| 1 | .TH fio 1 "December 2016" "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 of various fio actions. May be `all' for all types |
| 17 | or individual types separated by a comma (eg \-\-debug=io,file). `help' will |
| 18 | list all available tracing options. |
| 19 | .TP |
| 20 | .BI \-\-output \fR=\fPfilename |
| 21 | Write output to \fIfilename\fR. |
| 22 | .TP |
| 23 | .BI \-\-output-format \fR=\fPformat |
| 24 | Set the reporting format to \fInormal\fR, \fIterse\fR, \fIjson\fR, or |
| 25 | \fIjson+\fR. Multiple formats can be selected, separate by a comma. \fIterse\fR |
| 26 | is a CSV based format. \fIjson+\fR is like \fIjson\fR, except it adds a full |
| 27 | dump of the latency buckets. |
| 28 | .TP |
| 29 | .BI \-\-runtime \fR=\fPruntime |
| 30 | Limit run time to \fIruntime\fR seconds. |
| 31 | .TP |
| 32 | .B \-\-bandwidth\-log |
| 33 | Generate aggregate bandwidth logs. |
| 34 | .TP |
| 35 | .B \-\-minimal |
| 36 | Print statistics in a terse, semicolon-delimited format. |
| 37 | .TP |
| 38 | .B \-\-append-terse |
| 39 | Print statistics in selected mode AND terse, semicolon-delimited format. |
| 40 | Deprecated, use \-\-output-format instead to select multiple formats. |
| 41 | .TP |
| 42 | .B \-\-version |
| 43 | Display version information and exit. |
| 44 | .TP |
| 45 | .BI \-\-terse\-version \fR=\fPversion |
| 46 | Set terse version output format (Current version 3, or older version 2). |
| 47 | .TP |
| 48 | .B \-\-help |
| 49 | Display usage information and exit. |
| 50 | .TP |
| 51 | .B \-\-cpuclock-test |
| 52 | Perform test and validation of internal CPU clock |
| 53 | .TP |
| 54 | .BI \-\-crctest[\fR=\fPtest] |
| 55 | Test the speed of the builtin checksumming functions. If no argument is given, |
| 56 | all of them are tested. Or a comma separated list can be passed, in which |
| 57 | case the given ones are tested. |
| 58 | .TP |
| 59 | .BI \-\-cmdhelp \fR=\fPcommand |
| 60 | Print help information for \fIcommand\fR. May be `all' for all commands. |
| 61 | .TP |
| 62 | .BI \-\-enghelp \fR=\fPioengine[,command] |
| 63 | List all commands defined by \fIioengine\fR, or print help for \fIcommand\fR defined by \fIioengine\fR. |
| 64 | .TP |
| 65 | .BI \-\-showcmd \fR=\fPjobfile |
| 66 | Convert \fIjobfile\fR to a set of command-line options. |
| 67 | .TP |
| 68 | .BI \-\-eta \fR=\fPwhen |
| 69 | Specifies when real-time ETA estimate should be printed. \fIwhen\fR may |
| 70 | be one of `always', `never' or `auto'. |
| 71 | .TP |
| 72 | .BI \-\-eta\-newline \fR=\fPtime |
| 73 | Force an ETA newline for every `time` period passed. |
| 74 | .TP |
| 75 | .BI \-\-status\-interval \fR=\fPtime |
| 76 | Report full output status every `time` period passed. |
| 77 | .TP |
| 78 | .BI \-\-readonly |
| 79 | Turn on safety read-only checks, preventing any attempted write. |
| 80 | .TP |
| 81 | .BI \-\-section \fR=\fPsec |
| 82 | Only run section \fIsec\fR from job file. This option can be used multiple times to add more sections to run. |
| 83 | .TP |
| 84 | .BI \-\-alloc\-size \fR=\fPkb |
| 85 | Set the internal smalloc pool size to \fIkb\fP kilobytes. |
| 86 | .TP |
| 87 | .BI \-\-warnings\-fatal |
| 88 | All fio parser warnings are fatal, causing fio to exit with an error. |
| 89 | .TP |
| 90 | .BI \-\-max\-jobs \fR=\fPnr |
| 91 | Set the maximum allowed number of jobs (threads/processes) to support. |
| 92 | .TP |
| 93 | .BI \-\-server \fR=\fPargs |
| 94 | Start a backend server, with \fIargs\fP specifying what to listen to. See client/server section. |
| 95 | .TP |
| 96 | .BI \-\-daemonize \fR=\fPpidfile |
| 97 | Background a fio server, writing the pid to the given pid file. |
| 98 | .TP |
| 99 | .BI \-\-client \fR=\fPhost |
| 100 | Instead of running the jobs locally, send and run them on the given host or set of hosts. See client/server section. |
| 101 | .TP |
| 102 | .BI \-\-idle\-prof \fR=\fPoption |
| 103 | Report cpu idleness on a system or percpu basis (\fIoption\fP=system,percpu) or run unit work calibration only (\fIoption\fP=calibrate). |
| 104 | .SH "JOB FILE FORMAT" |
| 105 | Job files are in `ini' format. They consist of one or more |
| 106 | job definitions, which begin with a job name in square brackets and |
| 107 | extend to the next job name. The job name can be any ASCII string |
| 108 | except `global', which has a special meaning. Following the job name is |
| 109 | a sequence of zero or more parameters, one per line, that define the |
| 110 | behavior of the job. Any line starting with a `;' or `#' character is |
| 111 | considered a comment and ignored. |
| 112 | .P |
| 113 | If \fIjobfile\fR is specified as `-', the job file will be read from |
| 114 | standard input. |
| 115 | .SS "Global Section" |
| 116 | The global section contains default parameters for jobs specified in the |
| 117 | job file. A job is only affected by global sections residing above it, |
| 118 | and there may be any number of global sections. Specific job definitions |
| 119 | may override any parameter set in global sections. |
| 120 | .SH "JOB PARAMETERS" |
| 121 | .SS Types |
| 122 | Some parameters may take arguments of a specific type. |
| 123 | Anywhere a numeric value is required, an arithmetic expression may be used, |
| 124 | provided it is surrounded by parentheses. Supported operators are: |
| 125 | .RS |
| 126 | .RS |
| 127 | .TP |
| 128 | .B addition (+) |
| 129 | .TP |
| 130 | .B subtraction (-) |
| 131 | .TP |
| 132 | .B multiplication (*) |
| 133 | .TP |
| 134 | .B division (/) |
| 135 | .TP |
| 136 | .B modulus (%) |
| 137 | .TP |
| 138 | .B exponentiation (^) |
| 139 | .RE |
| 140 | .RE |
| 141 | .P |
| 142 | For time values in expressions, units are microseconds by default. This is |
| 143 | different than for time values not in expressions (not enclosed in |
| 144 | parentheses). The types used are: |
| 145 | .TP |
| 146 | .I str |
| 147 | String: a sequence of alphanumeric characters. |
| 148 | .TP |
| 149 | .I int |
| 150 | Integer. A whole number value, which may contain an integer prefix |
| 151 | and an integer suffix. |
| 152 | |
| 153 | [integer prefix]number[integer suffix] |
| 154 | |
| 155 | The optional integer prefix specifies the number's base. The default |
| 156 | is decimal. 0x specifies hexadecimal. |
| 157 | |
| 158 | The optional integer suffix specifies the number's units, and includes |
| 159 | an optional unit prefix and an optional unit. For quantities |
| 160 | of data, the default unit is bytes. For quantities of time, |
| 161 | the default unit is seconds. |
| 162 | |
| 163 | With \fBkb_base=1000\fR, fio follows international standards for unit prefixes. |
| 164 | To specify power-of-10 decimal values defined in the International |
| 165 | System of Units (SI): |
| 166 | .nf |
| 167 | ki means kilo (K) or 1000 |
| 168 | mi means mega (M) or 1000**2 |
| 169 | gi means giga (G) or 1000**3 |
| 170 | ti means tera (T) or 1000**4 |
| 171 | pi means peta (P) or 1000**5 |
| 172 | .fi |
| 173 | |
| 174 | To specify power-of-2 binary values defined in IEC 80000-13: |
| 175 | .nf |
| 176 | k means kibi (Ki) or 1024 |
| 177 | m means mebi (Mi) or 1024**2 |
| 178 | g means gibi (Gi) or 1024**3 |
| 179 | t means tebi (Ti) or 1024**4 |
| 180 | p means pebi (Pi) or 1024**5 |
| 181 | .fi |
| 182 | |
| 183 | With \fBkb_base=1024\fR (the default), the unit prefixes are opposite from |
| 184 | those specified in the SI and IEC 80000-13 standards to provide |
| 185 | compatibility with old scripts. For example, 4k means 4096. |
| 186 | |
| 187 | .nf |
| 188 | Examples with \fBkb_base=1000\fR: |
| 189 | 4 KiB: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB |
| 190 | 1 MiB: 1048576, 1m, 1024k |
| 191 | 1 MB: 1000000, 1mi, 1000ki |
| 192 | 1 TiB: 1073741824, 1t, 1024m, 1048576k |
| 193 | 1 TB: 1000000000, 1ti, 1000mi, 1000000ki |
| 194 | .fi |
| 195 | |
| 196 | .nf |
| 197 | Examples with \fBkb_base=1024\fR (default): |
| 198 | 4 KiB: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB |
| 199 | 1 MiB: 1048576, 1m, 1024k |
| 200 | 1 MB: 1000000, 1mi, 1000ki |
| 201 | 1 TiB: 1073741824, 1t, 1024m, 1048576k |
| 202 | 1 TB: 1000000000, 1ti, 1000mi, 1000000ki |
| 203 | .fi |
| 204 | |
| 205 | For quantities of data, an optional unit of 'B' may be included |
| 206 | (e.g., 'kb' is the same as 'k'). |
| 207 | |
| 208 | The integer suffix is not case sensitive (e.g., m/mi mean mebi/mega, |
| 209 | not milli). 'b' and 'B' both mean byte, not bit. |
| 210 | |
| 211 | To specify times (units are not case sensitive): |
| 212 | .nf |
| 213 | D means days |
| 214 | H means hours |
| 215 | M mean minutes |
| 216 | s or sec means seconds (default) |
| 217 | ms or msec means milliseconds |
| 218 | us or usec means microseconds |
| 219 | .fi |
| 220 | |
| 221 | .TP |
| 222 | .I bool |
| 223 | Boolean: a true or false value. `0' denotes false, `1' denotes true. |
| 224 | .TP |
| 225 | .I irange |
| 226 | Integer range: a range of integers specified in the format |
| 227 | \fIlower\fR:\fIupper\fR or \fIlower\fR\-\fIupper\fR. \fIlower\fR and |
| 228 | \fIupper\fR may contain a suffix as described above. If an option allows two |
| 229 | sets of ranges, they are separated with a `,' or `/' character. For example: |
| 230 | `8\-8k/8M\-4G'. |
| 231 | .TP |
| 232 | .I float_list |
| 233 | List of floating numbers: A list of floating numbers, separated by |
| 234 | a ':' character. |
| 235 | .SS "Parameter List" |
