
| 2780 | fio wait \fBthinktime\fR usecs after every block. This effectively makes any |
| 2781 | queue depth setting redundant, since no more than 1 I/O will be queued |
| 2782 | before we have to complete it and do our \fBthinktime\fR. In other words, this |
| 2783 | setting effectively caps the queue depth if the latter is larger. |
| 2784 | .TP |
| 2785 | .BI thinktime_blocks_type \fR=\fPstr |
| 2786 | Only valid if \fBthinktime\fR is set - control how \fBthinktime_blocks\fR triggers. |
| 2787 | The default is `complete', which triggers \fBthinktime\fR when fio completes |
| 2788 | \fBthinktime_blocks\fR blocks. If this is set to `issue', then the trigger happens |
| 2789 | at the issue side. |
| 2790 | .TP |
| 2791 | .BI thinktime_iotime \fR=\fPtime |
| 2792 | Only valid if \fBthinktime\fR is set - control \fBthinktime\fR interval by time. |
| 2793 | The \fBthinktime\fR stall is repeated after IOs are executed for |
| 2794 | \fBthinktime_iotime\fR. For example, `\-\-thinktime_iotime=9s \-\-thinktime=1s' |
| 2795 | repeat 10-second cycle with IOs for 9 seconds and stall for 1 second. When the |
| 2796 | unit is omitted, \fBthinktime_iotime\fR is interpreted as a number of seconds. |
| 2797 | If this option is used together with \fBthinktime_blocks\fR, the \fBthinktime\fR |
| 2798 | stall is repeated after \fBthinktime_iotime\fR or after \fBthinktime_blocks\fR |
| 2799 | IOs, whichever happens first. |
| 2800 | |
| 2801 | .TP |
| 2802 | .BI rate \fR=\fPint[,int][,int] |
| 2803 | Cap the bandwidth used by this job. The number is in bytes/sec, the normal |
| 2804 | suffix rules apply. Comma-separated values may be specified for reads, |
| 2805 | writes, and trims as described in \fBblocksize\fR. |
| 2806 | .RS |
| 2807 | .P |
| 2808 | For example, using `rate=1m,500k' would limit reads to 1MiB/sec and writes to |
| 2809 | 500KiB/sec. Capping only reads or writes can be done with `rate=,500k' or |
| 2810 | `rate=500k,' where the former will only limit writes (to 500KiB/sec) and the |
| 2811 | latter will only limit reads. |
| 2812 | .RE |
| 2813 | .TP |
| 2814 | .BI rate_min \fR=\fPint[,int][,int] |
| 2815 | Tell fio to do whatever it can to maintain at least this bandwidth. Failing |
| 2816 | to meet this requirement will cause the job to exit. Comma-separated values |
| 2817 | may be specified for reads, writes, and trims as described in |
| 2818 | \fBblocksize\fR. |
| 2819 | .TP |
| 2820 | .BI rate_iops \fR=\fPint[,int][,int] |
| 2821 | Cap the bandwidth to this number of IOPS. Basically the same as |
| 2822 | \fBrate\fR, just specified independently of bandwidth. If the job is |
| 2823 | given a block size range instead of a fixed value, the smallest block size |
| 2824 | is used as the metric. Comma-separated values may be specified for reads, |
| 2825 | writes, and trims as described in \fBblocksize\fR. |
| 2826 | .TP |
| 2827 | .BI rate_iops_min \fR=\fPint[,int][,int] |
| 2828 | If fio doesn't meet this rate of I/O, it will cause the job to exit. |
| 2829 | Comma-separated values may be specified for reads, writes, and trims as |
| 2830 | described in \fBblocksize\fR. |
| 2831 | .TP |
| 2832 | .BI rate_process \fR=\fPstr |
| 2833 | This option controls how fio manages rated I/O submissions. The default is |
| 2834 | `linear', which submits I/O in a linear fashion with fixed delays between |
| 2835 | I/Os that gets adjusted based on I/O completion rates. If this is set to |
| 2836 | `poisson', fio will submit I/O based on a more real world random request |
| 2837 | flow, known as the Poisson process |
| 2838 | (\fIhttps://en.wikipedia.org/wiki/Poisson_point_process\fR). The lambda will be |
| 2839 | 10^6 / IOPS for the given workload. |
| 2840 | .TP |
| 2841 | .BI rate_ignore_thinktime \fR=\fPbool |
| 2842 | By default, fio will attempt to catch up to the specified rate setting, if any |
| 2843 | kind of thinktime setting was used. If this option is set, then fio will |
| 2844 | ignore the thinktime and continue doing IO at the specified rate, instead of |
| 2845 | entering a catch-up mode after thinktime is done. |
| 2846 | .SS "I/O latency" |
| 2847 | .TP |
| 2848 | .BI latency_target \fR=\fPtime |
| 2849 | If set, fio will attempt to find the max performance point that the given |
| 2850 | workload will run at while maintaining a latency below this target. When |
| 2851 | the unit is omitted, the value is interpreted in microseconds. See |
| 2852 | \fBlatency_window\fR and \fBlatency_percentile\fR. |
| 2853 | .TP |
| 2854 | .BI latency_window \fR=\fPtime |
| 2855 | Used with \fBlatency_target\fR to specify the sample window that the job |
| 2856 | is run at varying queue depths to test the performance. When the unit is |
| 2857 | omitted, the value is interpreted in microseconds. |
| 2858 | .TP |
| 2859 | .BI latency_percentile \fR=\fPfloat |
| 2860 | The percentage of I/Os that must fall within the criteria specified by |
| 2861 | \fBlatency_target\fR and \fBlatency_window\fR. If not set, this |
| 2862 | defaults to 100.0, meaning that all I/Os must be equal or below to the value |
| 2863 | set by \fBlatency_target\fR. |
| 2864 | .TP |
| 2865 | .BI latency_run \fR=\fPbool |
| 2866 | Used with \fBlatency_target\fR. If false (default), fio will find the highest |
| 2867 | queue depth that meets \fBlatency_target\fR and exit. If true, fio will continue |
| 2868 | running and try to meet \fBlatency_target\fR by adjusting queue depth. |
| 2869 | .TP |
| 2870 | .BI max_latency \fR=\fPtime[,time][,time] |
| 2871 | If set, fio will exit the job with an ETIMEDOUT error if it exceeds this |
| 2872 | maximum latency. When the unit is omitted, the value is interpreted in |
| 2873 | microseconds. Comma-separated values may be specified for reads, writes, |
| 2874 | and trims as described in \fBblocksize\fR. |
| 2875 | .TP |
| 2876 | .BI rate_cycle \fR=\fPint |
| 2877 | Average bandwidth for \fBrate\fR and \fBrate_min\fR over this number |
| 2878 | of milliseconds. Defaults to 1000. |
| 2879 | .SS "I/O replay" |
| 2880 | .TP |
| 2881 | .BI write_iolog \fR=\fPstr |
| 2882 | Write the issued I/O patterns to the specified file. See |
| 2883 | \fBread_iolog\fR. Specify a separate file for each job, otherwise the |
| 2884 | iologs will be interspersed and the file may be corrupt. This file will be |
| 2885 | opened in append mode. |
| 2886 | .TP |
| 2887 | .BI read_iolog \fR=\fPstr |
| 2888 | Open an iolog with the specified filename and replay the I/O patterns it |
| 2889 | contains. This can be used to store a workload and replay it sometime |
| 2890 | later. The iolog given may also be a blktrace binary file, which allows fio |
| 2891 | to replay a workload captured by blktrace. See |
| 2892 | \fBblktrace\fR\|(8) for how to capture such logging data. For blktrace |
| 2893 | replay, the file needs to be turned into a blkparse binary data file first |
| 2894 | (`blkparse <device> \-o /dev/null \-d file_for_fio.bin'). |
| 2895 | You can specify a number of files by separating the names with a ':' character. |
| 2896 | See the \fBfilename\fR option for information on how to escape ':' |
| 2897 | characters within the file names. These files will be sequentially assigned to |
| 2898 | job clones created by \fBnumjobs\fR. '-' is a reserved name, meaning read from |
| 2899 | stdin, notably if \fBfilename\fR is set to '-' which means stdin as well, |
| 2900 | then this flag can't be set to '-'. |
| 2901 | .TP |
| 2902 | .BI read_iolog_chunked \fR=\fPbool |
| 2903 | Determines how iolog is read. If false (default) entire \fBread_iolog\fR will |
| 2904 | be read at once. If selected true, input from iolog will be read gradually. |
| 2905 | Useful when iolog is very large, or it is generated. |
| 2906 | .TP |
| 2907 | .BI merge_blktrace_file \fR=\fPstr |
| 2908 | When specified, rather than replaying the logs passed to \fBread_iolog\fR, |
| 2909 | the logs go through a merge phase which aggregates them into a single blktrace. |
| 2910 | The resulting file is then passed on as the \fBread_iolog\fR parameter. The |
| 2911 | intention here is to make the order of events consistent. This limits the |
| 2912 | influence of the scheduler compared to replaying multiple blktraces via |
| 2913 | concurrent jobs. |
| 2914 | .TP |
| 2915 | .BI merge_blktrace_scalars \fR=\fPfloat_list |
| 2916 | This is a percentage based option that is index paired with the list of files |
| 2917 | passed to \fBread_iolog\fR. When merging is performed, scale the time of each |
| 2918 | event by the corresponding amount. For example, |
| 2919 | `\-\-merge_blktrace_scalars="50:100"' runs the first trace in halftime and the |
| 2920 | second trace in realtime. This knob is separately tunable from |
| 2921 | \fBreplay_time_scale\fR which scales the trace during runtime and will not |
| 2922 | change the output of the merge unlike this option. |
| 2923 | .TP |
| 2924 | .BI merge_blktrace_iters \fR=\fPfloat_list |
| 2925 | This is a whole number option that is index paired with the list of files |
| 2926 | passed to \fBread_iolog\fR. When merging is performed, run each trace for |
| 2927 | the specified number of iterations. For example, |
| 2928 | `\-\-merge_blktrace_iters="2:1"' runs the first trace for two iterations |
| 2929 | and the second trace for one iteration. |
| 2930 | .TP |
| 2931 | .BI replay_no_stall \fR=\fPbool |
| 2932 | When replaying I/O with \fBread_iolog\fR the default behavior is to |
| 2933 | attempt to respect the timestamps within the log and replay them with the |
| 2934 | appropriate delay between IOPS. By setting this variable fio will not |
| 2935 | respect the timestamps and attempt to replay them as fast as possible while |
| 2936 | still respecting ordering. The result is the same I/O pattern to a given |
| 2937 | device, but different timings. |
| 2938 | .TP |
| 2939 | .BI replay_time_scale \fR=\fPint |
| 2940 | When replaying I/O with \fBread_iolog\fR, fio will honor the original timing |
| 2941 | in the trace. With this option, it's possible to scale the time. It's a |
| 2942 | percentage option, if set to 50 it means run at 50% the original IO rate in |
| 2943 | the trace. If set to 200, run at twice the original IO rate. Defaults to 100. |
| 2944 | .TP |
| 2945 | .BI replay_redirect \fR=\fPstr |
| 2946 | While replaying I/O patterns using \fBread_iolog\fR the default behavior |
| 2947 | is to replay the IOPS onto the major/minor device that each IOP was recorded |
| 2948 | from. This is sometimes undesirable because on a different machine those |
| 2949 | major/minor numbers can map to a different device. Changing hardware on the |
| 2950 | same system can also result in a different major/minor mapping. |
| 2951 | \fBreplay_redirect\fR causes all I/Os to be replayed onto the single specified |
| 2952 | device regardless of the device it was recorded |
| 2953 | from. i.e. `replay_redirect=/dev/sdc' would cause all I/O |
| 2954 | in the blktrace or iolog to be replayed onto `/dev/sdc'. This means |
| 2955 | multiple devices will be replayed onto a single device, if the trace |
| 2956 | contains multiple devices. If you want multiple devices to be replayed |
| 2957 | concurrently to multiple redirected devices you must blkparse your trace |
| 2958 | into separate traces and replay them with independent fio invocations. |
| 2959 | Unfortunately this also breaks the strict time ordering between multiple |
| 2960 | device accesses. |
| 2961 | .TP |
| 2962 | .BI replay_align \fR=\fPint |
| 2963 | Force alignment of the byte offsets in a trace to this value. The value |
| 2964 | must be a power of 2. |
| 2965 | .TP |
| 2966 | .BI replay_scale \fR=\fPint |
| 2967 | Scale bye offsets down by this factor when replaying traces. Should most |
| 2968 | likely use \fBreplay_align\fR as well. |
| 2969 | .SS "Threads, processes and job synchronization" |
| 2970 | .TP |
| 2971 | .BI replay_skip \fR=\fPstr |
| 2972 | Sometimes it's useful to skip certain IO types in a replay trace. This could |
| 2973 | be, for instance, eliminating the writes in the trace. Or not replaying the |
| 2974 | trims/discards, if you are redirecting to a device that doesn't support them. |
| 2975 | This option takes a comma separated list of read, write, trim, sync. |
| 2976 | .TP |
| 2977 | .BI thread |
| 2978 | Fio defaults to creating jobs by using fork, however if this option is |
| 2979 | given, fio will create jobs by using POSIX Threads' function |
| 2980 | \fBpthread_create\fR\|(3) to create threads instead. |
| 2981 | .TP |
| 2982 | .BI wait_for \fR=\fPstr |
| 2983 | If set, the current job won't be started until all workers of the specified |
| 2984 | waitee job are done. |
| 2985 | .\" ignore blank line here from HOWTO as it looks normal without it |
| 2986 | \fBwait_for\fR operates on the job name basis, so there are a few |
| 2987 | limitations. First, the waitee must be defined prior to the waiter job |
| 2988 | (meaning no forward references). Second, if a job is being referenced as a |
| 2989 | waitee, it must have a unique name (no duplicate waitees). |
| 2990 | .TP |
| 2991 | .BI nice \fR=\fPint |
| 2992 | Run the job with the given nice value. See man \fBnice\fR\|(2). |
| 2993 | .\" ignore blank line here from HOWTO as it looks normal without it |
| 2994 | On Windows, values less than \-15 set the process class to "High"; \-1 through |
| 2995 | \-15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle" |
| 2996 | priority class. |
| 2997 | .TP |
| 2998 | .BI prio \fR=\fPint |
| 2999 | Set the I/O priority value of this job. Linux limits us to a positive value |
| 3000 | between 0 and 7, with 0 being the highest. See man |
| 3001 | \fBionice\fR\|(1). Refer to an appropriate manpage for other operating |
| 3002 | systems since meaning of priority may differ. For per-command priority |
| 3003 | setting, see the I/O engine specific `cmdprio_percentage` and |
| 3004 | `cmdprio` options. |
| 3005 | .TP |
| 3006 | .BI prioclass \fR=\fPint |
| 3007 | Set the I/O priority class. See man \fBionice\fR\|(1). For per-command |
| 3008 | priority setting, see the I/O engine specific `cmdprio_percentage` and |
| 3009 | `cmdprio_class` options. |
| 3010 | .TP |
| 3011 | .BI cpus_allowed \fR=\fPstr |
| 3012 | Controls the same options as \fBcpumask\fR, but accepts a textual |
| 3013 | specification of the permitted CPUs instead and CPUs are indexed from 0. So |
| 3014 | to use CPUs 0 and 5 you would specify `cpus_allowed=0,5'. This option also |
| 3015 | allows a range of CPUs to be specified \-\- say you wanted a binding to CPUs |
| 3016 | 0, 5, and 8 to 15, you would set `cpus_allowed=0,5,8\-15'. |
| 3017 | .RS |
| 3018 | .P |
| 3019 | On Windows, when `cpus_allowed' is unset only CPUs from fio's current |
| 3020 | processor group will be used and affinity settings are inherited from the |
| 3021 | system. An fio build configured to target Windows 7 makes options that set |
| 3022 | CPUs processor group aware and values will set both the processor group |
| 3023 | and a CPU from within that group. For example, on a system where processor |
| 3024 | group 0 has 40 CPUs and processor group 1 has 32 CPUs, `cpus_allowed' |
| 3025 | values between 0 and 39 will bind CPUs from processor group 0 and |
| 3026 | `cpus_allowed' values between 40 and 71 will bind CPUs from processor |
| 3027 | group 1. When using `cpus_allowed_policy=shared' all CPUs specified by a |
| 3028 | single `cpus_allowed' option must be from the same processor group. For |
| 3029 | Windows fio builds not built for Windows 7, CPUs will only be selected from |
| 3030 | (and be relative to) whatever processor group fio happens to be running in |
| 3031 | and CPUs from other processor groups cannot be used. |
| 3032 | .RE |
| 3033 | .TP |
| 3034 | .BI cpus_allowed_policy \fR=\fPstr |
| 3035 | Set the policy of how fio distributes the CPUs specified by |
| 3036 | \fBcpus_allowed\fR or \fBcpumask\fR. Two policies are supported: |
| 3037 | .RS |
| 3038 | .RS |
| 3039 | .TP |
| 3040 | .B shared |
| 3041 | All jobs will share the CPU set specified. |
| 3042 | .TP |
| 3043 | .B split |
| 3044 | Each job will get a unique CPU from the CPU set. |
| 3045 | .RE |
| 3046 | .P |
| 3047 | \fBshared\fR is the default behavior, if the option isn't specified. If |
| 3048 | \fBsplit\fR is specified, then fio will assign one cpu per job. If not |
| 3049 | enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs |
| 3050 | in the set. |
| 3051 | .RE |
| 3052 | .TP |
| 3053 | .BI cpumask \fR=\fPint |
| 3054 | Set the CPU affinity of this job. The parameter given is a bit mask of |
| 3055 | allowed CPUs the job may run on. So if you want the allowed CPUs to be 1 |
| 3056 | and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man |
| 3057 | \fBsched_setaffinity\fR\|(2). This may not work on all supported |
| 3058 | operating systems or kernel versions. This option doesn't work well for a |
| 3059 | higher CPU count than what you can store in an integer mask, so it can only |
| 3060 | control cpus 1\-32. For boxes with larger CPU counts, use |
| 3061 | \fBcpus_allowed\fR. |
| 3062 | .TP |
| 3063 | .BI numa_cpu_nodes \fR=\fPstr |
| 3064 | Set this job running on specified NUMA nodes' CPUs. The arguments allow |
| 3065 | comma delimited list of cpu numbers, A\-B ranges, or `all'. Note, to enable |
| 3066 | NUMA options support, fio must be built on a system with libnuma\-dev(el) |
| 3067 | installed. |
| 3068 | .TP |
| 3069 | .BI numa_mem_policy \fR=\fPstr |
| 3070 | Set this job's memory policy and corresponding NUMA nodes. Format of the |
| 3071 | arguments: |
| 3072 | .RS |
| 3073 | .RS |
| 3074 | .P |
| 3075 | <mode>[:<nodelist>] |
| 3076 | .RE |
| 3077 | .P |
| 3078 | `mode' is one of the following memory policies: `default', `prefer', |
| 3079 | `bind', `interleave' or `local'. For `default' and `local' memory |
| 3080 | policies, no node needs to be specified. For `prefer', only one node is |
| 3081 | allowed. For `bind' and `interleave' the `nodelist' may be as |
| 3082 | follows: a comma delimited list of numbers, A\-B ranges, or `all'. |
| 3083 | .RE |
| 3084 | .TP |
| 3085 | .BI cgroup \fR=\fPstr |
| 3086 | Add job to this control group. If it doesn't exist, it will be created. The |
| 3087 | system must have a mounted cgroup blkio mount point for this to work. If |
| 3088 | your system doesn't have it mounted, you can do so with: |
| 3089 | .RS |
| 3090 | .RS |
| 3091 | .P |
| 3092 | # mount \-t cgroup \-o blkio none /cgroup |
| 3093 | .RE |
| 3094 | .RE |
| 3095 | .TP |
| 3096 | .BI cgroup_weight \fR=\fPint |
| 3097 | Set the weight of the cgroup to this value. See the documentation that comes |
| 3098 | with the kernel, allowed values are in the range of 100..1000. |
| 3099 | .TP |
| 3100 | .BI cgroup_nodelete \fR=\fPbool |
| 3101 | Normally fio will delete the cgroups it has created after the job |
| 3102 | completion. To override this behavior and to leave cgroups around after the |
| 3103 | job completion, set `cgroup_nodelete=1'. This can be useful if one wants |
| 3104 | to inspect various cgroup files after job completion. Default: false. |
| 3105 | .TP |
| 3106 | .BI flow_id \fR=\fPint |
| 3107 | The ID of the flow. If not specified, it defaults to being a global |
| 3108 | flow. See \fBflow\fR. |
| 3109 | .TP |
| 3110 | .BI flow \fR=\fPint |
| 3111 | Weight in token-based flow control. If this value is used, |
| 3112 | then fio regulates the activity between two or more jobs |
| 3113 | sharing the same flow_id. |
| 3114 | Fio attempts to keep each job activity proportional to other jobs' activities |
| 3115 | in the same flow_id group, with respect to requested weight per job. |
| 3116 | That is, if one job has `flow=3', another job has `flow=2' |
| 3117 | and another with `flow=1`, then there will be a roughly 3:2:1 ratio |
| 3118 | in how much one runs vs the others. |
| 3119 | .TP |
| 3120 | .BI flow_sleep \fR=\fPint |
| 3121 | The period of time, in microseconds, to wait after the flow counter |
| 3122 | has exceeded its proportion before retrying operations. |
| 3123 | .TP |
| 3124 | .BI stonewall "\fR,\fB wait_for_previous" |
| 3125 | Wait for preceding jobs in the job file to exit, before starting this |
| 3126 | one. Can be used to insert serialization points in the job file. A stone |
| 3127 | wall also implies starting a new reporting group, see |
| 3128 | \fBgroup_reporting\fR. Optionally you can use `stonewall=0` to disable or |
| 3129 | `stonewall=1` to enable it. |
| 3130 | .TP |
| 3131 | .BI exitall |
| 3132 | By default, fio will continue running all other jobs when one job finishes. |
| 3133 | Sometimes this is not the desired action. Setting \fBexitall\fR will instead |
| 3134 | make fio terminate all jobs in the same group, as soon as one job of that |
| 3135 | group finishes. |
| 3136 | .TP |
| 3137 | .BI exit_what \fR=\fPstr |
| 3138 | By default, fio will continue running all other jobs when one job finishes. |
| 3139 | Sometimes this is not the desired action. Setting \fBexitall\fR will instead |
| 3140 | make fio terminate all jobs in the same group. The option \fBexit_what\fR |
| 3141 | allows you to control which jobs get terminated when \fBexitall\fR is enabled. |
| 3142 | The default value is \fBgroup\fR. |
| 3143 | The allowed values are: |
| 3144 | .RS |
| 3145 | .RS |
| 3146 | .TP |
| 3147 | .B all |
| 3148 | terminates all jobs. |
| 3149 | .TP |
| 3150 | .B group |
| 3151 | is the default and does not change the behaviour of \fBexitall\fR. |
| 3152 | .TP |
| 3153 | .B stonewall |
| 3154 | terminates all currently running jobs across all groups and continues |
| 3155 | execution with the next stonewalled group. |
| 3156 | .RE |
| 3157 | .RE |
| 3158 | .TP |
| 3159 | .BI exec_prerun \fR=\fPstr |
| 3160 | Before running this job, issue the command specified through |
| 3161 | \fBsystem\fR\|(3). Output is redirected in a file called `jobname.prerun.txt'. |
| 3162 | .TP |
| 3163 | .BI exec_postrun \fR=\fPstr |
| 3164 | After the job completes, issue the command specified though |
| 3165 | \fBsystem\fR\|(3). Output is redirected in a file called `jobname.postrun.txt'. |
| 3166 | .TP |
| 3167 | .BI uid \fR=\fPint |
| 3168 | Instead of running as the invoking user, set the user ID to this value |
| 3169 | before the thread/process does any work. |
| 3170 | .TP |
| 3171 | .BI gid \fR=\fPint |
| 3172 | Set group ID, see \fBuid\fR. |
| 3173 | .SS "Verification" |
| 3174 | .TP |
| 3175 | .BI verify_only |
| 3176 | Do not perform specified workload, only verify data still matches previous |
| 3177 | invocation of this workload. This option allows one to check data multiple |
| 3178 | times at a later date without overwriting it. This option makes sense only |
| 3179 | for workloads that write data, and does not support workloads with the |
| 3180 | \fBtime_based\fR option set. |
| 3181 | .TP |
| 3182 | .BI do_verify \fR=\fPbool |
| 3183 | Run the verify phase after a write phase. Only valid if \fBverify\fR is |
| 3184 | set. Default: true. |
| 3185 | .TP |
| 3186 | .BI verify \fR=\fPstr |
| 3187 | If writing to a file, fio can verify the file contents after each iteration |
| 3188 | of the job. Each verification method also implies verification of special |
| 3189 | header, which is written to the beginning of each block. This header also |
| 3190 | includes meta information, like offset of the block, block number, timestamp |
| 3191 | when block was written, etc. \fBverify\fR can be combined with |
| 3192 | \fBverify_pattern\fR option. The allowed values are: |
| 3193 | .RS |
| 3194 | .RS |
| 3195 | .TP |
| 3196 | .B md5 |
| 3197 | Use an md5 sum of the data area and store it in the header of |
| 3198 | each block. |
| 3199 | .TP |
| 3200 | .B crc64 |
| 3201 | Use an experimental crc64 sum of the data area and store it in the |
| 3202 | header of each block. |
| 3203 | .TP |