| 236 | .TP |
| 237 | .BI name \fR=\fPstr |
| 238 | May be used to override the job name. On the command line, this parameter |
| 239 | has the special purpose of signalling the start of a new job. |
| 240 | .TP |
| 241 | .BI wait_for \fR=\fPstr |
| 242 | Specifies the name of the already defined job to wait for. Single waitee name |
| 243 | only may be specified. If set, the job won't be started until all workers of |
| 244 | the waitee job are done. Wait_for operates on the job name basis, so there are |
| 245 | a few limitations. First, the waitee must be defined prior to the waiter job |
| 246 | (meaning no forward references). Second, if a job is being referenced as a |
| 247 | waitee, it must have a unique name (no duplicate waitees). |
| 248 | .TP |
| 249 | .BI description \fR=\fPstr |
| 250 | Human-readable description of the job. It is printed when the job is run, but |
| 251 | otherwise has no special purpose. |
| 252 | .TP |
| 253 | .BI directory \fR=\fPstr |
| 254 | Prefix filenames with this directory. Used to place files in a location other |
| 255 | than `./'. |
| 256 | You can specify a number of directories by separating the names with a ':' |
| 257 | character. These directories will be assigned equally distributed to job clones |
| 258 | creates with \fInumjobs\fR as long as they are using generated filenames. |
| 259 | If specific \fIfilename(s)\fR are set fio will use the first listed directory, |
| 260 | and thereby matching the \fIfilename\fR semantic which generates a file each |
| 261 | clone if not specified, but let all clones use the same if set. See |
| 262 | \fIfilename\fR for considerations regarding escaping certain characters on |
| 263 | some platforms. |
| 264 | .TP |
| 265 | .BI filename \fR=\fPstr |
| 266 | .B fio |
| 267 | normally makes up a file name based on the job name, thread number, and file |
| 268 | number. If you want to share files between threads in a job or several jobs, |
| 269 | specify a \fIfilename\fR for each of them to override the default. |
| 270 | If the I/O engine is file-based, you can specify |
| 271 | a number of files by separating the names with a `:' character. `\-' is a |
| 272 | reserved name, meaning stdin or stdout, depending on the read/write direction |
| 273 | set. On Windows, disk devices are accessed as \\.\PhysicalDrive0 for the first |
| 274 | device, \\.\PhysicalDrive1 for the second etc. Note: Windows and FreeBSD |
| 275 | prevent write access to areas of the disk containing in-use data |
| 276 | (e.g. filesystems). If the wanted filename does need to include a colon, then |
| 277 | escape that with a '\\' character. For instance, if the filename is |
| 278 | "/dev/dsk/foo@3,0:c", then you would use filename="/dev/dsk/foo@3,0\\:c". |
| 279 | .TP |
| 280 | .BI filename_format \fR=\fPstr |
| 281 | If sharing multiple files between jobs, it is usually necessary to have |
| 282 | fio generate the exact names that you want. By default, fio will name a file |
| 283 | based on the default file format specification of |
| 284 | \fBjobname.jobnumber.filenumber\fP. With this option, that can be |
| 285 | customized. Fio will recognize and replace the following keywords in this |
| 286 | string: |
| 287 | .RS |
| 288 | .RS |
| 289 | .TP |
| 290 | .B $jobname |
| 291 | The name of the worker thread or process. |
| 292 | .TP |
| 293 | .B $jobnum |
| 294 | The incremental number of the worker thread or process. |
| 295 | .TP |
| 296 | .B $filenum |
| 297 | The incremental number of the file for that worker thread or process. |
| 298 | .RE |
| 299 | .P |
| 300 | To have dependent jobs share a set of files, this option can be set to |
| 301 | have fio generate filenames that are shared between the two. For instance, |
| 302 | if \fBtestfiles.$filenum\fR is specified, file number 4 for any job will |
| 303 | be named \fBtestfiles.4\fR. The default of \fB$jobname.$jobnum.$filenum\fR |
| 304 | will be used if no other format specifier is given. |
| 305 | .RE |
| 306 | .P |
| 307 | .TP |
| 308 | .BI unique_filename \fR=\fPbool |
| 309 | To avoid collisions between networked clients, fio defaults to prefixing |
| 310 | any generated filenames (with a directory specified) with the source of |
| 311 | the client connecting. To disable this behavior, set this option to 0. |
| 312 | .TP |
| 313 | .BI lockfile \fR=\fPstr |
| 314 | Fio defaults to not locking any files before it does IO to them. If a file or |
| 315 | file descriptor is shared, fio can serialize IO to that file to make the end |
| 316 | result consistent. This is usual for emulating real workloads that share files. |
| 317 | The lock modes are: |
| 318 | .RS |
| 319 | .RS |
| 320 | .TP |
| 321 | .B none |
| 322 | No locking. This is the default. |
| 323 | .TP |
| 324 | .B exclusive |
| 325 | Only one thread or process may do IO at a time, excluding all others. |
| 326 | .TP |
| 327 | .B readwrite |
| 328 | Read-write locking on the file. Many readers may access the file at the same |
| 329 | time, but writes get exclusive access. |
| 330 | .RE |
| 331 | .RE |
| 332 | .P |
| 333 | .BI opendir \fR=\fPstr |
| 334 | Recursively open any files below directory \fIstr\fR. |
| 335 | .TP |
| 336 | .BI readwrite \fR=\fPstr "\fR,\fP rw" \fR=\fPstr |
| 337 | Type of I/O pattern. Accepted values are: |
| 338 | .RS |
| 339 | .RS |
| 340 | .TP |
| 341 | .B read |
| 342 | Sequential reads. |
| 343 | .TP |
| 344 | .B write |
| 345 | Sequential writes. |
| 346 | .TP |
| 347 | .B trim |
| 348 | Sequential trims (Linux block devices only). |
| 349 | .TP |
| 350 | .B randread |
| 351 | Random reads. |
| 352 | .TP |
| 353 | .B randwrite |
| 354 | Random writes. |
| 355 | .TP |
| 356 | .B randtrim |
| 357 | Random trims (Linux block devices only). |
| 358 | .TP |
| 359 | .B rw, readwrite |
| 360 | Mixed sequential reads and writes. |
| 361 | .TP |
| 362 | .B randrw |
| 363 | Mixed random reads and writes. |
| 364 | .TP |
| 365 | .B trimwrite |
| 366 | Sequential trim and write mixed workload. Blocks will be trimmed first, then |
| 367 | the same blocks will be written to. |
| 368 | .RE |
| 369 | .P |
| 370 | Fio defaults to read if the option is not specified. |
| 371 | For mixed I/O, the default split is 50/50. For certain types of io the result |
| 372 | may still be skewed a bit, since the speed may be different. It is possible to |
| 373 | specify a number of IO's to do before getting a new offset, this is done by |
| 374 | appending a `:\fI<nr>\fR to the end of the string given. For a random read, it |
| 375 | would look like \fBrw=randread:8\fR for passing in an offset modifier with a |
| 376 | value of 8. If the postfix is used with a sequential IO pattern, then the value |
| 377 | specified will be added to the generated offset for each IO. For instance, |
| 378 | using \fBrw=write:4k\fR will skip 4k for every write. It turns sequential IO |
| 379 | into sequential IO with holes. See the \fBrw_sequencer\fR option. |
| 380 | .RE |
| 381 | .TP |
| 382 | .BI rw_sequencer \fR=\fPstr |
| 383 | If an offset modifier is given by appending a number to the \fBrw=<str>\fR line, |
| 384 | then this option controls how that number modifies the IO offset being |
| 385 | generated. Accepted values are: |
| 386 | .RS |
| 387 | .RS |
| 388 | .TP |
| 389 | .B sequential |
| 390 | Generate sequential offset |
| 391 | .TP |
| 392 | .B identical |
| 393 | Generate the same offset |
| 394 | .RE |
| 395 | .P |
| 396 | \fBsequential\fR is only useful for random IO, where fio would normally |
| 397 | generate a new random offset for every IO. If you append eg 8 to randread, you |
| 398 | would get a new random offset for every 8 IO's. The result would be a seek for |
| 399 | only every 8 IO's, instead of for every IO. Use \fBrw=randread:8\fR to specify |
| 400 | that. As sequential IO is already sequential, setting \fBsequential\fR for that |
| 401 | would not result in any differences. \fBidentical\fR behaves in a similar |
| 402 | fashion, except it sends the same offset 8 number of times before generating a |
| 403 | new offset. |
| 404 | .RE |
| 405 | .P |
| 406 | .TP |
| 407 | .BI kb_base \fR=\fPint |
| 408 | The base unit for a kilobyte. The defacto base is 2^10, 1024. Storage |
| 409 | manufacturers like to use 10^3 or 1000 as a base ten unit instead, for obvious |
| 410 | reasons. Allowed values are 1024 or 1000, with 1024 being the default. |
| 411 | .TP |
| 412 | .BI unified_rw_reporting \fR=\fPbool |
| 413 | Fio normally reports statistics on a per data direction basis, meaning that |
| 414 | reads, writes, and trims are accounted and reported separately. If this option is |
| 415 | set fio sums the results and reports them as "mixed" instead. |
| 416 | .TP |
| 417 | .BI randrepeat \fR=\fPbool |
| 418 | Seed the random number generator used for random I/O patterns in a predictable |
| 419 | way so the pattern is repeatable across runs. Default: true. |
| 420 | .TP |
| 421 | .BI allrandrepeat \fR=\fPbool |
| 422 | Seed all random number generators in a predictable way so results are |
| 423 | repeatable across runs. Default: false. |
| 424 | .TP |
| 425 | .BI randseed \fR=\fPint |
| 426 | Seed the random number generators based on this seed value, to be able to |
| 427 | control what sequence of output is being generated. If not set, the random |
| 428 | sequence depends on the \fBrandrepeat\fR setting. |
| 429 | .TP |
| 430 | .BI fallocate \fR=\fPstr |
| 431 | Whether pre-allocation is performed when laying down files. Accepted values |
| 432 | are: |
| 433 | .RS |
| 434 | .RS |
| 435 | .TP |
| 436 | .B none |
| 437 | Do not pre-allocate space. |
| 438 | .TP |
| 439 | .B posix |
| 440 | Pre-allocate via \fBposix_fallocate\fR\|(3). |
| 441 | .TP |
| 442 | .B keep |
| 443 | Pre-allocate via \fBfallocate\fR\|(2) with FALLOC_FL_KEEP_SIZE set. |
| 444 | .TP |
| 445 | .B 0 |
| 446 | Backward-compatible alias for 'none'. |
| 447 | .TP |
| 448 | .B 1 |
| 449 | Backward-compatible alias for 'posix'. |
| 450 | .RE |
| 451 | .P |
| 452 | May not be available on all supported platforms. 'keep' is only |
| 453 | available on Linux. If using ZFS on Solaris this must be set to 'none' |
| 454 | because ZFS doesn't support it. Default: 'posix'. |
| 455 | .RE |
| 456 | .TP |
| 457 | .BI fadvise_hint \fR=\fPstr |
| 458 | Use \fBposix_fadvise\fR\|(2) to advise the kernel what I/O patterns |
| 459 | are likely to be issued. Accepted values are: |
| 460 | .RS |
| 461 | .RS |
| 462 | .TP |
| 463 | .B 0 |
| 464 | Backwards compatible hint for "no hint". |
| 465 | .TP |
| 466 | .B 1 |
| 467 | Backwards compatible hint for "advise with fio workload type". This |
| 468 | uses \fBFADV_RANDOM\fR for a random workload, and \fBFADV_SEQUENTIAL\fR |
| 469 | for a sequential workload. |
| 470 | .TP |
| 471 | .B sequential |
| 472 | Advise using \fBFADV_SEQUENTIAL\fR |
| 473 | .TP |
| 474 | .B random |
| 475 | Advise using \fBFADV_RANDOM\fR |
| 476 | .RE |
| 477 | .RE |
| 478 | .TP |
| 479 | .BI fadvise_stream \fR=\fPint |
| 480 | Use \fBposix_fadvise\fR\|(2) to advise the kernel what stream ID the |
| 481 | writes issued belong to. Only supported on Linux. Note, this option |
| 482 | may change going forward. |
| 483 | .TP |
| 484 | .BI size \fR=\fPint |
| 485 | Total size of I/O for this job. \fBfio\fR will run until this many bytes have |
| 486 | been transferred, unless limited by other options (\fBruntime\fR, for instance, |
| 487 | or increased/descreased by \fBio_size\fR). Unless \fBnrfiles\fR and |
| 488 | \fBfilesize\fR options are given, this amount will be divided between the |
| 489 | available files for the job. If not set, fio will use the full size of the |
| 490 | given files or devices. If the files do not exist, size must be given. It is |
| 491 | also possible to give size as a percentage between 1 and 100. If size=20% is |
| 492 | given, fio will use 20% of the full size of the given files or devices. |
| 493 | .TP |
| 494 | .BI io_size \fR=\fPint "\fR,\fB io_limit \fR=\fPint |
| 495 | Normally fio operates within the region set by \fBsize\fR, which means that |
| 496 | the \fBsize\fR option sets both the region and size of IO to be performed. |
| 497 | Sometimes that is not what you want. With this option, it is possible to |
| 498 | define just the amount of IO that fio should do. For instance, if \fBsize\fR |
| 499 | is set to 20G and \fBio_limit\fR is set to 5G, fio will perform IO within |
| 500 | the first 20G but exit when 5G have been done. The opposite is also |
| 501 | possible - if \fBsize\fR is set to 20G, and \fBio_size\fR is set to 40G, then |
| 502 | fio will do 40G of IO within the 0..20G region. |
| 503 | .TP |
| 504 | .BI fill_device \fR=\fPbool "\fR,\fB fill_fs" \fR=\fPbool |
| 505 | Sets size to something really large and waits for ENOSPC (no space left on |
| 506 | device) as the terminating condition. Only makes sense with sequential write. |
| 507 | For a read workload, the mount point will be filled first then IO started on |
| 508 | the result. This option doesn't make sense if operating on a raw device node, |
| 509 | since the size of that is already known by the file system. Additionally, |
| 510 | writing beyond end-of-device will not return ENOSPC there. |
| 511 | .TP |
| 512 | .BI filesize \fR=\fPirange |
| 513 | Individual file sizes. May be a range, in which case \fBfio\fR will select sizes |
| 514 | for files at random within the given range, limited to \fBsize\fR in total (if |
| 515 | that is given). If \fBfilesize\fR is not specified, each created file is the |
| 516 | same size. |
| 517 | .TP |
| 518 | .BI file_append \fR=\fPbool |
| 519 | Perform IO after the end of the file. Normally fio will operate within the |
| 520 | size of a file. If this option is set, then fio will append to the file |
| 521 | instead. This has identical behavior to setting \fRoffset\fP to the size |
| 522 | of a file. This option is ignored on non-regular files. |
| 523 | .TP |
| 524 | .BI blocksize \fR=\fPint[,int][,int] "\fR,\fB bs" \fR=\fPint[,int][,int] |
| 525 | The block size in bytes for I/O units. Default: 4096. |
| 526 | A single value applies to reads, writes, and trims. |
| 527 | Comma-separated values may be specified for reads, writes, and trims. |
| 528 | Empty values separated by commas use the default value. A value not |
| 529 | terminated in a comma applies to subsequent types. |
| 530 | .nf |
| 531 | Examples: |
| 532 | bs=256k means 256k for reads, writes and trims |
| 533 | bs=8k,32k means 8k for reads, 32k for writes and trims |
| 534 | bs=8k,32k, means 8k for reads, 32k for writes, and default for trims |
| 535 | bs=,8k means default for reads, 8k for writes and trims |
| 536 | bs=,8k, means default for reads, 8k for writes, and default for writes |
| 537 | .fi |
| 538 | .TP |
| 539 | .BI blocksize_range \fR=\fPirange[,irange][,irange] "\fR,\fB bsrange" \fR=\fPirange[,irange][,irange] |
| 540 | A range of block sizes in bytes for I/O units. |
| 541 | The issued I/O unit will always be a multiple of the minimum size, unless |
| 542 | \fBblocksize_unaligned\fR is set. |
| 543 | Comma-separated ranges may be specified for reads, writes, and trims |
| 544 | as described in \fBblocksize\fR. |
| 545 | .nf |
| 546 | Example: bsrange=1k-4k,2k-8k. |
| 547 | .fi |
| 548 | .TP |
| 549 | .BI bssplit \fR=\fPstr[,str][,str] |
| 550 | This option allows even finer grained control of the block sizes issued, |
| 551 | not just even splits between them. With this option, you can weight various |
| 552 | block sizes for exact control of the issued IO for a job that has mixed |
| 553 | block sizes. The format of the option is bssplit=blocksize/percentage, |
| 554 | optionally adding as many definitions as needed separated by a colon. |
| 555 | Example: bssplit=4k/10:64k/50:32k/40 would issue 50% 64k blocks, 10% 4k |
| 556 | blocks and 40% 32k blocks. \fBbssplit\fR also supports giving separate |
| 557 | splits to reads, writes, and trims. |
| 558 | Comma-separated values may be specified for reads, writes, and trims |
| 559 | as described in \fBblocksize\fR. |
| 560 | .TP |
| 561 | .B blocksize_unaligned\fR,\fB bs_unaligned |
| 562 | If set, fio will issue I/O units with any size within \fBblocksize_range\fR, |
| 563 | not just multiples of the minimum size. This typically won't |
| 564 | work with direct I/O, as that normally requires sector alignment. |
| 565 | .TP |
| 566 | .BI bs_is_seq_rand \fR=\fPbool |
| 567 | If this option is set, fio will use the normal read,write blocksize settings as |
| 568 | sequential,random blocksize settings instead. Any random read or write will |
| 569 | use the WRITE blocksize settings, and any sequential read or write will use |
| 570 | the READ blocksize settings. |
| 571 | .TP |
| 572 | .BI blockalign \fR=\fPint[,int][,int] "\fR,\fB ba" \fR=\fPint[,int][,int] |
| 573 | Boundary to which fio will align random I/O units. Default: \fBblocksize\fR. |
| 574 | Minimum alignment is typically 512b for using direct IO, though it usually |
| 575 | depends on the hardware block size. This option is mutually exclusive with |
| 576 | using a random map for files, so it will turn off that option. |
| 577 | Comma-separated values may be specified for reads, writes, and trims |
| 578 | as described in \fBblocksize\fR. |
| 579 | .TP |
| 580 | .B zero_buffers |
| 581 | Initialize buffers with all zeros. Default: fill buffers with random data. |
| 582 | .TP |
| 583 | .B refill_buffers |
| 584 | If this option is given, fio will refill the IO buffers on every submit. The |
| 585 | default is to only fill it at init time and reuse that data. Only makes sense |
| 586 | if zero_buffers isn't specified, naturally. If data verification is enabled, |
| 587 | refill_buffers is also automatically enabled. |
| 588 | .TP |
| 589 | .BI scramble_buffers \fR=\fPbool |
| 590 | If \fBrefill_buffers\fR is too costly and the target is using data |
| 591 | deduplication, then setting this option will slightly modify the IO buffer |
| 592 | contents to defeat normal de-dupe attempts. This is not enough to defeat |
| 593 | more clever block compression attempts, but it will stop naive dedupe |
| 594 | of blocks. Default: true. |
| 595 | .TP |
| 596 | .BI buffer_compress_percentage \fR=\fPint |
| 597 | If this is set, then fio will attempt to provide IO buffer content (on WRITEs) |
| 598 | that compress to the specified level. Fio does this by providing a mix of |
| 599 | random data and a fixed pattern. The fixed pattern is either zeroes, or the |
| 600 | pattern specified by \fBbuffer_pattern\fR. If the pattern option is used, it |
| 601 | might skew the compression ratio slightly. Note that this is per block size |
| 602 | unit, for file/disk wide compression level that matches this setting. Note |
| 603 | that this is per block size unit, for file/disk wide compression level that |