| 3204 | .B crc32c |
| 3205 | Use a crc32c sum of the data area and store it in the header of |
| 3206 | each block. This will automatically use hardware acceleration |
| 3207 | (e.g. SSE4.2 on an x86 or CRC crypto extensions on ARM64) but will |
| 3208 | fall back to software crc32c if none is found. Generally the |
| 3209 | fastest checksum fio supports when hardware accelerated. |
| 3210 | .TP |
| 3211 | .B crc32c\-intel |
| 3212 | Synonym for crc32c. |
| 3213 | .TP |
| 3214 | .B crc32 |
| 3215 | Use a crc32 sum of the data area and store it in the header of each |
| 3216 | block. |
| 3217 | .TP |
| 3218 | .B crc16 |
| 3219 | Use a crc16 sum of the data area and store it in the header of each |
| 3220 | block. |
| 3221 | .TP |
| 3222 | .B crc7 |
| 3223 | Use a crc7 sum of the data area and store it in the header of each |
| 3224 | block. |
| 3225 | .TP |
| 3226 | .B xxhash |
| 3227 | Use xxhash as the checksum function. Generally the fastest software |
| 3228 | checksum that fio supports. |
| 3229 | .TP |
| 3230 | .B sha512 |
| 3231 | Use sha512 as the checksum function. |
| 3232 | .TP |
| 3233 | .B sha256 |
| 3234 | Use sha256 as the checksum function. |
| 3235 | .TP |
| 3236 | .B sha1 |
| 3237 | Use optimized sha1 as the checksum function. |
| 3238 | .TP |
| 3239 | .B sha3\-224 |
| 3240 | Use optimized sha3\-224 as the checksum function. |
| 3241 | .TP |
| 3242 | .B sha3\-256 |
| 3243 | Use optimized sha3\-256 as the checksum function. |
| 3244 | .TP |
| 3245 | .B sha3\-384 |
| 3246 | Use optimized sha3\-384 as the checksum function. |
| 3247 | .TP |
| 3248 | .B sha3\-512 |
| 3249 | Use optimized sha3\-512 as the checksum function. |
| 3250 | .TP |
| 3251 | .B meta |
| 3252 | This option is deprecated, since now meta information is included in |
| 3253 | generic verification header and meta verification happens by |
| 3254 | default. For detailed information see the description of the |
| 3255 | \fBverify\fR setting. This option is kept because of |
| 3256 | compatibility's sake with old configurations. Do not use it. |
| 3257 | .TP |
| 3258 | .B pattern |
| 3259 | Verify a strict pattern. Normally fio includes a header with some |
| 3260 | basic information and checksumming, but if this option is set, only |
| 3261 | the specific pattern set with \fBverify_pattern\fR is verified. |
| 3262 | .TP |
| 3263 | .B null |
| 3264 | Only pretend to verify. Useful for testing internals with |
| 3265 | `ioengine=null', not for much else. |
| 3266 | .RE |
| 3267 | .P |
| 3268 | This option can be used for repeated burn\-in tests of a system to make sure |
| 3269 | that the written data is also correctly read back. If the data direction |
| 3270 | given is a read or random read, fio will assume that it should verify a |
| 3271 | previously written file. If the data direction includes any form of write, |
| 3272 | the verify will be of the newly written data. |
| 3273 | .P |
| 3274 | To avoid false verification errors, do not use the norandommap option when |
| 3275 | verifying data with async I/O engines and I/O depths > 1. Or use the |
| 3276 | norandommap and the lfsr random generator together to avoid writing to the |
| 3277 | same offset with multiple outstanding I/Os. |
| 3278 | .RE |
| 3279 | .TP |
| 3280 | .BI verify_offset \fR=\fPint |
| 3281 | Swap the verification header with data somewhere else in the block before |
| 3282 | writing. It is swapped back before verifying. |
| 3283 | .TP |
| 3284 | .BI verify_interval \fR=\fPint |
| 3285 | Write the verification header at a finer granularity than the |
| 3286 | \fBblocksize\fR. It will be written for chunks the size of |
| 3287 | \fBverify_interval\fR. \fBblocksize\fR should divide this evenly. |
| 3288 | .TP |
| 3289 | .BI verify_pattern \fR=\fPstr |
| 3290 | If set, fio will fill the I/O buffers with this pattern. Fio defaults to |
| 3291 | filling with totally random bytes, but sometimes it's interesting to fill |
| 3292 | with a known pattern for I/O verification purposes. Depending on the width |
| 3293 | of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time (it can |
| 3294 | be either a decimal or a hex number). The \fBverify_pattern\fR if larger than |
| 3295 | a 32\-bit quantity has to be a hex number that starts with either "0x" or |
| 3296 | "0X". Use with \fBverify\fR. Also, \fBverify_pattern\fR supports %o |
| 3297 | format, which means that for each block offset will be written and then |
| 3298 | verified back, e.g.: |
| 3299 | .RS |
| 3300 | .RS |
| 3301 | .P |
| 3302 | verify_pattern=%o |
| 3303 | .RE |
| 3304 | .P |
| 3305 | Or use combination of everything: |
| 3306 | .RS |
| 3307 | .P |
| 3308 | verify_pattern=0xff%o"abcd"\-12 |
| 3309 | .RE |
| 3310 | .RE |
| 3311 | .TP |
| 3312 | .BI verify_fatal \fR=\fPbool |
| 3313 | Normally fio will keep checking the entire contents before quitting on a |
| 3314 | block verification failure. If this option is set, fio will exit the job on |
| 3315 | the first observed failure. Default: false. |
| 3316 | .TP |
| 3317 | .BI verify_dump \fR=\fPbool |
| 3318 | If set, dump the contents of both the original data block and the data block |
| 3319 | we read off disk to files. This allows later analysis to inspect just what |
| 3320 | kind of data corruption occurred. Off by default. |
| 3321 | .TP |
| 3322 | .BI verify_async \fR=\fPint |
| 3323 | Fio will normally verify I/O inline from the submitting thread. This option |
| 3324 | takes an integer describing how many async offload threads to create for I/O |
| 3325 | verification instead, causing fio to offload the duty of verifying I/O |
| 3326 | contents to one or more separate threads. If using this offload option, even |
| 3327 | sync I/O engines can benefit from using an \fBiodepth\fR setting higher |
| 3328 | than 1, as it allows them to have I/O in flight while verifies are running. |
| 3329 | Defaults to 0 async threads, i.e. verification is not asynchronous. |
| 3330 | .TP |
| 3331 | .BI verify_async_cpus \fR=\fPstr |
| 3332 | Tell fio to set the given CPU affinity on the async I/O verification |
| 3333 | threads. See \fBcpus_allowed\fR for the format used. |
| 3334 | .TP |
| 3335 | .BI verify_backlog \fR=\fPint |
| 3336 | Fio will normally verify the written contents of a job that utilizes verify |
| 3337 | once that job has completed. In other words, everything is written then |
| 3338 | everything is read back and verified. You may want to verify continually |
| 3339 | instead for a variety of reasons. Fio stores the meta data associated with |
| 3340 | an I/O block in memory, so for large verify workloads, quite a bit of memory |
| 3341 | would be used up holding this meta data. If this option is enabled, fio will |
| 3342 | write only N blocks before verifying these blocks. |
| 3343 | .TP |
| 3344 | .BI verify_backlog_batch \fR=\fPint |
| 3345 | Control how many blocks fio will verify if \fBverify_backlog\fR is |
| 3346 | set. If not set, will default to the value of \fBverify_backlog\fR |
| 3347 | (meaning the entire queue is read back and verified). If |
| 3348 | \fBverify_backlog_batch\fR is less than \fBverify_backlog\fR then not all |
| 3349 | blocks will be verified, if \fBverify_backlog_batch\fR is larger than |
| 3350 | \fBverify_backlog\fR, some blocks will be verified more than once. |
| 3351 | .TP |
| 3352 | .BI verify_state_save \fR=\fPbool |
| 3353 | When a job exits during the write phase of a verify workload, save its |
| 3354 | current state. This allows fio to replay up until that point, if the verify |
| 3355 | state is loaded for the verify read phase. The format of the filename is, |
| 3356 | roughly: |
| 3357 | .RS |
| 3358 | .RS |
| 3359 | .P |
| 3360 | <type>\-<jobname>\-<jobindex>\-verify.state. |
| 3361 | .RE |
| 3362 | .P |
| 3363 | <type> is "local" for a local run, "sock" for a client/server socket |
| 3364 | connection, and "ip" (192.168.0.1, for instance) for a networked |
| 3365 | client/server connection. Defaults to true. |
| 3366 | .RE |
| 3367 | .TP |
| 3368 | .BI verify_state_load \fR=\fPbool |
| 3369 | If a verify termination trigger was used, fio stores the current write state |
| 3370 | of each thread. This can be used at verification time so that fio knows how |
| 3371 | far it should verify. Without this information, fio will run a full |
| 3372 | verification pass, according to the settings in the job file used. Default |
| 3373 | false. |
| 3374 | .TP |
| 3375 | .BI trim_percentage \fR=\fPint |
| 3376 | Number of verify blocks to discard/trim. |
| 3377 | .TP |
| 3378 | .BI trim_verify_zero \fR=\fPbool |
| 3379 | Verify that trim/discarded blocks are returned as zeros. |
| 3380 | .TP |
| 3381 | .BI trim_backlog \fR=\fPint |
| 3382 | Verify that trim/discarded blocks are returned as zeros. |
| 3383 | .TP |
| 3384 | .BI trim_backlog_batch \fR=\fPint |
| 3385 | Trim this number of I/O blocks. |
| 3386 | .TP |
| 3387 | .BI experimental_verify \fR=\fPbool |
| 3388 | Enable experimental verification. Standard verify records I/O metadata for |
| 3389 | later use during the verification phase. Experimental verify instead resets the |
| 3390 | file after the write phase and then replays I/Os for the verification phase. |
| 3391 | .SS "Steady state" |
| 3392 | .TP |
| 3393 | .BI steadystate \fR=\fPstr:float "\fR,\fP ss" \fR=\fPstr:float |
| 3394 | Define the criterion and limit for assessing steady state performance. The |
| 3395 | first parameter designates the criterion whereas the second parameter sets |
| 3396 | the threshold. When the criterion falls below the threshold for the |
| 3397 | specified duration, the job will stop. For example, `iops_slope:0.1%' will |
| 3398 | direct fio to terminate the job when the least squares regression slope |
| 3399 | falls below 0.1% of the mean IOPS. If \fBgroup_reporting\fR is enabled |
| 3400 | this will apply to all jobs in the group. Below is the list of available |
| 3401 | steady state assessment criteria. All assessments are carried out using only |
| 3402 | data from the rolling collection window. Threshold limits can be expressed |
| 3403 | as a fixed value or as a percentage of the mean in the collection window. |
| 3404 | .RS |
| 3405 | .P |
| 3406 | When using this feature, most jobs should include the \fBtime_based\fR |
| 3407 | and \fBruntime\fR options or the \fBloops\fR option so that fio does not |
| 3408 | stop running after it has covered the full size of the specified file(s) |
| 3409 | or device(s). |
| 3410 | .RS |
| 3411 | .RS |
| 3412 | .TP |
| 3413 | .B iops |
| 3414 | Collect IOPS data. Stop the job if all individual IOPS measurements |
| 3415 | are within the specified limit of the mean IOPS (e.g., `iops:2' |
| 3416 | means that all individual IOPS values must be within 2 of the mean, |
| 3417 | whereas `iops:0.2%' means that all individual IOPS values must be |
| 3418 | within 0.2% of the mean IOPS to terminate the job). |
| 3419 | .TP |
| 3420 | .B iops_slope |
| 3421 | Collect IOPS data and calculate the least squares regression |
| 3422 | slope. Stop the job if the slope falls below the specified limit. |
| 3423 | .TP |
| 3424 | .B bw |
| 3425 | Collect bandwidth data. Stop the job if all individual bandwidth |
| 3426 | measurements are within the specified limit of the mean bandwidth. |
| 3427 | .TP |
| 3428 | .B bw_slope |
| 3429 | Collect bandwidth data and calculate the least squares regression |
| 3430 | slope. Stop the job if the slope falls below the specified limit. |
| 3431 | .RE |
| 3432 | .RE |
| 3433 | .TP |
| 3434 | .BI steadystate_duration \fR=\fPtime "\fR,\fP ss_dur" \fR=\fPtime |
| 3435 | A rolling window of this duration will be used to judge whether steady state |
| 3436 | has been reached. Data will be collected once per second. The default is 0 |
| 3437 | which disables steady state detection. When the unit is omitted, the |
| 3438 | value is interpreted in seconds. |
| 3439 | .TP |