| 604 | matches this setting, you'll also want to set refill_buffers. |
| 605 | .TP |
| 606 | .BI buffer_compress_chunk \fR=\fPint |
| 607 | See \fBbuffer_compress_percentage\fR. This setting allows fio to manage how |
| 608 | big the ranges of random data and zeroed data is. Without this set, fio will |
| 609 | provide \fBbuffer_compress_percentage\fR of blocksize random data, followed by |
| 610 | the remaining zeroed. With this set to some chunk size smaller than the block |
| 611 | size, fio can alternate random and zeroed data throughout the IO buffer. |
| 612 | .TP |
| 613 | .BI buffer_pattern \fR=\fPstr |
| 614 | If set, fio will fill the IO buffers with this pattern. If not set, the contents |
| 615 | of IO buffers is defined by the other options related to buffer contents. The |
| 616 | setting can be any pattern of bytes, and can be prefixed with 0x for hex |
| 617 | values. It may also be a string, where the string must then be wrapped with |
| 618 | "", e.g.: |
| 619 | .RS |
| 620 | .RS |
| 621 | \fBbuffer_pattern\fR="abcd" |
| 622 | .RS |
| 623 | or |
| 624 | .RE |
| 625 | \fBbuffer_pattern\fR=-12 |
| 626 | .RS |
| 627 | or |
| 628 | .RE |
| 629 | \fBbuffer_pattern\fR=0xdeadface |
| 630 | .RE |
| 631 | .LP |
| 632 | Also you can combine everything together in any order: |
| 633 | .LP |
| 634 | .RS |
| 635 | \fBbuffer_pattern\fR=0xdeadface"abcd"-12 |
| 636 | .RE |
| 637 | .RE |
| 638 | .TP |
| 639 | .BI dedupe_percentage \fR=\fPint |
| 640 | If set, fio will generate this percentage of identical buffers when writing. |
| 641 | These buffers will be naturally dedupable. The contents of the buffers depend |
| 642 | on what other buffer compression settings have been set. It's possible to have |
| 643 | the individual buffers either fully compressible, or not at all. This option |
| 644 | only controls the distribution of unique buffers. |
| 645 | .TP |
| 646 | .BI nrfiles \fR=\fPint |
| 647 | Number of files to use for this job. Default: 1. |
| 648 | .TP |
| 649 | .BI openfiles \fR=\fPint |
| 650 | Number of files to keep open at the same time. Default: \fBnrfiles\fR. |
| 651 | .TP |
| 652 | .BI file_service_type \fR=\fPstr |
| 653 | Defines how files to service are selected. The following types are defined: |
| 654 | .RS |
| 655 | .RS |
| 656 | .TP |
| 657 | .B random |
| 658 | Choose a file at random. |
| 659 | .TP |
| 660 | .B roundrobin |
| 661 | Round robin over opened files (default). |
| 662 | .TP |
| 663 | .B sequential |
| 664 | Do each file in the set sequentially. |
| 665 | .TP |
| 666 | .B zipf |
| 667 | Use a zipfian distribution to decide what file to access. |
| 668 | .TP |
| 669 | .B pareto |
| 670 | Use a pareto distribution to decide what file to access. |
| 671 | .TP |
| 672 | .B gauss |
| 673 | Use a gaussian (normal) distribution to decide what file to access. |
| 674 | .RE |
| 675 | .P |
| 676 | For \fBrandom\fR, \fBroundrobin\fR, and \fBsequential\fR, a postfix can be |
| 677 | appended to tell fio how many I/Os to issue before switching to a new file. |
| 678 | For example, specifying \fBfile_service_type=random:8\fR would cause fio to |
| 679 | issue \fI8\fR I/Os before selecting a new file at random. For the non-uniform |
| 680 | distributions, a floating point postfix can be given to influence how the |
| 681 | distribution is skewed. See \fBrandom_distribution\fR for a description of how |
| 682 | that would work. |
| 683 | .RE |
| 684 | .TP |
| 685 | .BI ioengine \fR=\fPstr |
| 686 | Defines how the job issues I/O. The following types are defined: |
| 687 | .RS |
| 688 | .RS |
| 689 | .TP |
| 690 | .B sync |
| 691 | Basic \fBread\fR\|(2) or \fBwrite\fR\|(2) I/O. \fBfseek\fR\|(2) is used to |
| 692 | position the I/O location. |
| 693 | .TP |
| 694 | .B psync |
| 695 | Basic \fBpread\fR\|(2) or \fBpwrite\fR\|(2) I/O. |
| 696 | Default on all supported operating systems except for Windows. |
| 697 | .TP |
| 698 | .B vsync |
| 699 | Basic \fBreadv\fR\|(2) or \fBwritev\fR\|(2) I/O. Will emulate queuing by |
| 700 | coalescing adjacent IOs into a single submission. |
| 701 | .TP |
| 702 | .B pvsync |
| 703 | Basic \fBpreadv\fR\|(2) or \fBpwritev\fR\|(2) I/O. |
| 704 | .TP |
| 705 | .B pvsync2 |
| 706 | Basic \fBpreadv2\fR\|(2) or \fBpwritev2\fR\|(2) I/O. |
| 707 | .TP |
| 708 | .B libaio |
| 709 | Linux native asynchronous I/O. This ioengine defines engine specific options. |
| 710 | .TP |
| 711 | .B posixaio |
| 712 | POSIX asynchronous I/O using \fBaio_read\fR\|(3) and \fBaio_write\fR\|(3). |
| 713 | .TP |
| 714 | .B solarisaio |
| 715 | Solaris native asynchronous I/O. |
| 716 | .TP |
| 717 | .B windowsaio |
| 718 | Windows native asynchronous I/O. Default on Windows. |
| 719 | .TP |
| 720 | .B mmap |
| 721 | File is memory mapped with \fBmmap\fR\|(2) and data copied using |
| 722 | \fBmemcpy\fR\|(3). |
| 723 | .TP |
| 724 | .B splice |
| 725 | \fBsplice\fR\|(2) is used to transfer the data and \fBvmsplice\fR\|(2) to |
| 726 | transfer data from user-space to the kernel. |
| 727 | .TP |
| 728 | .B sg |
| 729 | SCSI generic sg v3 I/O. May be either synchronous using the SG_IO ioctl, or if |
| 730 | the target is an sg character device, we use \fBread\fR\|(2) and |
| 731 | \fBwrite\fR\|(2) for asynchronous I/O. |
| 732 | .TP |
| 733 | .B null |
| 734 | Doesn't transfer any data, just pretends to. Mainly used to exercise \fBfio\fR |
| 735 | itself and for debugging and testing purposes. |
| 736 | .TP |
| 737 | .B net |
| 738 | Transfer over the network. The protocol to be used can be defined with the |
| 739 | \fBprotocol\fR parameter. Depending on the protocol, \fBfilename\fR, |
| 740 | \fBhostname\fR, \fBport\fR, or \fBlisten\fR must be specified. |
| 741 | This ioengine defines engine specific options. |
| 742 | .TP |
| 743 | .B netsplice |
| 744 | Like \fBnet\fR, but uses \fBsplice\fR\|(2) and \fBvmsplice\fR\|(2) to map data |
| 745 | and send/receive. This ioengine defines engine specific options. |
| 746 | .TP |
| 747 | .B cpuio |
| 748 | Doesn't transfer any data, but burns CPU cycles according to \fBcpuload\fR and |
| 749 | \fBcpuchunks\fR parameters. A job never finishes unless there is at least one |
| 750 | non-cpuio job. |
| 751 | .TP |
| 752 | .B guasi |
| 753 | The GUASI I/O engine is the Generic Userspace Asynchronous Syscall Interface |
| 754 | approach to asynchronous I/O. |
| 755 | .br |
| 756 | See <http://www.xmailserver.org/guasi\-lib.html>. |
| 757 | .TP |
| 758 | .B rdma |
| 759 | The RDMA I/O engine supports both RDMA memory semantics (RDMA_WRITE/RDMA_READ) |
| 760 | and channel semantics (Send/Recv) for the InfiniBand, RoCE and iWARP protocols. |
| 761 | .TP |
| 762 | .B external |
| 763 | Loads an external I/O engine object file. Append the engine filename as |
| 764 | `:\fIenginepath\fR'. |
| 765 | .TP |
| 766 | .B falloc |
| 767 | IO engine that does regular linux native fallocate call to simulate data |
| 768 | transfer as fio ioengine |
| 769 | .br |
| 770 | DDIR_READ does fallocate(,mode = FALLOC_FL_KEEP_SIZE,) |
| 771 | .br |
| 772 | DIR_WRITE does fallocate(,mode = 0) |
| 773 | .br |
| 774 | DDIR_TRIM does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE) |
| 775 | .TP |
| 776 | .B e4defrag |
| 777 | IO engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate defragment activity |
| 778 | request to DDIR_WRITE event |
| 779 | .TP |
| 780 | .B rbd |
| 781 | IO engine supporting direct access to Ceph Rados Block Devices (RBD) via librbd |
| 782 | without the need to use the kernel rbd driver. This ioengine defines engine specific |
| 783 | options. |
| 784 | .TP |
| 785 | .B gfapi |
| 786 | Using Glusterfs libgfapi sync interface to direct access to Glusterfs volumes without |
| 787 | having to go through FUSE. This ioengine defines engine specific |
| 788 | options. |
| 789 | .TP |
| 790 | .B gfapi_async |
| 791 | Using Glusterfs libgfapi async interface to direct access to Glusterfs volumes without |
| 792 | having to go through FUSE. This ioengine defines engine specific |
| 793 | options. |
| 794 | .TP |
| 795 | .B libhdfs |
| 796 | Read and write through Hadoop (HDFS). The \fBfilename\fR option is used to |
| 797 | specify host,port of the hdfs name-node to connect. This engine interprets |
| 798 | offsets a little differently. In HDFS, files once created cannot be modified. |
| 799 | So random writes are not possible. To imitate this, libhdfs engine expects |
| 800 | bunch of small files to be created over HDFS, and engine will randomly pick a |
| 801 | file out of those files based on the offset generated by fio backend. (see the |
| 802 | example job file to create such files, use rw=write option). Please note, you |
| 803 | might want to set necessary environment variables to work with hdfs/libhdfs |
| 804 | properly. |
| 805 | .TP |
| 806 | .B mtd |
| 807 | Read, write and erase an MTD character device (e.g., /dev/mtd0). Discards are |
| 808 | treated as erases. Depending on the underlying device type, the I/O may have |
| 809 | to go in a certain pattern, e.g., on NAND, writing sequentially to erase blocks |
| 810 | and discarding before overwriting. The trimwrite mode works well for this |
| 811 | constraint. |
| 812 | .TP |
| 813 | .B pmemblk |
| 814 | Read and write using filesystem DAX to a file on a filesystem mounted with |
| 815 | DAX on a persistent memory device through the NVML libpmemblk library. |
| 816 | .TP |
| 817 | .B dev-dax |
| 818 | Read and write using device DAX to a persistent memory device |
| 819 | (e.g., /dev/dax0.0) through the NVML libpmem library. |
| 820 | .RE |
| 821 | .P |
| 822 | .RE |
| 823 | .TP |
| 824 | .BI iodepth \fR=\fPint |
| 825 | Number of I/O units to keep in flight against the file. Note that increasing |
| 826 | iodepth beyond 1 will not affect synchronous ioengines (except for small |
| 827 | degress when verify_async is in use). Even async engines may impose OS |
| 828 | restrictions causing the desired depth not to be achieved. This may happen on |
| 829 | Linux when using libaio and not setting \fBdirect\fR=1, since buffered IO is |
| 830 | not async on that OS. Keep an eye on the IO depth distribution in the |
| 831 | fio output to verify that the achieved depth is as expected. Default: 1. |
| 832 | .TP |
| 833 | .BI iodepth_batch \fR=\fPint "\fR,\fP iodepth_batch_submit" \fR=\fPint |
| 834 | This defines how many pieces of IO to submit at once. It defaults to 1 |
| 835 | which means that we submit each IO as soon as it is available, but can |
| 836 | be raised to submit bigger batches of IO at the time. If it is set to 0 |
| 837 | the \fBiodepth\fR value will be used. |
| 838 | .TP |
| 839 | .BI iodepth_batch_complete_min \fR=\fPint "\fR,\fP iodepth_batch_complete" \fR=\fPint |
| 840 | This defines how many pieces of IO to retrieve at once. It defaults to 1 which |
| 841 | means that we'll ask for a minimum of 1 IO in the retrieval process from the |
| 842 | kernel. The IO retrieval will go on until we hit the limit set by |
| 843 | \fBiodepth_low\fR. If this variable is set to 0, then fio will always check for |
| 844 | completed events before queuing more IO. This helps reduce IO latency, at the |
| 845 | cost of more retrieval system calls. |
| 846 | .TP |
| 847 | .BI iodepth_batch_complete_max \fR=\fPint |
| 848 | This defines maximum pieces of IO to |
| 849 | retrieve at once. This variable should be used along with |
| 850 | \fBiodepth_batch_complete_min\fR=int variable, specifying the range |
| 851 | of min and max amount of IO which should be retrieved. By default |
| 852 | it is equal to \fBiodepth_batch_complete_min\fR value. |
| 853 | |
| 854 | Example #1: |
| 855 | .RS |
| 856 | .RS |
| 857 | \fBiodepth_batch_complete_min\fR=1 |
| 858 | .LP |
| 859 | \fBiodepth_batch_complete_max\fR=<iodepth> |
| 860 | .RE |
| 861 | |
| 862 | which means that we will retrieve at least 1 IO and up to the |
| 863 | whole submitted queue depth. If none of IO has been completed |
| 864 | yet, we will wait. |
| 865 | |
| 866 | Example #2: |
| 867 | .RS |
| 868 | \fBiodepth_batch_complete_min\fR=0 |
| 869 | .LP |
| 870 | \fBiodepth_batch_complete_max\fR=<iodepth> |
| 871 | .RE |
| 872 | |
| 873 | which means that we can retrieve up to the whole submitted |
| 874 | queue depth, but if none of IO has been completed yet, we will |
| 875 | NOT wait and immediately exit the system call. In this example |
| 876 | we simply do polling. |
| 877 | .RE |
| 878 | .TP |
| 879 | .BI iodepth_low \fR=\fPint |
| 880 | Low watermark indicating when to start filling the queue again. Default: |
| 881 | \fBiodepth\fR. |
| 882 | .TP |
| 883 | .BI io_submit_mode \fR=\fPstr |
| 884 | This option controls how fio submits the IO to the IO engine. The default is |
| 885 | \fBinline\fR, which means that the fio job threads submit and reap IO directly. |
| 886 | If set to \fBoffload\fR, the job threads will offload IO submission to a |
| 887 | dedicated pool of IO threads. This requires some coordination and thus has a |
| 888 | bit of extra overhead, especially for lower queue depth IO where it can |
| 889 | increase latencies. The benefit is that fio can manage submission rates |
| 890 | independently of the device completion rates. This avoids skewed latency |
| 891 | reporting if IO gets back up on the device side (the coordinated omission |
| 892 | problem). |
| 893 | .TP |
| 894 | .BI direct \fR=\fPbool |
| 895 | If true, use non-buffered I/O (usually O_DIRECT). Default: false. |
| 896 | .TP |
| 897 | .BI atomic \fR=\fPbool |
| 898 | If value is true, attempt to use atomic direct IO. Atomic writes are guaranteed |
| 899 | to be stable once acknowledged by the operating system. Only Linux supports |
| 900 | O_ATOMIC right now. |
| 901 | .TP |
| 902 | .BI buffered \fR=\fPbool |
| 903 | If true, use buffered I/O. This is the opposite of the \fBdirect\fR parameter. |
| 904 | Default: true. |
| 905 | .TP |
| 906 | .BI offset \fR=\fPint |
| 907 | Offset in the file to start I/O. Data before the offset will not be touched. |
| 908 | .TP |
| 909 | .BI offset_increment \fR=\fPint |
| 910 | If this is provided, then the real offset becomes the |
| 911 | offset + offset_increment * thread_number, where the thread number is a |
| 912 | counter that starts at 0 and is incremented for each sub-job (i.e. when |
| 913 | numjobs option is specified). This option is useful if there are several jobs |
| 914 | which are intended to operate on a file in parallel disjoint segments, with |
| 915 | even spacing between the starting points. |
| 916 | .TP |
| 917 | .BI number_ios \fR=\fPint |
| 918 | Fio will normally perform IOs until it has exhausted the size of the region |
| 919 | set by \fBsize\fR, or if it exhaust the allocated time (or hits an error |
| 920 | condition). With this setting, the range/size can be set independently of |
| 921 | the number of IOs to perform. When fio reaches this number, it will exit |
| 922 | normally and report status. Note that this does not extend the amount |
| 923 | of IO that will be done, it will only stop fio if this condition is met |
| 924 | before other end-of-job criteria. |
| 925 | .TP |
| 926 | .BI fsync \fR=\fPint |
| 927 | How many I/Os to perform before issuing an \fBfsync\fR\|(2) of dirty data. If |
| 928 | 0, don't sync. Default: 0. |
| 929 | .TP |
| 930 | .BI fdatasync \fR=\fPint |
| 931 | Like \fBfsync\fR, but uses \fBfdatasync\fR\|(2) instead to only sync the |
| 932 | data parts of the file. Default: 0. |
| 933 | .TP |
| 934 | .BI write_barrier \fR=\fPint |
| 935 | Make every Nth write a barrier write. |
| 936 | .TP |
| 937 | .BI sync_file_range \fR=\fPstr:int |
| 938 | Use \fBsync_file_range\fR\|(2) for every \fRval\fP number of write operations. Fio will |
| 939 | track range of writes that have happened since the last \fBsync_file_range\fR\|(2) call. |
| 940 | \fRstr\fP can currently be one or more of: |
| 941 | .RS |
| 942 | .TP |
| 943 | .B wait_before |
| 944 | SYNC_FILE_RANGE_WAIT_BEFORE |
| 945 | .TP |
| 946 | .B write |
| 947 | SYNC_FILE_RANGE_WRITE |
| 948 | .TP |
| 949 | .B wait_after |
| 950 | SYNC_FILE_RANGE_WRITE |
| 951 | .TP |
| 952 | .RE |
| 953 | .P |
| 954 | So if you do sync_file_range=wait_before,write:8, fio would use |
| 955 | \fBSYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE\fP for every 8 writes. |
| 956 | Also see the \fBsync_file_range\fR\|(2) man page. This option is Linux specific. |
| 957 | .TP |
| 958 | .BI overwrite \fR=\fPbool |
| 959 | If writing, setup the file first and do overwrites. Default: false. |
| 960 | .TP |
| 961 | .BI end_fsync \fR=\fPbool |
| 962 | Sync file contents when a write stage has completed. Default: false. |
| 963 | .TP |
| 964 | .BI fsync_on_close \fR=\fPbool |
| 965 | If true, sync file contents on close. This differs from \fBend_fsync\fR in that |
| 966 | it will happen on every close, not just at the end of the job. Default: false. |
| 967 | .TP |
| 968 | .BI rwmixread \fR=\fPint |
| 969 | Percentage of a mixed workload that should be reads. Default: 50. |
| 970 | .TP |
| 971 | .BI rwmixwrite \fR=\fPint |
| 972 | Percentage of a mixed workload that should be writes. If \fBrwmixread\fR and |
| 973 | \fBrwmixwrite\fR are given and do not sum to 100%, the latter of the two |
| 974 | overrides the first. This may interfere with a given rate setting, if fio is |
| 975 | asked to limit reads or writes to a certain rate. If that is the case, then |
| 976 | the distribution may be skewed. Default: 50. |
| 977 | .TP |
| 978 | .BI random_distribution \fR=\fPstr:float |
| 979 | By default, fio will use a completely uniform random distribution when asked |
| 980 | to perform random IO. Sometimes it is useful to skew the distribution in |
| 981 | specific ways, ensuring that some parts of the data is more hot than others. |
| 982 | Fio includes the following distribution models: |
| 983 | .RS |
| 984 | .TP |
| 985 | .B random |
| 986 | Uniform random distribution |
| 987 | .TP |
| 988 | .B zipf |
| 989 | Zipf distribution |
| 990 | .TP |
| 991 | .B pareto |
| 992 | Pareto distribution |
| 993 | .TP |
| 994 | .B gauss |
| 995 | Normal (gaussian) distribution |
| 996 | .TP |
| 997 | .B zoned |
| 998 | Zoned random distribution |
| 999 | .TP |
| 1000 | .RE |
| 1001 | When using a \fBzipf\fR or \fBpareto\fR distribution, an input value is also |
| 1002 | needed to define the access pattern. For \fBzipf\fR, this is the zipf theta. |
| 1003 | For \fBpareto\fR, it's the pareto power. Fio includes a test program, genzipf, |
| 1004 | that can be used visualize what the given input values will yield in terms of |
| 1005 | hit rates. If you wanted to use \fBzipf\fR with a theta of 1.2, you would use |
| 1006 | random_distribution=zipf:1.2 as the option. If a non-uniform model is used, |
| 1007 | fio will disable use of the random map. For the \fBgauss\fR distribution, a |