| 3440 | .BI steadystate_ramp_time \fR=\fPtime "\fR,\fP ss_ramp" \fR=\fPtime |
| 3441 | Allow the job to run for the specified duration before beginning data |
| 3442 | collection for checking the steady state job termination criterion. The |
| 3443 | default is 0. When the unit is omitted, the value is interpreted in seconds. |
| 3444 | .SS "Measurements and reporting" |
| 3445 | .TP |
| 3446 | .BI per_job_logs \fR=\fPbool |
| 3447 | If set, this generates bw/clat/iops log with per file private filenames. If |
| 3448 | not set, jobs with identical names will share the log filename. Default: |
| 3449 | true. |
| 3450 | .TP |
| 3451 | .BI group_reporting |
| 3452 | It may sometimes be interesting to display statistics for groups of jobs as |
| 3453 | a whole instead of for each individual job. This is especially true if |
| 3454 | \fBnumjobs\fR is used; looking at individual thread/process output |
| 3455 | quickly becomes unwieldy. To see the final report per-group instead of |
| 3456 | per-job, use \fBgroup_reporting\fR. Jobs in a file will be part of the |
| 3457 | same reporting group, unless if separated by a \fBstonewall\fR, or by |
| 3458 | using \fBnew_group\fR. |
| 3459 | .TP |
| 3460 | .BI new_group |
| 3461 | Start a new reporting group. See: \fBgroup_reporting\fR. If not given, |
| 3462 | all jobs in a file will be part of the same reporting group, unless |
| 3463 | separated by a \fBstonewall\fR. |
| 3464 | .TP |
| 3465 | .BI stats \fR=\fPbool |
| 3466 | By default, fio collects and shows final output results for all jobs |
| 3467 | that run. If this option is set to 0, then fio will ignore it in |
| 3468 | the final stat output. |
| 3469 | .TP |
| 3470 | .BI write_bw_log \fR=\fPstr |
| 3471 | If given, write a bandwidth log for this job. Can be used to store data of |
| 3472 | the bandwidth of the jobs in their lifetime. |
| 3473 | .RS |
| 3474 | .P |
| 3475 | If no str argument is given, the default filename of |
| 3476 | `jobname_type.x.log' is used. Even when the argument is given, fio |
| 3477 | will still append the type of log. So if one specifies: |
| 3478 | .RS |
| 3479 | .P |
| 3480 | write_bw_log=foo |
| 3481 | .RE |
| 3482 | .P |
| 3483 | The actual log name will be `foo_bw.x.log' where `x' is the index |
| 3484 | of the job (1..N, where N is the number of jobs). If |
| 3485 | \fBper_job_logs\fR is false, then the filename will not include the |
| 3486 | `.x` job index. |
| 3487 | .P |
| 3488 | The included \fBfio_generate_plots\fR script uses gnuplot to turn these |
| 3489 | text files into nice graphs. See the \fBLOG FILE FORMATS\fR section for how data is |
| 3490 | structured within the file. |
| 3491 | .RE |
| 3492 | .TP |
| 3493 | .BI write_lat_log \fR=\fPstr |
| 3494 | Same as \fBwrite_bw_log\fR, except this option creates I/O |
| 3495 | submission (e.g., `name_slat.x.log'), completion (e.g., |
| 3496 | `name_clat.x.log'), and total (e.g., `name_lat.x.log') latency |
| 3497 | files instead. See \fBwrite_bw_log\fR for details about the |
| 3498 | filename format and the \fBLOG FILE FORMATS\fR section for how data is structured |
| 3499 | within the files. |
| 3500 | .TP |
| 3501 | .BI write_hist_log \fR=\fPstr |
| 3502 | Same as \fBwrite_bw_log\fR but writes an I/O completion latency |
| 3503 | histogram file (e.g., `name_hist.x.log') instead. Note that this |
| 3504 | file will be empty unless \fBlog_hist_msec\fR has also been set. |
| 3505 | See \fBwrite_bw_log\fR for details about the filename format and |
| 3506 | the \fBLOG FILE FORMATS\fR section for how data is structured |
| 3507 | within the file. |
| 3508 | .TP |
| 3509 | .BI write_iops_log \fR=\fPstr |
| 3510 | Same as \fBwrite_bw_log\fR, but writes an IOPS file (e.g. |
| 3511 | `name_iops.x.log`) instead. Because fio defaults to individual |
| 3512 | I/O logging, the value entry in the IOPS log will be 1 unless windowed |
| 3513 | logging (see \fBlog_avg_msec\fR) has been enabled. See |
| 3514 | \fBwrite_bw_log\fR for details about the filename format and \fBLOG |
| 3515 | FILE FORMATS\fR for how data is structured within the file. |
| 3516 | .TP |
| 3517 | .BI log_entries \fR=\fPint |
| 3518 | By default, fio will log an entry in the iops, latency, or bw log for |
| 3519 | every I/O that completes. The initial number of I/O log entries is 1024. |
| 3520 | When the log entries are all used, new log entries are dynamically |
| 3521 | allocated. This dynamic log entry allocation may negatively impact |
| 3522 | time-related statistics such as I/O tail latencies (e.g. 99.9th percentile |
| 3523 | completion latency). This option allows specifying a larger initial |
| 3524 | number of log entries to avoid run-time allocation of new log entries, |
| 3525 | resulting in more precise time-related I/O statistics. |
| 3526 | Also see \fBlog_avg_msec\fR as well. Defaults to 1024. |
| 3527 | .TP |
| 3528 | .BI log_avg_msec \fR=\fPint |
| 3529 | By default, fio will log an entry in the iops, latency, or bw log for every |
| 3530 | I/O that completes. When writing to the disk log, that can quickly grow to a |
| 3531 | very large size. Setting this option makes fio average the each log entry |
| 3532 | over the specified period of time, reducing the resolution of the log. See |
| 3533 | \fBlog_max_value\fR as well. Defaults to 0, logging all entries. |
| 3534 | Also see \fBLOG FILE FORMATS\fR section. |
| 3535 | .TP |
| 3536 | .BI log_hist_msec \fR=\fPint |
| 3537 | Same as \fBlog_avg_msec\fR, but logs entries for completion latency |
| 3538 | histograms. Computing latency percentiles from averages of intervals using |
| 3539 | \fBlog_avg_msec\fR is inaccurate. Setting this option makes fio log |
| 3540 | histogram entries over the specified period of time, reducing log sizes for |
| 3541 | high IOPS devices while retaining percentile accuracy. See |
| 3542 | \fBlog_hist_coarseness\fR and \fBwrite_hist_log\fR as well. |
| 3543 | Defaults to 0, meaning histogram logging is disabled. |
| 3544 | .TP |
| 3545 | .BI log_hist_coarseness \fR=\fPint |
| 3546 | Integer ranging from 0 to 6, defining the coarseness of the resolution of |
| 3547 | the histogram logs enabled with \fBlog_hist_msec\fR. For each increment |
| 3548 | in coarseness, fio outputs half as many bins. Defaults to 0, for which |
| 3549 | histogram logs contain 1216 latency bins. See \fBLOG FILE FORMATS\fR section. |
| 3550 | .TP |
| 3551 | .BI log_max_value \fR=\fPbool |
| 3552 | If \fBlog_avg_msec\fR is set, fio logs the average over that window. If |
| 3553 | you instead want to log the maximum value, set this option to 1. Defaults to |
| 3554 | 0, meaning that averaged values are logged. |
| 3555 | .TP |
| 3556 | .BI log_offset \fR=\fPbool |
| 3557 | If this is set, the iolog options will include the byte offset for the I/O |
| 3558 | entry as well as the other data values. Defaults to 0 meaning that |
| 3559 | offsets are not present in logs. Also see \fBLOG FILE FORMATS\fR section. |
| 3560 | .TP |
| 3561 | .BI log_prio \fR=\fPbool |
| 3562 | If this is set, the iolog options will include the I/O priority for the I/O |
| 3563 | entry as well as the other data values. Defaults to 0 meaning that |
| 3564 | I/O priorities are not present in logs. Also see \fBLOG FILE FORMATS\fR section. |
| 3565 | .TP |
| 3566 | .BI log_compression \fR=\fPint |
| 3567 | If this is set, fio will compress the I/O logs as it goes, to keep the |
| 3568 | memory footprint lower. When a log reaches the specified size, that chunk is |
| 3569 | removed and compressed in the background. Given that I/O logs are fairly |
| 3570 | highly compressible, this yields a nice memory savings for longer runs. The |
| 3571 | downside is that the compression will consume some background CPU cycles, so |
| 3572 | it may impact the run. This, however, is also true if the logging ends up |
| 3573 | consuming most of the system memory. So pick your poison. The I/O logs are |
| 3574 | saved normally at the end of a run, by decompressing the chunks and storing |
| 3575 | them in the specified log file. This feature depends on the availability of |
| 3576 | zlib. |
| 3577 | .TP |
| 3578 | .BI log_compression_cpus \fR=\fPstr |
| 3579 | Define the set of CPUs that are allowed to handle online log compression for |
| 3580 | the I/O jobs. This can provide better isolation between performance |
| 3581 | sensitive jobs, and background compression work. See \fBcpus_allowed\fR for |
| 3582 | the format used. |
| 3583 | .TP |
| 3584 | .BI log_store_compressed \fR=\fPbool |
| 3585 | If set, fio will store the log files in a compressed format. They can be |
| 3586 | decompressed with fio, using the \fB\-\-inflate\-log\fR command line |
| 3587 | parameter. The files will be stored with a `.fz' suffix. |
| 3588 | .TP |
| 3589 | .BI log_unix_epoch \fR=\fPbool |
| 3590 | If set, fio will log Unix timestamps to the log files produced by enabling |
| 3591 | write_type_log for each log type, instead of the default zero-based |
| 3592 | timestamps. |
| 3593 | .TP |
| 3594 | .BI log_alternate_epoch \fR=\fPbool |
| 3595 | If set, fio will log timestamps based on the epoch used by the clock specified |
| 3596 | in the \fBlog_alternate_epoch_clock_id\fR option, to the log files produced by |
| 3597 | enabling write_type_log for each log type, instead of the default zero-based |
| 3598 | timestamps. |
| 3599 | .TP |
| 3600 | .BI log_alternate_epoch_clock_id \fR=\fPint |
| 3601 | Specifies the clock_id to be used by clock_gettime to obtain the alternate epoch |
| 3602 | if either \fBBlog_unix_epoch\fR or \fBlog_alternate_epoch\fR are true. Otherwise has no |
| 3603 | effect. Default value is 0, or CLOCK_REALTIME. |
| 3604 | .TP |
| 3605 | .BI block_error_percentiles \fR=\fPbool |
| 3606 | If set, record errors in trim block-sized units from writes and trims and |
| 3607 | output a histogram of how many trims it took to get to errors, and what kind |
| 3608 | of error was encountered. |
| 3609 | .TP |
| 3610 | .BI bwavgtime \fR=\fPint |
| 3611 | Average the calculated bandwidth over the given time. Value is specified in |
| 3612 | milliseconds. If the job also does bandwidth logging through |
| 3613 | \fBwrite_bw_log\fR, then the minimum of this option and |
| 3614 | \fBlog_avg_msec\fR will be used. Default: 500ms. |
| 3615 | .TP |
| 3616 | .BI iopsavgtime \fR=\fPint |
| 3617 | Average the calculated IOPS over the given time. Value is specified in |
| 3618 | milliseconds. If the job also does IOPS logging through |
| 3619 | \fBwrite_iops_log\fR, then the minimum of this option and |
| 3620 | \fBlog_avg_msec\fR will be used. Default: 500ms. |
| 3621 | .TP |
| 3622 | .BI disk_util \fR=\fPbool |
| 3623 | Generate disk utilization statistics, if the platform supports it. |
| 3624 | Default: true. |
| 3625 | .TP |
| 3626 | .BI disable_lat \fR=\fPbool |
| 3627 | Disable measurements of total latency numbers. Useful only for cutting back |
| 3628 | the number of calls to \fBgettimeofday\fR\|(2), as that does impact |
| 3629 | performance at really high IOPS rates. Note that to really get rid of a |
| 3630 | large amount of these calls, this option must be used with |
| 3631 | \fBdisable_slat\fR and \fBdisable_bw_measurement\fR as well. |
| 3632 | .TP |
| 3633 | .BI disable_clat \fR=\fPbool |
| 3634 | Disable measurements of completion latency numbers. See |
| 3635 | \fBdisable_lat\fR. |
| 3636 | .TP |
| 3637 | .BI disable_slat \fR=\fPbool |
| 3638 | Disable measurements of submission latency numbers. See |
| 3639 | \fBdisable_lat\fR. |
| 3640 | .TP |
| 3641 | .BI disable_bw_measurement \fR=\fPbool "\fR,\fP disable_bw" \fR=\fPbool |
| 3642 | Disable measurements of throughput/bandwidth numbers. See |
| 3643 | \fBdisable_lat\fR. |
| 3644 | .TP |
| 3645 | .BI slat_percentiles \fR=\fPbool |
| 3646 | Report submission latency percentiles. Submission latency is not recorded |
| 3647 | for synchronous ioengines. |
| 3648 | .TP |
| 3649 | .BI clat_percentiles \fR=\fPbool |