| 1008 | normal deviation is supplied as a value between 0 and 100. |
| 1009 | .P |
| 1010 | .RS |
| 1011 | For a \fBzoned\fR distribution, fio supports specifying percentages of IO |
| 1012 | access that should fall within what range of the file or device. For example, |
| 1013 | given a criteria of: |
| 1014 | .P |
| 1015 | .RS |
| 1016 | 60% of accesses should be to the first 10% |
| 1017 | .RE |
| 1018 | .RS |
| 1019 | 30% of accesses should be to the next 20% |
| 1020 | .RE |
| 1021 | .RS |
| 1022 | 8% of accesses should be to to the next 30% |
| 1023 | .RE |
| 1024 | .RS |
| 1025 | 2% of accesses should be to the next 40% |
| 1026 | .RE |
| 1027 | .P |
| 1028 | we can define that through zoning of the random accesses. For the above |
| 1029 | example, the user would do: |
| 1030 | .P |
| 1031 | .RS |
| 1032 | .B random_distribution=zoned:60/10:30/20:8/30:2/40 |
| 1033 | .RE |
| 1034 | .P |
| 1035 | similarly to how \fBbssplit\fR works for setting ranges and percentages of block |
| 1036 | sizes. Like \fBbssplit\fR, it's possible to specify separate zones for reads, |
| 1037 | writes, and trims. If just one set is given, it'll apply to all of them. |
| 1038 | .RE |
| 1039 | .TP |
| 1040 | .BI percentage_random \fR=\fPint[,int][,int] |
| 1041 | For a random workload, set how big a percentage should be random. This defaults |
| 1042 | to 100%, in which case the workload is fully random. It can be set from |
| 1043 | anywhere from 0 to 100. Setting it to 0 would make the workload fully |
| 1044 | sequential. It is possible to set different values for reads, writes, and |
| 1045 | trim. To do so, simply use a comma separated list. See \fBblocksize\fR. |
| 1046 | .TP |
| 1047 | .B norandommap |
| 1048 | Normally \fBfio\fR will cover every block of the file when doing random I/O. If |
| 1049 | this parameter is given, a new offset will be chosen without looking at past |
| 1050 | I/O history. This parameter is mutually exclusive with \fBverify\fR. |
| 1051 | .TP |
| 1052 | .BI softrandommap \fR=\fPbool |
| 1053 | See \fBnorandommap\fR. If fio runs with the random block map enabled and it |
| 1054 | fails to allocate the map, if this option is set it will continue without a |
| 1055 | random block map. As coverage will not be as complete as with random maps, this |
| 1056 | option is disabled by default. |
| 1057 | .TP |
| 1058 | .BI random_generator \fR=\fPstr |
| 1059 | Fio supports the following engines for generating IO offsets for random IO: |
| 1060 | .RS |
| 1061 | .TP |
| 1062 | .B tausworthe |
| 1063 | Strong 2^88 cycle random number generator |
| 1064 | .TP |
| 1065 | .B lfsr |
| 1066 | Linear feedback shift register generator |
| 1067 | .TP |
| 1068 | .B tausworthe64 |
| 1069 | Strong 64-bit 2^258 cycle random number generator |
| 1070 | .TP |
| 1071 | .RE |
| 1072 | .P |
| 1073 | Tausworthe is a strong random number generator, but it requires tracking on the |
| 1074 | side if we want to ensure that blocks are only read or written once. LFSR |
| 1075 | guarantees that we never generate the same offset twice, and it's also less |
| 1076 | computationally expensive. It's not a true random generator, however, though |
| 1077 | for IO purposes it's typically good enough. LFSR only works with single block |
| 1078 | sizes, not with workloads that use multiple block sizes. If used with such a |
| 1079 | workload, fio may read or write some blocks multiple times. The default |
| 1080 | value is tausworthe, unless the required space exceeds 2^32 blocks. If it does, |
| 1081 | then tausworthe64 is selected automatically. |
| 1082 | .TP |
| 1083 | .BI nice \fR=\fPint |
| 1084 | Run job with given nice value. See \fBnice\fR\|(2). |
| 1085 | .TP |
| 1086 | .BI prio \fR=\fPint |
| 1087 | Set I/O priority value of this job between 0 (highest) and 7 (lowest). See |
| 1088 | \fBionice\fR\|(1). |
| 1089 | .TP |
| 1090 | .BI prioclass \fR=\fPint |
| 1091 | Set I/O priority class. See \fBionice\fR\|(1). |
| 1092 | .TP |
| 1093 | .BI thinktime \fR=\fPint |
| 1094 | Stall job for given number of microseconds between issuing I/Os. |
| 1095 | .TP |
| 1096 | .BI thinktime_spin \fR=\fPint |
| 1097 | Pretend to spend CPU time for given number of microseconds, sleeping the rest |
| 1098 | of the time specified by \fBthinktime\fR. Only valid if \fBthinktime\fR is set. |
| 1099 | .TP |
| 1100 | .BI thinktime_blocks \fR=\fPint |
| 1101 | Only valid if thinktime is set - control how many blocks to issue, before |
| 1102 | waiting \fBthinktime\fR microseconds. If not set, defaults to 1 which will |
| 1103 | make fio wait \fBthinktime\fR microseconds after every block. This |
| 1104 | effectively makes any queue depth setting redundant, since no more than 1 IO |
| 1105 | will be queued before we have to complete it and do our thinktime. In other |
| 1106 | words, this setting effectively caps the queue depth if the latter is larger. |
| 1107 | Default: 1. |
| 1108 | .TP |
| 1109 | .BI rate \fR=\fPint[,int][,int] |
| 1110 | Cap bandwidth used by this job. The number is in bytes/sec, the normal postfix |
| 1111 | rules apply. You can use \fBrate\fR=500k to limit reads and writes to 500k each, |
| 1112 | or you can specify reads, write, and trim limits separately. |
| 1113 | Using \fBrate\fR=1m,500k would |
| 1114 | limit reads to 1MiB/sec and writes to 500KiB/sec. Capping only reads or writes |
| 1115 | can be done with \fBrate\fR=,500k or \fBrate\fR=500k,. The former will only |
| 1116 | limit writes (to 500KiB/sec), the latter will only limit reads. |
| 1117 | .TP |
| 1118 | .BI rate_min \fR=\fPint[,int][,int] |
| 1119 | Tell \fBfio\fR to do whatever it can to maintain at least the given bandwidth. |
| 1120 | Failing to meet this requirement will cause the job to exit. The same format |
| 1121 | as \fBrate\fR is used for read vs write vs trim separation. |
| 1122 | .TP |
| 1123 | .BI rate_iops \fR=\fPint[,int][,int] |
| 1124 | Cap the bandwidth to this number of IOPS. Basically the same as rate, just |
| 1125 | specified independently of bandwidth. The same format as \fBrate\fR is used for |
| 1126 | read vs write vs trim separation. If \fBblocksize\fR is a range, the smallest block |
| 1127 | size is used as the metric. |
| 1128 | .TP |
| 1129 | .BI rate_iops_min \fR=\fPint[,int][,int] |
| 1130 | If this rate of I/O is not met, the job will exit. The same format as \fBrate\fR |
| 1131 | is used for read vs write vs trim separation. |
| 1132 | .TP |
| 1133 | .BI rate_process \fR=\fPstr |
| 1134 | This option controls how fio manages rated IO submissions. The default is |
| 1135 | \fBlinear\fR, which submits IO in a linear fashion with fixed delays between |
| 1136 | IOs that gets adjusted based on IO completion rates. If this is set to |
| 1137 | \fBpoisson\fR, fio will submit IO based on a more real world random request |
| 1138 | flow, known as the Poisson process |
| 1139 | (https://en.wikipedia.org/wiki/Poisson_process). The lambda will be |
| 1140 | 10^6 / IOPS for the given workload. |
| 1141 | .TP |
| 1142 | .BI rate_cycle \fR=\fPint |
| 1143 | Average bandwidth for \fBrate\fR and \fBrate_min\fR over this number of |
| 1144 | milliseconds. Default: 1000ms. |
| 1145 | .TP |
| 1146 | .BI latency_target \fR=\fPint |
| 1147 | If set, fio will attempt to find the max performance point that the given |
| 1148 | workload will run at while maintaining a latency below this target. The |
| 1149 | values is given in microseconds. See \fBlatency_window\fR and |
| 1150 | \fBlatency_percentile\fR. |
| 1151 | .TP |
| 1152 | .BI latency_window \fR=\fPint |
| 1153 | Used with \fBlatency_target\fR to specify the sample window that the job |
| 1154 | is run at varying queue depths to test the performance. The value is given |
| 1155 | in microseconds. |
| 1156 | .TP |
| 1157 | .BI latency_percentile \fR=\fPfloat |
| 1158 | The percentage of IOs that must fall within the criteria specified by |
| 1159 | \fBlatency_target\fR and \fBlatency_window\fR. If not set, this defaults |
| 1160 | to 100.0, meaning that all IOs must be equal or below to the value set |
| 1161 | by \fBlatency_target\fR. |
| 1162 | .TP |
| 1163 | .BI max_latency \fR=\fPint |
| 1164 | If set, fio will exit the job if it exceeds this maximum latency. It will exit |
| 1165 | with an ETIME error. |
| 1166 | .TP |
| 1167 | .BI cpumask \fR=\fPint |
| 1168 | Set CPU affinity for this job. \fIint\fR is a bitmask of allowed CPUs the job |
| 1169 | may run on. See \fBsched_setaffinity\fR\|(2). |
| 1170 | .TP |
| 1171 | .BI cpus_allowed \fR=\fPstr |
| 1172 | Same as \fBcpumask\fR, but allows a comma-delimited list of CPU numbers. |
| 1173 | .TP |
| 1174 | .BI cpus_allowed_policy \fR=\fPstr |
| 1175 | Set the policy of how fio distributes the CPUs specified by \fBcpus_allowed\fR |
| 1176 | or \fBcpumask\fR. Two policies are supported: |
| 1177 | .RS |
| 1178 | .RS |
| 1179 | .TP |
| 1180 | .B shared |
| 1181 | All jobs will share the CPU set specified. |
| 1182 | .TP |
| 1183 | .B split |
| 1184 | Each job will get a unique CPU from the CPU set. |
| 1185 | .RE |
| 1186 | .P |
| 1187 | \fBshared\fR is the default behaviour, if the option isn't specified. If |
| 1188 | \fBsplit\fR is specified, then fio will assign one cpu per job. If not enough |
| 1189 | CPUs are given for the jobs listed, then fio will roundrobin the CPUs in |
| 1190 | the set. |
| 1191 | .RE |
| 1192 | .P |
| 1193 | .TP |
| 1194 | .BI numa_cpu_nodes \fR=\fPstr |
| 1195 | Set this job running on specified NUMA nodes' CPUs. The arguments allow |
| 1196 | comma delimited list of cpu numbers, A-B ranges, or 'all'. |
| 1197 | .TP |
| 1198 | .BI numa_mem_policy \fR=\fPstr |
| 1199 | Set this job's memory policy and corresponding NUMA nodes. Format of |
| 1200 | the arguments: |
| 1201 | .RS |
| 1202 | .TP |
| 1203 | .B <mode>[:<nodelist>] |
| 1204 | .TP |
| 1205 | .B mode |
| 1206 | is one of the following memory policy: |
| 1207 | .TP |
| 1208 | .B default, prefer, bind, interleave, local |
| 1209 | .TP |
| 1210 | .RE |
| 1211 | For \fBdefault\fR and \fBlocal\fR memory policy, no \fBnodelist\fR is |
| 1212 | needed to be specified. For \fBprefer\fR, only one node is |
| 1213 | allowed. For \fBbind\fR and \fBinterleave\fR, \fBnodelist\fR allows |
| 1214 | comma delimited list of numbers, A-B ranges, or 'all'. |
| 1215 | .TP |
| 1216 | .BI startdelay \fR=\fPirange |
| 1217 | Delay start of job for the specified number of seconds. Supports all time |
| 1218 | suffixes to allow specification of hours, minutes, seconds and |
| 1219 | milliseconds - seconds are the default if a unit is omitted. |
| 1220 | Can be given as a range which causes each thread to choose randomly out of the |
| 1221 | range. |
| 1222 | .TP |
| 1223 | .BI runtime \fR=\fPint |
| 1224 | Terminate processing after the specified number of seconds. |
| 1225 | .TP |
| 1226 | .B time_based |
| 1227 | If given, run for the specified \fBruntime\fR duration even if the files are |
| 1228 | completely read or written. The same workload will be repeated as many times |
| 1229 | as \fBruntime\fR allows. |
| 1230 | .TP |
| 1231 | .BI ramp_time \fR=\fPint |
| 1232 | If set, fio will run the specified workload for this amount of time before |
| 1233 | logging any performance numbers. Useful for letting performance settle before |
| 1234 | logging results, thus minimizing the runtime required for stable results. Note |
| 1235 | that the \fBramp_time\fR is considered lead in time for a job, thus it will |
| 1236 | increase the total runtime if a special timeout or runtime is specified. |
| 1237 | .TP |
| 1238 | .BI steadystate \fR=\fPstr:float "\fR,\fP ss" \fR=\fPstr:float |
| 1239 | Define the criterion and limit for assessing steady state performance. The |
| 1240 | first parameter designates the criterion whereas the second parameter sets the |
| 1241 | threshold. When the criterion falls below the threshold for the specified |
| 1242 | duration, the job will stop. For example, iops_slope:0.1% will direct fio |
| 1243 | to terminate the job when the least squares regression slope falls below 0.1% |
| 1244 | of the mean IOPS. If group_reporting is enabled this will apply to all jobs in |
| 1245 | the group. All assessments are carried out using only data from the rolling |
| 1246 | collection window. Threshold limits can be expressed as a fixed value or as a |
| 1247 | percentage of the mean in the collection window. Below are the available steady |
| 1248 | state assessment criteria. |
| 1249 | .RS |
| 1250 | .RS |
| 1251 | .TP |
| 1252 | .B iops |
| 1253 | Collect IOPS data. Stop the job if all individual IOPS measurements are within |
| 1254 | the specified limit of the mean IOPS (e.g., iops:2 means that all individual |
| 1255 | IOPS values must be within 2 of the mean, whereas iops:0.2% means that all |
| 1256 | individual IOPS values must be within 0.2% of the mean IOPS to terminate the |
| 1257 | job). |
| 1258 | .TP |
| 1259 | .B iops_slope |
| 1260 | Collect IOPS data and calculate the least squares regression slope. Stop the |
| 1261 | job if the slope falls below the specified limit. |
| 1262 | .TP |
| 1263 | .B bw |
| 1264 | Collect bandwidth data. Stop the job if all individual bandwidth measurements |
| 1265 | are within the specified limit of the mean bandwidth. |
| 1266 | .TP |
| 1267 | .B bw_slope |
| 1268 | Collect bandwidth data and calculate the least squares regression slope. Stop |
| 1269 | the job if the slope falls below the specified limit. |
| 1270 | .RE |
| 1271 | .RE |
| 1272 | .TP |
| 1273 | .BI steadystate_duration \fR=\fPtime "\fR,\fP ss_dur" \fR=\fPtime |
| 1274 | A rolling window of this duration will be used to judge whether steady state |
| 1275 | has been reached. Data will be collected once per second. The default is 0 |
| 1276 | which disables steady state detection. |
| 1277 | .TP |
| 1278 | .BI steadystate_ramp_time \fR=\fPtime "\fR,\fP ss_ramp" \fR=\fPtime |
| 1279 | Allow the job to run for the specified duration before beginning data collection |
| 1280 | for checking the steady state job termination criterion. The default is 0. |
| 1281 | .TP |
| 1282 | .BI invalidate \fR=\fPbool |
| 1283 | Invalidate buffer-cache for the file prior to starting I/O. Default: true. |
| 1284 | .TP |
| 1285 | .BI sync \fR=\fPbool |
| 1286 | Use synchronous I/O for buffered writes. For the majority of I/O engines, |
| 1287 | this means using O_SYNC. Default: false. |
| 1288 | .TP |
| 1289 | .BI iomem \fR=\fPstr "\fR,\fP mem" \fR=\fPstr |
| 1290 | Allocation method for I/O unit buffer. Allowed values are: |
| 1291 | .RS |
| 1292 | .RS |
| 1293 | .TP |
| 1294 | .B malloc |
| 1295 | Allocate memory with \fBmalloc\fR\|(3). Default memory type. |
| 1296 | .TP |
| 1297 | .B shm |
| 1298 | Use shared memory buffers allocated through \fBshmget\fR\|(2). |
| 1299 | .TP |
| 1300 | .B shmhuge |
| 1301 | Same as \fBshm\fR, but use huge pages as backing. |
| 1302 | .TP |
| 1303 | .B mmap |
| 1304 | Use \fBmmap\fR\|(2) for allocation. Uses anonymous memory unless a filename |
| 1305 | is given after the option in the format `:\fIfile\fR'. |
| 1306 | .TP |
| 1307 | .B mmaphuge |
| 1308 | Same as \fBmmap\fR, but use huge files as backing. |
| 1309 | .TP |
| 1310 | .B mmapshared |
| 1311 | Same as \fBmmap\fR, but use a MMAP_SHARED mapping. |
| 1312 | .RE |
| 1313 | .P |
| 1314 | The amount of memory allocated is the maximum allowed \fBblocksize\fR for the |
| 1315 | job multiplied by \fBiodepth\fR. For \fBshmhuge\fR or \fBmmaphuge\fR to work, |
| 1316 | the system must have free huge pages allocated. \fBmmaphuge\fR also needs to |
| 1317 | have hugetlbfs mounted, and \fIfile\fR must point there. At least on Linux, |
| 1318 | huge pages must be manually allocated. See \fB/proc/sys/vm/nr_hugehages\fR |
| 1319 | and the documentation for that. Normally you just need to echo an appropriate |
| 1320 | number, eg echoing 8 will ensure that the OS has 8 huge pages ready for |
| 1321 | use. |
| 1322 | .RE |
| 1323 | .TP |
| 1324 | .BI iomem_align \fR=\fPint "\fR,\fP mem_align" \fR=\fPint |
| 1325 | This indicates the memory alignment of the IO memory buffers. Note that the |
| 1326 | given alignment is applied to the first IO unit buffer, if using \fBiodepth\fR |
| 1327 | the alignment of the following buffers are given by the \fBbs\fR used. In |
| 1328 | other words, if using a \fBbs\fR that is a multiple of the page sized in the |
| 1329 | system, all buffers will be aligned to this value. If using a \fBbs\fR that |
| 1330 | is not page aligned, the alignment of subsequent IO memory buffers is the |
| 1331 | sum of the \fBiomem_align\fR and \fBbs\fR used. |
| 1332 | .TP |
| 1333 | .BI hugepage\-size \fR=\fPint |
| 1334 | Defines the size of a huge page. Must be at least equal to the system setting. |
| 1335 | Should be a multiple of 1MiB. Default: 4MiB. |
| 1336 | .TP |
| 1337 | .B exitall |
| 1338 | Terminate all jobs when one finishes. Default: wait for each job to finish. |
| 1339 | .TP |
| 1340 | .B exitall_on_error \fR=\fPbool |
| 1341 | Terminate all jobs if one job finishes in error. Default: wait for each job |
| 1342 | to finish. |
| 1343 | .TP |
| 1344 | .BI bwavgtime \fR=\fPint |
| 1345 | Average bandwidth calculations over the given time in milliseconds. If the job |
| 1346 | also does bandwidth logging through \fBwrite_bw_log\fR, then the minimum of |
| 1347 | this option and \fBlog_avg_msec\fR will be used. Default: 500ms. |
| 1348 | .TP |
| 1349 | .BI iopsavgtime \fR=\fPint |
| 1350 | Average IOPS calculations over the given time in milliseconds. If the job |
| 1351 | also does IOPS logging through \fBwrite_iops_log\fR, then the minimum of |
| 1352 | this option and \fBlog_avg_msec\fR will be used. Default: 500ms. |
| 1353 | .TP |
| 1354 | .BI create_serialize \fR=\fPbool |
| 1355 | If true, serialize file creation for the jobs. Default: true. |
| 1356 | .TP |
| 1357 | .BI create_fsync \fR=\fPbool |
| 1358 | \fBfsync\fR\|(2) data file after creation. Default: true. |
| 1359 | .TP |
| 1360 | .BI create_on_open \fR=\fPbool |
| 1361 | If true, the files are not created until they are opened for IO by the job. |
| 1362 | .TP |
| 1363 | .BI create_only \fR=\fPbool |
| 1364 | If true, fio will only run the setup phase of the job. If files need to be |
| 1365 | laid out or updated on disk, only that will be done. The actual job contents |
| 1366 | are not executed. |
| 1367 | .TP |
| 1368 | .BI allow_file_create \fR=\fPbool |
| 1369 | If true, fio is permitted to create files as part of its workload. This is |
| 1370 | the default behavior. If this option is false, then fio will error out if the |
| 1371 | files it needs to use don't already exist. Default: true. |
| 1372 | .TP |
| 1373 | .BI allow_mounted_write \fR=\fPbool |
| 1374 | If this isn't set, fio will abort jobs that are destructive (eg that write) |