| 3650 | Report completion latency percentiles. |
| 3651 | .TP |
| 3652 | .BI lat_percentiles \fR=\fPbool |
| 3653 | Report total latency percentiles. Total latency is the sum of submission |
| 3654 | latency and completion latency. |
| 3655 | .TP |
| 3656 | .BI percentile_list \fR=\fPfloat_list |
| 3657 | Overwrite the default list of percentiles for latencies and the |
| 3658 | block error histogram. Each number is a floating point number in the range |
| 3659 | (0,100], and the maximum length of the list is 20. Use ':' to separate the |
| 3660 | numbers. For example, `\-\-percentile_list=99.5:99.9' will cause fio to |
| 3661 | report the latency durations below which 99.5% and 99.9% of the observed |
| 3662 | latencies fell, respectively. |
| 3663 | .TP |
| 3664 | .BI significant_figures \fR=\fPint |
| 3665 | If using \fB\-\-output\-format\fR of `normal', set the significant figures |
| 3666 | to this value. Higher values will yield more precise IOPS and throughput |
| 3667 | units, while lower values will round. Requires a minimum value of 1 and a |
| 3668 | maximum value of 10. Defaults to 4. |
| 3669 | .SS "Error handling" |
| 3670 | .TP |
| 3671 | .BI exitall_on_error |
| 3672 | When one job finishes in error, terminate the rest. The default is to wait |
| 3673 | for each job to finish. |
| 3674 | .TP |
| 3675 | .BI continue_on_error \fR=\fPstr |
| 3676 | Normally fio will exit the job on the first observed failure. If this option |
| 3677 | is set, fio will continue the job when there is a 'non-fatal error' (EIO or |
| 3678 | EILSEQ) until the runtime is exceeded or the I/O size specified is |
| 3679 | completed. If this option is used, there are two more stats that are |
| 3680 | appended, the total error count and the first error. The error field given |
| 3681 | in the stats is the first error that was hit during the run. |
| 3682 | .RS |
| 3683 | .P |
| 3684 | Note: a write error from the device may go unnoticed by fio when using buffered |
| 3685 | IO, as the write() (or similar) system call merely dirties the kernel pages, |
| 3686 | unless `sync' or `direct' is used. Device IO errors occur when the dirty data is |
| 3687 | actually written out to disk. If fully sync writes aren't desirable, `fsync' or |
| 3688 | `fdatasync' can be used as well. This is specific to writes, as reads are always |
| 3689 | synchronous. |
| 3690 | .RS |
| 3691 | .P |
| 3692 | The allowed values are: |
| 3693 | .RS |
| 3694 | .RS |
| 3695 | .TP |
| 3696 | .B none |
| 3697 | Exit on any I/O or verify errors. |
| 3698 | .TP |
| 3699 | .B read |
| 3700 | Continue on read errors, exit on all others. |
| 3701 | .TP |
| 3702 | .B write |
| 3703 | Continue on write errors, exit on all others. |
| 3704 | .TP |
| 3705 | .B io |
| 3706 | Continue on any I/O error, exit on all others. |
| 3707 | .TP |
| 3708 | .B verify |
| 3709 | Continue on verify errors, exit on all others. |
| 3710 | .TP |
| 3711 | .B all |
| 3712 | Continue on all errors. |
| 3713 | .TP |
| 3714 | .B 0 |
| 3715 | Backward-compatible alias for 'none'. |
| 3716 | .TP |
| 3717 | .B 1 |
| 3718 | Backward-compatible alias for 'all'. |
| 3719 | .RE |
| 3720 | .RE |
| 3721 | .TP |
| 3722 | .BI ignore_error \fR=\fPstr |
| 3723 | Sometimes you want to ignore some errors during test in that case you can |
| 3724 | specify error list for each error type, instead of only being able to |
| 3725 | ignore the default 'non-fatal error' using \fBcontinue_on_error\fR. |
| 3726 | `ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST' errors for |
| 3727 | given error type is separated with ':'. Error may be symbol ('ENOSPC', 'ENOMEM') |
| 3728 | or integer. Example: |
| 3729 | .RS |
| 3730 | .RS |
| 3731 | .P |
| 3732 | ignore_error=EAGAIN,ENOSPC:122 |
| 3733 | .RE |
| 3734 | .P |
| 3735 | This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from |
| 3736 | WRITE. This option works by overriding \fBcontinue_on_error\fR with |
| 3737 | the list of errors for each error type if any. |
| 3738 | .RE |
| 3739 | .TP |
| 3740 | .BI error_dump \fR=\fPbool |
| 3741 | If set dump every error even if it is non fatal, true by default. If |
| 3742 | disabled only fatal error will be dumped. |
| 3743 | .SS "Running predefined workloads" |
| 3744 | Fio includes predefined profiles that mimic the I/O workloads generated by |
| 3745 | other tools. |
| 3746 | .TP |
| 3747 | .BI profile \fR=\fPstr |
| 3748 | The predefined workload to run. Current profiles are: |
| 3749 | .RS |
| 3750 | .RS |
| 3751 | .TP |
| 3752 | .B tiobench |
| 3753 | Threaded I/O bench (tiotest/tiobench) like workload. |
| 3754 | .TP |
| 3755 | .B act |
| 3756 | Aerospike Certification Tool (ACT) like workload. |
| 3757 | .RE |
| 3758 | .RE |
| 3759 | .P |
| 3760 | To view a profile's additional options use \fB\-\-cmdhelp\fR after specifying |
| 3761 | the profile. For example: |
| 3762 | .RS |
| 3763 | .TP |
| 3764 | $ fio \-\-profile=act \-\-cmdhelp |
| 3765 | .RE |
| 3766 | .SS "Act profile options" |
| 3767 | .TP |
| 3768 | .BI device\-names \fR=\fPstr |
| 3769 | Devices to use. |
| 3770 | .TP |
| 3771 | .BI load \fR=\fPint |
| 3772 | ACT load multiplier. Default: 1. |
| 3773 | .TP |
| 3774 | .BI test\-duration\fR=\fPtime |
| 3775 | How long the entire test takes to run. When the unit is omitted, the value |
| 3776 | is given in seconds. Default: 24h. |
| 3777 | .TP |
| 3778 | .BI threads\-per\-queue\fR=\fPint |
| 3779 | Number of read I/O threads per device. Default: 8. |
| 3780 | .TP |
| 3781 | .BI read\-req\-num\-512\-blocks\fR=\fPint |
| 3782 | Number of 512B blocks to read at the time. Default: 3. |
| 3783 | .TP |
| 3784 | .BI large\-block\-op\-kbytes\fR=\fPint |
| 3785 | Size of large block ops in KiB (writes). Default: 131072. |
| 3786 | .TP |
| 3787 | .BI prep |
| 3788 | Set to run ACT prep phase. |
| 3789 | .SS "Tiobench profile options" |
| 3790 | .TP |
| 3791 | .BI size\fR=\fPstr |
| 3792 | Size in MiB. |
| 3793 | .TP |
| 3794 | .BI block\fR=\fPint |
| 3795 | Block size in bytes. Default: 4096. |
| 3796 | .TP |
| 3797 | .BI numruns\fR=\fPint |
| 3798 | Number of runs. |
| 3799 | .TP |
| 3800 | .BI dir\fR=\fPstr |
| 3801 | Test directory. |
| 3802 | .TP |
| 3803 | .BI threads\fR=\fPint |
| 3804 | Number of threads. |
| 3805 | .SH OUTPUT |
| 3806 | Fio spits out a lot of output. While running, fio will display the status of the |
| 3807 | jobs created. An example of that would be: |
| 3808 | .P |
| 3809 | .nf |
| 3810 | 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] |
| 3811 | .fi |
| 3812 | .P |
| 3813 | The characters inside the first set of square brackets denote the current status of |
| 3814 | each thread. The first character is the first job defined in the job file, and so |
| 3815 | forth. The possible values (in typical life cycle order) are: |
| 3816 | .RS |
| 3817 | .TP |
| 3818 | .PD 0 |
| 3819 | .B P |
| 3820 | Thread setup, but not started. |
| 3821 | .TP |
| 3822 | .B C |
| 3823 | Thread created. |
| 3824 | .TP |
| 3825 | .B I |
| 3826 | Thread initialized, waiting or generating necessary data. |
| 3827 | .TP |
| 3828 | .B p |
| 3829 | Thread running pre-reading file(s). |
| 3830 | .TP |
| 3831 | .B / |
| 3832 | Thread is in ramp period. |
| 3833 | .TP |
| 3834 | .B R |
| 3835 | Running, doing sequential reads. |
| 3836 | .TP |
| 3837 | .B r |
| 3838 | Running, doing random reads. |
| 3839 | .TP |
| 3840 | .B W |
| 3841 | Running, doing sequential writes. |
| 3842 | .TP |
| 3843 | .B w |
| 3844 | Running, doing random writes. |
| 3845 | .TP |
| 3846 | .B M |
| 3847 | Running, doing mixed sequential reads/writes. |
| 3848 | .TP |
| 3849 | .B m |
| 3850 | Running, doing mixed random reads/writes. |
| 3851 | .TP |
| 3852 | .B D |
| 3853 | Running, doing sequential trims. |
| 3854 | .TP |
| 3855 | .B d |
| 3856 | Running, doing random trims. |
| 3857 | .TP |
| 3858 | .B F |
| 3859 | Running, currently waiting for \fBfsync\fR\|(2). |
| 3860 | .TP |
| 3861 | .B V |
| 3862 | Running, doing verification of written data. |
| 3863 | .TP |
| 3864 | .B f |
| 3865 | Thread finishing. |
| 3866 | .TP |
| 3867 | .B E |
| 3868 | Thread exited, not reaped by main thread yet. |
| 3869 | .TP |
| 3870 | .B \- |
| 3871 | Thread reaped. |
| 3872 | .TP |
| 3873 | .B X |
| 3874 | Thread reaped, exited with an error. |
| 3875 | .TP |
| 3876 | .B K |
| 3877 | Thread reaped, exited due to signal. |
| 3878 | .PD |
| 3879 | .RE |
| 3880 | .P |
| 3881 | Fio will condense the thread string as not to take up more space on the command |
| 3882 | line than needed. For instance, if you have 10 readers and 10 writers running, |
| 3883 | the output would look like this: |
| 3884 | .P |
| 3885 | .nf |
| 3886 | 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] |
| 3887 | .fi |
| 3888 | .P |
| 3889 | Note that the status string is displayed in order, so it's possible to tell which of |
| 3890 | the jobs are currently doing what. In the example above this means that jobs 1\-\-10 |
| 3891 | are readers and 11\-\-20 are writers. |
| 3892 | .P |
| 3893 | The other values are fairly self explanatory \-\- number of threads currently |
| 3894 | running and doing I/O, the number of currently open files (f=), the estimated |
| 3895 | completion percentage, the rate of I/O since last check (read speed listed first, |
| 3896 | then write speed and optionally trim speed) in terms of bandwidth and IOPS, |
| 3897 | and time to completion for the current running group. It's impossible to estimate |
| 3898 | runtime of the following groups (if any). |
| 3899 | .P |
| 3900 | When fio is done (or interrupted by Ctrl\-C), it will show the data for |
| 3901 | each thread, group of threads, and disks in that order. For each overall thread (or |
| 3902 | group) the output looks like: |
| 3903 | .P |
| 3904 | .nf |
| 3905 | Client1: (groupid=0, jobs=1): err= 0: pid=16109: Sat Jun 24 12:07:54 2017 |
| 3906 | write: IOPS=88, BW=623KiB/s (638kB/s)(30.4MiB/50032msec) |
| 3907 | slat (nsec): min=500, max=145500, avg=8318.00, stdev=4781.50 |
| 3908 | clat (usec): min=170, max=78367, avg=4019.02, stdev=8293.31 |
| 3909 | lat (usec): min=174, max=78375, avg=4027.34, stdev=8291.79 |
| 3910 | clat percentiles (usec): |
| 3911 | | 1.00th=[ 302], 5.00th=[ 326], 10.00th=[ 343], 20.00th=[ 363], |
| 3912 | | 30.00th=[ 392], 40.00th=[ 404], 50.00th=[ 416], 60.00th=[ 445], |
| 3913 | | 70.00th=[ 816], 80.00th=[ 6718], 90.00th=[12911], 95.00th=[21627], |
| 3914 | | 99.00th=[43779], 99.50th=[51643], 99.90th=[68682], 99.95th=[72877], |
| 3915 | | 99.99th=[78119] |
| 3916 | bw ( KiB/s): min= 532, max= 686, per=0.10%, avg=622.87, stdev=24.82, samples= 100 |
| 3917 | iops : min= 76, max= 98, avg=88.98, stdev= 3.54, samples= 100 |
| 3918 | lat (usec) : 250=0.04%, 500=64.11%, 750=4.81%, 1000=2.79% |
| 3919 | lat (msec) : 2=4.16%, 4=1.84%, 10=4.90%, 20=11.33%, 50=5.37% |
| 3920 | lat (msec) : 100=0.65% |
| 3921 | cpu : usr=0.27%, sys=0.18%, ctx=12072, majf=0, minf=21 |
| 3922 | IO depths : 1=85.0%, 2=13.1%, 4=1.8%, 8=0.1%, 16=0.0%, 32=0.0%, >=64=0.0% |
| 3923 | submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% |
| 3924 | complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% |
| 3925 | issued rwt: total=0,4450,0, short=0,0,0, dropped=0,0,0 |
| 3926 | latency : target=0, window=0, percentile=100.00%, depth=8 |
| 3927 | .fi |
| 3928 | .P |
| 3929 | The job name (or first job's name when using \fBgroup_reporting\fR) is printed, |
| 3930 | along with the group id, count of jobs being aggregated, last error id seen (which |
| 3931 | is 0 when there are no errors), pid/tid of that thread and the time the job/group |
| 3932 | completed. Below are the I/O statistics for each data direction performed (showing |
| 3933 | writes in the example above). In the order listed, they denote: |
| 3934 | .RS |
| 3935 | .TP |
| 3936 | .B read/write/trim |
| 3937 | The string before the colon shows the I/O direction the statistics |
| 3938 | are for. \fIIOPS\fR is the average I/Os performed per second. \fIBW\fR |
| 3939 | is the average bandwidth rate shown as: value in power of 2 format |
| 3940 | (value in power of 10 format). The last two values show: (total |
| 3941 | I/O performed in power of 2 format / \fIruntime\fR of that thread). |
| 3942 | .TP |
| 3943 | .B slat |
| 3944 | Submission latency (\fImin\fR being the minimum, \fImax\fR being the |
| 3945 | maximum, \fIavg\fR being the average, \fIstdev\fR being the standard |
| 3946 | deviation). This is the time it took to submit the I/O. For |
| 3947 | sync I/O this row is not displayed as the slat is really the |
| 3948 | completion latency (since queue/complete is one operation there). |
| 3949 | This value can be in nanoseconds, microseconds or milliseconds \-\-\- |
| 3950 | fio will choose the most appropriate base and print that (in the |
| 3951 | example above nanoseconds was the best scale). Note: in \fB\-\-minimal\fR mode |
| 3952 | latencies are always expressed in microseconds. |
| 3953 | .TP |
| 3954 | .B clat |
| 3955 | Completion latency. Same names as slat, this denotes the time from |
| 3956 | submission to completion of the I/O pieces. For sync I/O, clat will |
| 3957 | usually be equal (or very close) to 0, as the time from submit to |
| 3958 | complete is basically just CPU time (I/O has already been done, see slat |
| 3959 | explanation). |
| 3960 | .TP |
| 3961 | .B lat |
| 3962 | Total latency. Same names as slat and clat, this denotes the time from |
| 3963 | when fio created the I/O unit to completion of the I/O operation. |
| 3964 | .TP |
| 3965 | .B bw |
| 3966 | Bandwidth statistics based on samples. Same names as the xlat stats, |
| 3967 | but also includes the number of samples taken (\fIsamples\fR) and an |
| 3968 | approximate percentage of total aggregate bandwidth this thread |
| 3969 | received in its group (\fIper\fR). This last value is only really |
| 3970 | useful if the threads in this group are on the same disk, since they |
| 3971 | are then competing for disk access. |
| 3972 | .TP |
| 3973 | .B iops |
| 3974 | IOPS statistics based on samples. Same names as \fBbw\fR. |
| 3975 | .TP |
| 3976 | .B lat (nsec/usec/msec) |
| 3977 | The distribution of I/O completion latencies. This is the time from when |
| 3978 | I/O leaves fio and when it gets completed. Unlike the separate |
| 3979 | read/write/trim sections above, the data here and in the remaining |
| 3980 | sections apply to all I/Os for the reporting group. 250=0.04% means that |
| 3981 | 0.04% of the I/Os completed in under 250us. 500=64.11% means that 64.11% |
| 3982 | of the I/Os required 250 to 499us for completion. |
| 3983 | .TP |
| 3984 | .B cpu |
| 3985 | CPU usage. User and system time, along with the number of context |
| 3986 | switches this thread went through, usage of system and user time, and |
| 3987 | finally the number of major and minor page faults. The CPU utilization |
| 3988 | numbers are averages for the jobs in that reporting group, while the |
| 3989 | context and fault counters are summed. |
| 3990 | .TP |
| 3991 | .B IO depths |
| 3992 | The distribution of I/O depths over the job lifetime. The numbers are |
| 3993 | divided into powers of 2 and each entry covers depths from that value |
| 3994 | up to those that are lower than the next entry \-\- e.g., 16= covers |
| 3995 | depths from 16 to 31. Note that the range covered by a depth |
| 3996 | distribution entry can be different to the range covered by the |
| 3997 | equivalent \fBsubmit\fR/\fBcomplete\fR distribution entry. |
| 3998 | .TP |
| 3999 | .B IO submit |
| 4000 | How many pieces of I/O were submitting in a single submit call. Each |
| 4001 | entry denotes that amount and below, until the previous entry \-\- e.g., |
| 4002 | 16=100% means that we submitted anywhere between 9 to 16 I/Os per submit |
| 4003 | call. Note that the range covered by a \fBsubmit\fR distribution entry can |
| 4004 | be different to the range covered by the equivalent depth distribution |
| 4005 | entry. |
| 4006 | .TP |
| 4007 | .B IO complete |
| 4008 | Like the above \fBsubmit\fR number, but for completions instead. |
| 4009 | .TP |
| 4010 | .B IO issued rwt |
| 4011 | The number of \fBread/write/trim\fR requests issued, and how many of them were |
| 4012 | short or dropped. |
| 4013 | .TP |
| 4014 | .B IO latency |
| 4015 | These values are for \fBlatency_target\fR and related options. When |
| 4016 | these options are engaged, this section describes the I/O depth required |
| 4017 | to meet the specified latency target. |
| 4018 | .RE |
| 4019 | .P |
| 4020 | After each client has been listed, the group statistics are printed. They |
| 4021 | will look like this: |
| 4022 | .P |
| 4023 | .nf |
| 4024 | Run status group 0 (all jobs): |
| 4025 | 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 |
| 4026 | WRITE: bw=1231KiB/s (1261kB/s), 616KiB/s\-621KiB/s (630kB/s\-636kB/s), io=64.0MiB (67.1MB), run=52747\-53223msec |
| 4027 | .fi |
| 4028 | .P |
| 4029 | For each data direction it prints: |
| 4030 | .RS |
| 4031 | .TP |
| 4032 | .B bw |
| 4033 | Aggregate bandwidth of threads in this group followed by the |
| 4034 | minimum and maximum bandwidth of all the threads in this group. |
| 4035 | Values outside of brackets are power-of-2 format and those |
| 4036 | within are the equivalent value in a power-of-10 format. |
| 4037 | .TP |
| 4038 | .B io |
| 4039 | Aggregate I/O performed of all threads in this group. The |
| 4040 | format is the same as \fBbw\fR. |
| 4041 | .TP |
| 4042 | .B run |
| 4043 | The smallest and longest runtimes of the threads in this group. |
| 4044 | .RE |
| 4045 | .P |
| 4046 | And finally, the disk statistics are printed. This is Linux specific. |
| 4047 | They will look like this: |
| 4048 | .P |
| 4049 | .nf |
| 4050 | Disk stats (read/write): |
| 4051 | sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00% |
| 4052 | .fi |
| 4053 | .P |
| 4054 | Each value is printed for both reads and writes, with reads first. The |
| 4055 | numbers denote: |
| 4056 | .RS |
| 4057 | .TP |
| 4058 | .B ios |
| 4059 | Number of I/Os performed by all groups. |
| 4060 | .TP |
| 4061 | .B merge |
| 4062 | Number of merges performed by the I/O scheduler. |
| 4063 | .TP |
| 4064 | .B ticks |
| 4065 | Number of ticks we kept the disk busy. |
| 4066 | .TP |
| 4067 | .B in_queue |
| 4068 | Total time spent in the disk queue. |
| 4069 | .TP |
| 4070 | .B util |
| 4071 | The disk utilization. A value of 100% means we kept the disk |
| 4072 | busy constantly, 50% would be a disk idling half of the time. |
| 4073 | .RE |
| 4074 | .P |
| 4075 | It is also possible to get fio to dump the current output while it is running, |
| 4076 | without terminating the job. To do that, send fio the USR1 signal. You can |
| 4077 | also get regularly timed dumps by using the \fB\-\-status\-interval\fR |
| 4078 | parameter, or by creating a file in `/tmp' named |
| 4079 | `fio\-dump\-status'. If fio sees this file, it will unlink it and dump the |
| 4080 | current output status. |
| 4081 | .SH TERSE OUTPUT |
| 4082 | For scripted usage where you typically want to generate tables or graphs of the |
| 4083 | results, fio can output the results in a semicolon separated format. The format |
| 4084 | is one long line of values, such as: |
| 4085 | .P |
| 4086 | .nf |
| 4087 | 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% |
| 4088 | A description of this job goes here. |
| 4089 | .fi |
| 4090 | .P |
| 4091 | The job description (if provided) follows on a second line for terse v2. |
| 4092 | It appears on the same line for other terse versions. |
| 4093 | .P |
| 4094 | To enable terse output, use the \fB\-\-minimal\fR or |
| 4095 | `\-\-output\-format=terse' command line options. The |
| 4096 | first value is the version of the terse output format. If the output has to be |
| 4097 | changed for some reason, this number will be incremented by 1 to signify that |
| 4098 | change. |
| 4099 | .P |
| 4100 | Split up, the format is as follows (comments in brackets denote when a |
| 4101 | field was introduced or whether it's specific to some terse version): |
| 4102 | .P |
| 4103 | .nf |
| 4104 | terse version, fio version [v3], jobname, groupid, error |
| 4105 | .fi |
| 4106 | .RS |
| 4107 | .P |
| 4108 | .B |
| 4109 | READ status: |
| 4110 | .RE |
| 4111 | .P |
| 4112 | .nf |
| 4113 | Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec) |
| 4114 | Submission latency: min, max, mean, stdev (usec) |
| 4115 | Completion latency: min, max, mean, stdev (usec) |
| 4116 | Completion latency percentiles: 20 fields (see below) |
| 4117 | Total latency: min, max, mean, stdev (usec) |
| 4118 | Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5] |
| 4119 | IOPS [v5]: min, max, mean, stdev, number of samples |
| 4120 | .fi |
| 4121 | .RS |
| 4122 | .P |
| 4123 | .B |
| 4124 | WRITE status: |
| 4125 | .RE |
| 4126 | .P |
| 4127 | .nf |
| 4128 | Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec) |
| 4129 | Submission latency: min, max, mean, stdev (usec) |
| 4130 | Completion latency: min, max, mean, stdev (usec) |
| 4131 | Completion latency percentiles: 20 fields (see below) |
| 4132 | Total latency: min, max, mean, stdev (usec) |
| 4133 | Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5] |
| 4134 | IOPS [v5]: min, max, mean, stdev, number of samples |
| 4135 | .fi |
| 4136 | .RS |
| 4137 | .P |
| 4138 | .B |
| 4139 | TRIM status [all but version 3]: |
| 4140 | .RE |
| 4141 | .P |
| 4142 | .nf |
| 4143 | Fields are similar to \fBREAD/WRITE\fR status. |
| 4144 | .fi |
| 4145 | .RS |
| 4146 | .P |
| 4147 | .B |
| 4148 | CPU usage: |
| 4149 | .RE |
| 4150 | .P |
| 4151 | .nf |
| 4152 | user, system, context switches, major faults, minor faults |
| 4153 | .fi |
| 4154 | .RS |
| 4155 | .P |
| 4156 | .B |
| 4157 | I/O depths: |
| 4158 | .RE |
| 4159 | .P |
| 4160 | .nf |
| 4161 | <=1, 2, 4, 8, 16, 32, >=64 |
| 4162 | .fi |
| 4163 | .RS |
| 4164 | .P |
| 4165 | .B |
| 4166 | I/O latencies microseconds: |
| 4167 | .RE |
| 4168 | .P |
| 4169 | .nf |
| 4170 | <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000 |
| 4171 | .fi |
| 4172 | .RS |
| 4173 | .P |
| 4174 | .B |
| 4175 | I/O latencies milliseconds: |
| 4176 | .RE |
| 4177 | .P |
| 4178 | .nf |
| 4179 | <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000 |
| 4180 | .fi |
| 4181 | .RS |
| 4182 | .P |
| 4183 | .B |
| 4184 | Disk utilization [v3]: |
| 4185 | .RE |
| 4186 | .P |
| 4187 | .nf |
| 4188 | disk name, read ios, write ios, read merges, write merges, read ticks, write ticks, time spent in queue, disk utilization percentage |
| 4189 | .fi |
| 4190 | .RS |
| 4191 | .P |
| 4192 | .B |
| 4193 | Additional Info (dependent on continue_on_error, default off): |
| 4194 | .RE |
| 4195 | .P |
| 4196 | .nf |
| 4197 | total # errors, first error code |
| 4198 | .fi |
| 4199 | .RS |
| 4200 | .P |
| 4201 | .B |
| 4202 | Additional Info (dependent on description being set): |
| 4203 | .RE |
| 4204 | .P |
| 4205 | .nf |
| 4206 | Text description |
| 4207 | .fi |
| 4208 | .P |
| 4209 | Completion latency percentiles can be a grouping of up to 20 sets, so for the |
| 4210 | terse output fio writes all of them. Each field will look like this: |
| 4211 | .P |
| 4212 | .nf |
| 4213 | 1.00%=6112 |
| 4214 | .fi |
| 4215 | .P |
| 4216 | which is the Xth percentile, and the `usec' latency associated with it. |
| 4217 | .P |
| 4218 | For \fBDisk utilization\fR, all disks used by fio are shown. So for each disk there |
| 4219 | will be a disk utilization section. |
| 4220 | .P |
| 4221 | Below is a single line containing short names for each of the fields in the |
| 4222 | minimal output v3, separated by semicolons: |
| 4223 | .P |
| 4224 | .nf |
| 4225 | terse_version_3;fio_version;jobname;groupid;error;read_kb;read_bandwidth_kb;read_iops;read_runtime_ms;read_slat_min_us;read_slat_max_us;read_slat_mean_us;read_slat_dev_us;read_clat_min_us;read_clat_max_us;read_clat_mean_us;read_clat_dev_us;read_clat_pct01;read_clat_pct02;read_clat_pct03;read_clat_pct04;read_clat_pct05;read_clat_pct06;read_clat_pct07;read_clat_pct08;read_clat_pct09;read_clat_pct10;read_clat_pct11;read_clat_pct12;read_clat_pct13;read_clat_pct14;read_clat_pct15;read_clat_pct16;read_clat_pct17;read_clat_pct18;read_clat_pct19;read_clat_pct20;read_tlat_min_us;read_lat_max_us;read_lat_mean_us;read_lat_dev_us;read_bw_min_kb;read_bw_max_kb;read_bw_agg_pct;read_bw_mean_kb;read_bw_dev_kb;write_kb;write_bandwidth_kb;write_iops;write_runtime_ms;write_slat_min_us;write_slat_max_us;write_slat_mean_us;write_slat_dev_us;write_clat_min_us;write_clat_max_us;write_clat_mean_us;write_clat_dev_us;write_clat_pct01;write_clat_pct02;write_clat_pct03;write_clat_pct04;write_clat_pct05;write_clat_pct06;write_clat_pct07;write_clat_pct08;write_clat_pct09;write_clat_pct10;write_clat_pct11;write_clat_pct12;write_clat_pct13;write_clat_pct14;write_clat_pct15;write_clat_pct16;write_clat_pct17;write_clat_pct18;write_clat_pct19;write_clat_pct20;write_tlat_min_us;write_lat_max_us;write_lat_mean_us;write_lat_dev_us;write_bw_min_kb;write_bw_max_kb;write_bw_agg_pct;write_bw_mean_kb;write_bw_dev_kb;cpu_user;cpu_sys;cpu_csw;cpu_mjf;cpu_minf;iodepth_1;iodepth_2;iodepth_4;iodepth_8;iodepth_16;iodepth_32;iodepth_64;lat_2us;lat_4us;lat_10us;lat_20us;lat_50us;lat_100us;lat_250us;lat_500us;lat_750us;lat_1000us;lat_2ms;lat_4ms;lat_10ms;lat_20ms;lat_50ms;lat_100ms;lat_250ms;lat_500ms;lat_750ms;lat_1000ms;lat_2000ms;lat_over_2000ms;disk_name;disk_read_iops;disk_write_iops;disk_read_merges;disk_write_merges;disk_read_ticks;write_ticks;disk_queue_time;disk_util |
| 4226 | .fi |
| 4227 | .P |
| 4228 | In client/server mode terse output differs from what appears when jobs are run |
| 4229 | locally. Disk utilization data is omitted from the standard terse output and |
| 4230 | for v3 and later appears on its own separate line at the end of each terse |
| 4231 | reporting cycle. |
| 4232 | .SH JSON OUTPUT |
| 4233 | The \fBjson\fR output format is intended to be both human readable and convenient |
| 4234 | for automated parsing. For the most part its sections mirror those of the |
| 4235 | \fBnormal\fR output. The \fBruntime\fR value is reported in msec and the \fBbw\fR value is |
| 4236 | reported in 1024 bytes per second units. |
| 4237 | .fi |
| 4238 | .SH JSON+ OUTPUT |
| 4239 | The \fBjson+\fR output format is identical to the \fBjson\fR output format except that it |
| 4240 | adds a full dump of the completion latency bins. Each \fBbins\fR object contains a |
| 4241 | set of (key, value) pairs where keys are latency durations and values count how |
| 4242 | many I/Os had completion latencies of the corresponding duration. For example, |
| 4243 | consider: |
| 4244 | .RS |
| 4245 | .P |
| 4246 | "bins" : { "87552" : 1, "89600" : 1, "94720" : 1, "96768" : 1, "97792" : 1, "99840" : 1, "100864" : 2, "103936" : 6, "104960" : 534, "105984" : 5995, "107008" : 7529, ... } |
| 4247 | .RE |
| 4248 | .P |
| 4249 | This data indicates that one I/O required 87,552ns to complete, two I/Os required |
| 4250 | 100,864ns to complete, and 7529 I/Os required 107,008ns to complete. |
| 4251 | .P |
| 4252 | Also included with fio is a Python script \fBfio_jsonplus_clat2csv\fR that takes |
| 4253 | json+ output and generates CSV-formatted latency data suitable for plotting. |
| 4254 | .P |
| 4255 | The latency durations actually represent the midpoints of latency intervals. |
| 4256 | For details refer to `stat.h' in the fio source. |
| 4257 | .SH TRACE FILE FORMAT |
| 4258 | There are two trace file format that you can encounter. The older (v1) format is |
| 4259 | unsupported since version 1.20\-rc3 (March 2008). It will still be described |
| 4260 | below in case that you get an old trace and want to understand it. |
| 4261 | .P |
| 4262 | In any case the trace is a simple text file with a single action per line. |
| 4263 | .TP |
| 4264 | .B Trace file format v1 |
| 4265 | Each line represents a single I/O action in the following format: |
| 4266 | .RS |
| 4267 | .RS |
| 4268 | .P |
| 4269 | rw, offset, length |
| 4270 | .RE |
| 4271 | .P |
| 4272 | where `rw=0/1' for read/write, and the `offset' and `length' entries being in bytes. |
| 4273 | .P |
| 4274 | This format is not supported in fio versions >= 1.20\-rc3. |
| 4275 | .RE |
| 4276 | .TP |
| 4277 | .B Trace file format v2 |
| 4278 | The second version of the trace file format was added in fio version 1.17. It |
| 4279 | allows one to access more than one file per trace and has a bigger set of possible |
| 4280 | file actions. |
| 4281 | .RS |
| 4282 | .P |
| 4283 | The first line of the trace file has to be: |
| 4284 | .RS |
| 4285 | .P |
| 4286 | "fio version 2 iolog" |
| 4287 | .RE |
| 4288 | .P |
| 4289 | Following this can be lines in two different formats, which are described below. |
| 4290 | .P |
| 4291 | .B |
| 4292 | The file management format: |
| 4293 | .RS |
| 4294 | filename action |
| 4295 | .P |
| 4296 | The `filename' is given as an absolute path. The `action' can be one of these: |
| 4297 | .RS |
| 4298 | .TP |
| 4299 | .B add |
| 4300 | Add the given `filename' to the trace. |
| 4301 | .TP |
| 4302 | .B open |
| 4303 | Open the file with the given `filename'. The `filename' has to have |
| 4304 | been added with the \fBadd\fR action before. |
| 4305 | .TP |
| 4306 | .B close |
| 4307 | Close the file with the given `filename'. The file has to have been |
| 4308 | \fBopen\fRed before. |
| 4309 | .RE |
| 4310 | .RE |
| 4311 | .P |
| 4312 | .B |
| 4313 | The file I/O action format: |
| 4314 | .RS |
| 4315 | filename action offset length |
| 4316 | .P |
| 4317 | The `filename' is given as an absolute path, and has to have been \fBadd\fRed and |
| 4318 | \fBopen\fRed before it can be used with this format. The `offset' and `length' are |
| 4319 | given in bytes. The `action' can be one of these: |
| 4320 | .RS |
| 4321 | .TP |
| 4322 | .B wait |
| 4323 | Wait for `offset' microseconds. Everything below 100 is discarded. |
| 4324 | The time is relative to the previous `wait' statement. Note that action `wait` |
| 4325 | is not allowed as of version 3, as the same behavior can be achieved using |
| 4326 | timestamps. |
| 4327 | .TP |
| 4328 | .B read |
| 4329 | Read `length' bytes beginning from `offset'. |
| 4330 | .TP |
| 4331 | .B write |
| 4332 | Write `length' bytes beginning from `offset'. |
| 4333 | .TP |
| 4334 | .B sync |
| 4335 | \fBfsync\fR\|(2) the file. |
| 4336 | .TP |
| 4337 | .B datasync |
| 4338 | \fBfdatasync\fR\|(2) the file. |
| 4339 | .TP |
| 4340 | .B trim |
| 4341 | Trim the given file from the given `offset' for `length' bytes. |
| 4342 | .RE |
| 4343 | .RE |
| 4344 | .RE |
| 4345 | .TP |
| 4346 | .B Trace file format v3 |
| 4347 | The third version of the trace file format was added in fio version 3.31. It |
| 4348 | forces each action to have a timestamp associated with it. |
| 4349 | .RS |
| 4350 | .P |
| 4351 | The first line of the trace file has to be: |
| 4352 | .RS |
| 4353 | .P |
| 4354 | "fio version 3 iolog" |
| 4355 | .RE |
| 4356 | .P |
| 4357 | Following this can be lines in two different formats, which are described below. |
| 4358 | .P |
| 4359 | .B |
| 4360 | The file management format: |
| 4361 | .RS |
| 4362 | timestamp filename action |
| 4363 | .P |
| 4364 | .RE |
| 4365 | .B |
| 4366 | The file I/O action format: |
| 4367 | .RS |
| 4368 | timestamp filename action offset length |
| 4369 | .P |
| 4370 | The `timestamp` is relative to the beginning of the run (ie starts at 0). The |
| 4371 | `filename`, `action`, `offset` and `length` are identical to version 2, except |
| 4372 | that version 3 does not allow the `wait` action. |
| 4373 | .RE |
| 4374 | .RE |
| 4375 | .SH I/O REPLAY \- MERGING TRACES |
| 4376 | Colocation is a common practice used to get the most out of a machine. |
| 4377 | Knowing which workloads play nicely with each other and which ones don't is |
| 4378 | a much harder task. While fio can replay workloads concurrently via multiple |
| 4379 | jobs, it leaves some variability up to the scheduler making results harder to |
| 4380 | reproduce. Merging is a way to make the order of events consistent. |
| 4381 | .P |
| 4382 | Merging is integrated into I/O replay and done when a \fBmerge_blktrace_file\fR |
| 4383 | is specified. The list of files passed to \fBread_iolog\fR go through the merge |
| 4384 | process and output a single file stored to the specified file. The output file is |
| 4385 | passed on as if it were the only file passed to \fBread_iolog\fR. An example would |
| 4386 | look like: |
| 4387 | .RS |
| 4388 | .P |
| 4389 | $ fio \-\-read_iolog="<file1>:<file2>" \-\-merge_blktrace_file="<output_file>" |
| 4390 | .RE |
| 4391 | .P |
| 4392 | Creating only the merged file can be done by passing the command line argument |
| 4393 | \fBmerge-blktrace-only\fR. |
| 4394 | .P |
| 4395 | Scaling traces can be done to see the relative impact of any particular trace |
| 4396 | being slowed down or sped up. \fBmerge_blktrace_scalars\fR takes in a colon |
| 4397 | separated list of percentage scalars. It is index paired with the files passed |
| 4398 | to \fBread_iolog\fR. |
| 4399 | .P |
| 4400 | With scaling, it may be desirable to match the running time of all traces. |
| 4401 | This can be done with \fBmerge_blktrace_iters\fR. It is index paired with |
| 4402 | \fBread_iolog\fR just like \fBmerge_blktrace_scalars\fR. |
| 4403 | .P |
| 4404 | In an example, given two traces, A and B, each 60s long. If we want to see |
| 4405 | the impact of trace A issuing IOs twice as fast and repeat trace A over the |
| 4406 | runtime of trace B, the following can be done: |
| 4407 | .RS |
| 4408 | .P |
| 4409 | $ fio \-\-read_iolog="<trace_a>:"<trace_b>" \-\-merge_blktrace_file"<output_file>" \-\-merge_blktrace_scalars="50:100" \-\-merge_blktrace_iters="2:1" |
| 4410 | .RE |
| 4411 | .P |
| 4412 | This runs trace A at 2x the speed twice for approximately the same runtime as |
| 4413 | a single run of trace B. |
| 4414 | .SH CPU IDLENESS PROFILING |
| 4415 | In some cases, we want to understand CPU overhead in a test. For example, we |
| 4416 | test patches for the specific goodness of whether they reduce CPU usage. |
| 4417 | Fio implements a balloon approach to create a thread per CPU that runs at idle |
| 4418 | priority, meaning that it only runs when nobody else needs the cpu. |
| 4419 | By measuring the amount of work completed by the thread, idleness of each CPU |
| 4420 | can be derived accordingly. |
| 4421 | .P |
| 4422 | An unit work is defined as touching a full page of unsigned characters. Mean and |
| 4423 | standard deviation of time to complete an unit work is reported in "unit work" |
| 4424 | section. Options can be chosen to report detailed percpu idleness or overall |
| 4425 | system idleness by aggregating percpu stats. |
| 4426 | .SH VERIFICATION AND TRIGGERS |
| 4427 | Fio is usually run in one of two ways, when data verification is done. The first |
| 4428 | is a normal write job of some sort with verify enabled. When the write phase has |
| 4429 | completed, fio switches to reads and verifies everything it wrote. The second |
| 4430 | model is running just the write phase, and then later on running the same job |
| 4431 | (but with reads instead of writes) to repeat the same I/O patterns and verify |
| 4432 | the contents. Both of these methods depend on the write phase being completed, |
| 4433 | as fio otherwise has no idea how much data was written. |
| 4434 | .P |
| 4435 | With verification triggers, fio supports dumping the current write state to |
| 4436 | local files. Then a subsequent read verify workload can load this state and know |
| 4437 | exactly where to stop. This is useful for testing cases where power is cut to a |
| 4438 | server in a managed fashion, for instance. |
| 4439 | .P |
| 4440 | A verification trigger consists of two things: |
| 4441 | .RS |
| 4442 | .P |
| 4443 | 1) Storing the write state of each job. |
| 4444 | .P |
| 4445 | 2) Executing a trigger command. |
| 4446 | .RE |
| 4447 | .P |
| 4448 | The write state is relatively small, on the order of hundreds of bytes to single |
| 4449 | kilobytes. It contains information on the number of completions done, the last X |
| 4450 | completions, etc. |
| 4451 | .P |
| 4452 | A trigger is invoked either through creation ('touch') of a specified file in |
| 4453 | the system, or through a timeout setting. If fio is run with |
| 4454 | `\-\-trigger\-file=/tmp/trigger\-file', then it will continually |
| 4455 | check for the existence of `/tmp/trigger\-file'. When it sees this file, it |
| 4456 | will fire off the trigger (thus saving state, and executing the trigger |
| 4457 | command). |
| 4458 | .P |
| 4459 | For client/server runs, there's both a local and remote trigger. If fio is |
| 4460 | running as a server backend, it will send the job states back to the client for |
| 4461 | safe storage, then execute the remote trigger, if specified. If a local trigger |
| 4462 | is specified, the server will still send back the write state, but the client |
| 4463 | will then execute the trigger. |
| 4464 | .RE |
| 4465 | .P |
| 4466 | .B Verification trigger example |
| 4467 | .RS |
| 4468 | Let's say we want to run a powercut test on the remote Linux machine 'server'. |
| 4469 | Our write workload is in `write\-test.fio'. We want to cut power to 'server' at |
| 4470 | some point during the run, and we'll run this test from the safety or our local |
| 4471 | machine, 'localbox'. On the server, we'll start the fio backend normally: |
| 4472 | .RS |
| 4473 | .P |
| 4474 | server# fio \-\-server |
| 4475 | .RE |
| 4476 | .P |
| 4477 | and on the client, we'll fire off the workload: |
| 4478 | .RS |
| 4479 | .P |
| 4480 | localbox$ fio \-\-client=server \-\-trigger\-file=/tmp/my\-trigger \-\-trigger\-remote="bash \-c "echo b > /proc/sysrq\-triger"" |
| 4481 | .RE |
| 4482 | .P |
| 4483 | We set `/tmp/my\-trigger' as the trigger file, and we tell fio to execute: |
| 4484 | .RS |
| 4485 | .P |
| 4486 | echo b > /proc/sysrq\-trigger |
| 4487 | .RE |
| 4488 | .P |
| 4489 | on the server once it has received the trigger and sent us the write state. This |
| 4490 | will work, but it's not really cutting power to the server, it's merely |
| 4491 | abruptly rebooting it. If we have a remote way of cutting power to the server |
| 4492 | through IPMI or similar, we could do that through a local trigger command |
| 4493 | instead. Let's assume we have a script that does IPMI reboot of a given hostname, |
| 4494 | ipmi\-reboot. On localbox, we could then have run fio with a local trigger |
| 4495 | instead: |
| 4496 | .RS |
| 4497 | .P |
| 4498 | localbox$ fio \-\-client=server \-\-trigger\-file=/tmp/my\-trigger \-\-trigger="ipmi\-reboot server" |
| 4499 | .RE |
| 4500 | .P |
| 4501 | For this case, fio would wait for the server to send us the write state, then |
| 4502 | execute `ipmi\-reboot server' when that happened. |
| 4503 | .RE |
| 4504 | .P |
| 4505 | .B Loading verify state |
| 4506 | .RS |
| 4507 | To load stored write state, a read verification job file must contain the |
| 4508 | \fBverify_state_load\fR option. If that is set, fio will load the previously |
| 4509 | stored state. For a local fio run this is done by loading the files directly, |
| 4510 | and on a client/server run, the server backend will ask the client to send the |
| 4511 | files over and load them from there. |
| 4512 | .RE |
| 4513 | .SH LOG FILE FORMATS |
| 4514 | Fio supports a variety of log file formats, for logging latencies, bandwidth, |
| 4515 | and IOPS. The logs share a common format, which looks like this: |
| 4516 | .RS |
| 4517 | .P |
| 4518 | time (msec), value, data direction, block size (bytes), offset (bytes), |
| 4519 | command priority |
| 4520 | .RE |
| 4521 | .P |
| 4522 | `Time' for the log entry is always in milliseconds. The `value' logged depends |
| 4523 | on the type of log, it will be one of the following: |
| 4524 | .RS |
| 4525 | .TP |
| 4526 | .B Latency log |
| 4527 | Value is latency in nsecs |
| 4528 | .TP |
| 4529 | .B Bandwidth log |
| 4530 | Value is in KiB/sec |
| 4531 | .TP |
| 4532 | .B IOPS log |
| 4533 | Value is IOPS |
| 4534 | .RE |
| 4535 | .P |
| 4536 | `Data direction' is one of the following: |
| 4537 | .RS |
| 4538 | .TP |
| 4539 | .B 0 |
| 4540 | I/O is a READ |
| 4541 | .TP |
| 4542 | .B 1 |
| 4543 | I/O is a WRITE |
| 4544 | .TP |
| 4545 | .B 2 |
| 4546 | I/O is a TRIM |
| 4547 | .RE |
| 4548 | .P |
| 4549 | The entry's `block size' is always in bytes. The `offset' is the position in bytes |
| 4550 | from the start of the file for that particular I/O. The logging of the offset can be |
| 4551 | toggled with \fBlog_offset\fR. |
| 4552 | .P |
| 4553 | If \fBlog_prio\fR is not set, the entry's `Command priority` is 1 for an IO executed |
| 4554 | with the highest RT priority class (\fBprioclass\fR=1 or \fBcmdprio_class\fR=1) and 0 |
| 4555 | otherwise. This is controlled by the \fBprioclass\fR option and the ioengine specific |
| 4556 | \fBcmdprio_percentage\fR \fBcmdprio_class\fR options. If \fBlog_prio\fR is set, the |
| 4557 | entry's `Command priority` is the priority set for the IO, as a 16-bits hexadecimal |
| 4558 | number with the lowest 13 bits indicating the priority value (\fBprio\fR and |
| 4559 | \fBcmdprio\fR options) and the highest 3 bits indicating the IO priority class |
| 4560 | (\fBprioclass\fR and \fBcmdprio_class\fR options). |
| 4561 | .P |
| 4562 | Fio defaults to logging every individual I/O but when windowed logging is set |
| 4563 | through \fBlog_avg_msec\fR, either the average (by default) or the maximum |
| 4564 | (\fBlog_max_value\fR is set) `value' seen over the specified period of time |
| 4565 | is recorded. Each `data direction' seen within the window period will aggregate |
| 4566 | its values in a separate row. Further, when using windowed logging the `block |
| 4567 | size' and `offset' entries will always contain 0. |
| 4568 | .SH CLIENT / SERVER |
| 4569 | Normally fio is invoked as a stand-alone application on the machine where the |
| 4570 | I/O workload should be generated. However, the backend and frontend of fio can |
| 4571 | be run separately i.e., the fio server can generate an I/O workload on the "Device |
| 4572 | Under Test" while being controlled by a client on another machine. |
| 4573 | .P |
| 4574 | Start the server on the machine which has access to the storage DUT: |
| 4575 | .RS |
| 4576 | .P |
| 4577 | $ fio \-\-server=args |
| 4578 | .RE |
| 4579 | .P |
| 4580 | where `args' defines what fio listens to. The arguments are of the form |
| 4581 | `type,hostname' or `IP,port'. `type' is either `ip' (or ip4) for TCP/IP |
| 4582 | v4, `ip6' for TCP/IP v6, or `sock' for a local unix domain socket. |
| 4583 | `hostname' is either a hostname or IP address, and `port' is the port to listen |
| 4584 | to (only valid for TCP/IP, not a local socket). Some examples: |
| 4585 | .RS |
| 4586 | .TP |
| 4587 | 1) \fBfio \-\-server\fR |
| 4588 | Start a fio server, listening on all interfaces on the default port (8765). |
| 4589 | .TP |
| 4590 | 2) \fBfio \-\-server=ip:hostname,4444\fR |
| 4591 | Start a fio server, listening on IP belonging to hostname and on port 4444. |
| 4592 | .TP |
| 4593 | 3) \fBfio \-\-server=ip6:::1,4444\fR |
| 4594 | Start a fio server, listening on IPv6 localhost ::1 and on port 4444. |
| 4595 | .TP |
| 4596 | 4) \fBfio \-\-server=,4444\fR |
| 4597 | Start a fio server, listening on all interfaces on port 4444. |
| 4598 | .TP |
| 4599 | 5) \fBfio \-\-server=1.2.3.4\fR |
| 4600 | Start a fio server, listening on IP 1.2.3.4 on the default port. |
| 4601 | .TP |
| 4602 | 6) \fBfio \-\-server=sock:/tmp/fio.sock\fR |
| 4603 | Start a fio server, listening on the local socket `/tmp/fio.sock'. |
| 4604 | .RE |
| 4605 | .P |
| 4606 | Once a server is running, a "client" can connect to the fio server with: |
| 4607 | .RS |
| 4608 | .P |
| 4609 | $ fio <local\-args> \-\-client=<server> <remote\-args> <job file(s)> |
| 4610 | .RE |
| 4611 | .P |
| 4612 | where `local\-args' are arguments for the client where it is running, `server' |
| 4613 | is the connect string, and `remote\-args' and `job file(s)' are sent to the |
| 4614 | server. The `server' string follows the same format as it does on the server |
| 4615 | side, to allow IP/hostname/socket and port strings. |
| 4616 | .P |
| 4617 | Fio can connect to multiple servers this way: |
| 4618 | .RS |
| 4619 | .P |
| 4620 | $ fio \-\-client=<server1> <job file(s)> \-\-client=<server2> <job file(s)> |
| 4621 | .RE |
| 4622 | .P |
| 4623 | If the job file is located on the fio server, then you can tell the server to |
| 4624 | load a local file as well. This is done by using \fB\-\-remote\-config\fR: |
| 4625 | .RS |
| 4626 | .P |
| 4627 | $ fio \-\-client=server \-\-remote\-config /path/to/file.fio |
| 4628 | .RE |
| 4629 | .P |
| 4630 | Then fio will open this local (to the server) job file instead of being passed |
| 4631 | one from the client. |
| 4632 | .P |
| 4633 | If you have many servers (example: 100 VMs/containers), you can input a pathname |
| 4634 | of a file containing host IPs/names as the parameter value for the |
| 4635 | \fB\-\-client\fR option. For example, here is an example `host.list' |
| 4636 | file containing 2 hostnames: |
| 4637 | .RS |
| 4638 | .P |
| 4639 | .PD 0 |
| 4640 | host1.your.dns.domain |
| 4641 | .P |
| 4642 | host2.your.dns.domain |
| 4643 | .PD |
| 4644 | .RE |
| 4645 | .P |
| 4646 | The fio command would then be: |
| 4647 | .RS |
| 4648 | .P |
| 4649 | $ fio \-\-client=host.list <job file(s)> |
| 4650 | .RE |
| 4651 | .P |
| 4652 | In this mode, you cannot input server-specific parameters or job files \-\- all |
| 4653 | servers receive the same job file. |
| 4654 | .P |
| 4655 | In order to let `fio \-\-client' runs use a shared filesystem from multiple |
| 4656 | hosts, `fio \-\-client' now prepends the IP address of the server to the |
| 4657 | filename. For example, if fio is using the directory `/mnt/nfs/fio' and is |
| 4658 | writing filename `fileio.tmp', with a \fB\-\-client\fR `hostfile' |
| 4659 | containing two hostnames `h1' and `h2' with IP addresses 192.168.10.120 and |
| 4660 | 192.168.10.121, then fio will create two files: |
| 4661 | .RS |
| 4662 | .P |
| 4663 | .PD 0 |
| 4664 | /mnt/nfs/fio/192.168.10.120.fileio.tmp |
| 4665 | .P |
| 4666 | /mnt/nfs/fio/192.168.10.121.fileio.tmp |
| 4667 | .PD |
| 4668 | .RE |
| 4669 | .P |
| 4670 | Terse output in client/server mode will differ slightly from what is produced |
| 4671 | when fio is run in stand-alone mode. See the terse output section for details. |
| 4672 | .SH AUTHORS |
| 4673 | .B fio |
| 4674 | was written by Jens Axboe <axboe@kernel.dk>. |
| 4675 | .br |
| 4676 | This man page was written by Aaron Carroll <aaronc@cse.unsw.edu.au> based |
| 4677 | on documentation by Jens Axboe. |
| 4678 | .br |
| 4679 | This man page was rewritten by Tomohiro Kusumi <tkusumi@tuxera.com> based |
| 4680 | on documentation by Jens Axboe. |
| 4681 | .SH "REPORTING BUGS" |
| 4682 | Report bugs to the \fBfio\fR mailing list <fio@vger.kernel.org>. |
| 4683 | .br |
| 4684 | See \fBREPORTING\-BUGS\fR. |
| 4685 | .P |
| 4686 | \fBREPORTING\-BUGS\fR: \fIhttp://git.kernel.dk/cgit/fio/plain/REPORTING\-BUGS\fR |
| 4687 | .SH "SEE ALSO" |
| 4688 | For further documentation see \fBHOWTO\fR and \fBREADME\fR. |
| 4689 | .br |
| 4690 | Sample jobfiles are available in the `examples/' directory. |
| 4691 | .br |
| 4692 | These are typically located under `/usr/share/doc/fio'. |
| 4693 | .P |
| 4694 | \fBHOWTO\fR: \fIhttp://git.kernel.dk/cgit/fio/plain/HOWTO\fR |
| 4695 | .br |
| 4696 | \fBREADME\fR: \fIhttp://git.kernel.dk/cgit/fio/plain/README\fR |