| 1375 | to what appears to be a mounted device or partition. This should help catch |
| 1376 | creating inadvertently destructive tests, not realizing that the test will |
| 1377 | destroy data on the mounted file system. Default: false. |
| 1378 | .TP |
| 1379 | .BI pre_read \fR=\fPbool |
| 1380 | If this is given, files will be pre-read into memory before starting the given |
| 1381 | IO operation. This will also clear the \fR \fBinvalidate\fR flag, since it is |
| 1382 | pointless to pre-read and then drop the cache. This will only work for IO |
| 1383 | engines that are seekable, since they allow you to read the same data |
| 1384 | multiple times. Thus it will not work on eg network or splice IO. |
| 1385 | .TP |
| 1386 | .BI unlink \fR=\fPbool |
| 1387 | Unlink job files when done. Default: false. |
| 1388 | .TP |
| 1389 | .BI unlink_each_loop \fR=\fPbool |
| 1390 | Unlink job files after each iteration or loop. Default: false. |
| 1391 | .TP |
| 1392 | .BI loops \fR=\fPint |
| 1393 | Specifies the number of iterations (runs of the same workload) of this job. |
| 1394 | Default: 1. |
| 1395 | .TP |
| 1396 | .BI verify_only \fR=\fPbool |
| 1397 | Do not perform the specified workload, only verify data still matches previous |
| 1398 | invocation of this workload. This option allows one to check data multiple |
| 1399 | times at a later date without overwriting it. This option makes sense only for |
| 1400 | workloads that write data, and does not support workloads with the |
| 1401 | \fBtime_based\fR option set. |
| 1402 | .TP |
| 1403 | .BI do_verify \fR=\fPbool |
| 1404 | Run the verify phase after a write phase. Only valid if \fBverify\fR is set. |
| 1405 | Default: true. |
| 1406 | .TP |
| 1407 | .BI verify \fR=\fPstr |
| 1408 | Method of verifying file contents after each iteration of the job. Each |
| 1409 | verification method also implies verification of special header, which is |
| 1410 | written to the beginning of each block. This header also includes meta |
| 1411 | information, like offset of the block, block number, timestamp when block |
| 1412 | was written, etc. \fBverify\fR=str can be combined with \fBverify_pattern\fR=str |
| 1413 | option. The allowed values are: |
| 1414 | .RS |
| 1415 | .RS |
| 1416 | .TP |
| 1417 | .B md5 crc16 crc32 crc32c crc32c-intel crc64 crc7 sha256 sha512 sha1 sha3-224 sha3-256 sha3-384 sha3-512 xxhash |
| 1418 | Store appropriate checksum in the header of each block. crc32c-intel is |
| 1419 | hardware accelerated SSE4.2 driven, falls back to regular crc32c if |
| 1420 | not supported by the system. |
| 1421 | .TP |
| 1422 | .B meta |
| 1423 | This option is deprecated, since now meta information is included in generic |
| 1424 | verification header and meta verification happens by default. For detailed |
| 1425 | information see the description of the \fBverify\fR=str setting. This option |
| 1426 | is kept because of compatibility's sake with old configurations. Do not use it. |
| 1427 | .TP |
| 1428 | .B pattern |
| 1429 | Verify a strict pattern. Normally fio includes a header with some basic |
| 1430 | information and checksumming, but if this option is set, only the |
| 1431 | specific pattern set with \fBverify_pattern\fR is verified. |
| 1432 | .TP |
| 1433 | .B null |
| 1434 | Pretend to verify. Used for testing internals. |
| 1435 | .RE |
| 1436 | |
| 1437 | This option can be used for repeated burn-in tests of a system to make sure |
| 1438 | that the written data is also correctly read back. If the data direction given |
| 1439 | is a read or random read, fio will assume that it should verify a previously |
| 1440 | written file. If the data direction includes any form of write, the verify will |
| 1441 | be of the newly written data. |
| 1442 | .RE |
| 1443 | .TP |
| 1444 | .BI verifysort \fR=\fPbool |
| 1445 | If true, written verify blocks are sorted if \fBfio\fR deems it to be faster to |
| 1446 | read them back in a sorted manner. Default: true. |
| 1447 | .TP |
| 1448 | .BI verifysort_nr \fR=\fPint |
| 1449 | Pre-load and sort verify blocks for a read workload. |
| 1450 | .TP |
| 1451 | .BI verify_offset \fR=\fPint |
| 1452 | Swap the verification header with data somewhere else in the block before |
| 1453 | writing. It is swapped back before verifying. |
| 1454 | .TP |
| 1455 | .BI verify_interval \fR=\fPint |
| 1456 | Write the verification header for this number of bytes, which should divide |
| 1457 | \fBblocksize\fR. Default: \fBblocksize\fR. |
| 1458 | .TP |
| 1459 | .BI verify_pattern \fR=\fPstr |
| 1460 | If set, fio will fill the io buffers with this pattern. Fio defaults to filling |
| 1461 | with totally random bytes, but sometimes it's interesting to fill with a known |
| 1462 | pattern for io verification purposes. Depending on the width of the pattern, |
| 1463 | fio will fill 1/2/3/4 bytes of the buffer at the time(it can be either a |
| 1464 | decimal or a hex number). The verify_pattern if larger than a 32-bit quantity |
| 1465 | has to be a hex number that starts with either "0x" or "0X". Use with |
| 1466 | \fBverify\fP=str. Also, verify_pattern supports %o format, which means that for |
| 1467 | each block offset will be written and then verified back, e.g.: |
| 1468 | .RS |
| 1469 | .RS |
| 1470 | \fBverify_pattern\fR=%o |
| 1471 | .RE |
| 1472 | Or use combination of everything: |
| 1473 | .LP |
| 1474 | .RS |
| 1475 | \fBverify_pattern\fR=0xff%o"abcd"-21 |
| 1476 | .RE |
| 1477 | .RE |
| 1478 | .TP |
| 1479 | .BI verify_fatal \fR=\fPbool |
| 1480 | If true, exit the job on the first observed verification failure. Default: |
| 1481 | false. |
| 1482 | .TP |
| 1483 | .BI verify_dump \fR=\fPbool |
| 1484 | If set, dump the contents of both the original data block and the data block we |
| 1485 | read off disk to files. This allows later analysis to inspect just what kind of |
| 1486 | data corruption occurred. Off by default. |
| 1487 | .TP |
| 1488 | .BI verify_async \fR=\fPint |
| 1489 | Fio will normally verify IO inline from the submitting thread. This option |
| 1490 | takes an integer describing how many async offload threads to create for IO |
| 1491 | verification instead, causing fio to offload the duty of verifying IO contents |
| 1492 | to one or more separate threads. If using this offload option, even sync IO |
| 1493 | engines can benefit from using an \fBiodepth\fR setting higher than 1, as it |
| 1494 | allows them to have IO in flight while verifies are running. |
| 1495 | .TP |
| 1496 | .BI verify_async_cpus \fR=\fPstr |
| 1497 | Tell fio to set the given CPU affinity on the async IO verification threads. |
| 1498 | See \fBcpus_allowed\fP for the format used. |
| 1499 | .TP |
| 1500 | .BI verify_backlog \fR=\fPint |
| 1501 | Fio will normally verify the written contents of a job that utilizes verify |
| 1502 | once that job has completed. In other words, everything is written then |
| 1503 | everything is read back and verified. You may want to verify continually |
| 1504 | instead for a variety of reasons. Fio stores the meta data associated with an |
| 1505 | IO block in memory, so for large verify workloads, quite a bit of memory would |
| 1506 | be used up holding this meta data. If this option is enabled, fio will write |
| 1507 | only N blocks before verifying these blocks. |
| 1508 | .TP |
| 1509 | .BI verify_backlog_batch \fR=\fPint |
| 1510 | Control how many blocks fio will verify if verify_backlog is set. If not set, |
| 1511 | will default to the value of \fBverify_backlog\fR (meaning the entire queue is |
| 1512 | read back and verified). If \fBverify_backlog_batch\fR is less than |
| 1513 | \fBverify_backlog\fR then not all blocks will be verified, if |
| 1514 | \fBverify_backlog_batch\fR is larger than \fBverify_backlog\fR, some blocks |
| 1515 | will be verified more than once. |
| 1516 | .TP |
| 1517 | .BI trim_percentage \fR=\fPint |
| 1518 | Number of verify blocks to discard/trim. |
| 1519 | .TP |
| 1520 | .BI trim_verify_zero \fR=\fPbool |
| 1521 | Verify that trim/discarded blocks are returned as zeroes. |
| 1522 | .TP |
| 1523 | .BI trim_backlog \fR=\fPint |
| 1524 | Trim after this number of blocks are written. |
| 1525 | .TP |
| 1526 | .BI trim_backlog_batch \fR=\fPint |
| 1527 | Trim this number of IO blocks. |
| 1528 | .TP |
| 1529 | .BI experimental_verify \fR=\fPbool |
| 1530 | Enable experimental verification. |
| 1531 | .TP |
| 1532 | .BI verify_state_save \fR=\fPbool |
| 1533 | When a job exits during the write phase of a verify workload, save its |
| 1534 | current state. This allows fio to replay up until that point, if the |
| 1535 | verify state is loaded for the verify read phase. |
| 1536 | .TP |
| 1537 | .BI verify_state_load \fR=\fPbool |
| 1538 | If a verify termination trigger was used, fio stores the current write |
| 1539 | state of each thread. This can be used at verification time so that fio |
| 1540 | knows how far it should verify. Without this information, fio will run |
| 1541 | a full verification pass, according to the settings in the job file used. |
| 1542 | .TP |
| 1543 | .B stonewall "\fR,\fP wait_for_previous" |
| 1544 | Wait for preceding jobs in the job file to exit before starting this one. |
| 1545 | \fBstonewall\fR implies \fBnew_group\fR. |
| 1546 | .TP |
| 1547 | .B new_group |
| 1548 | Start a new reporting group. If not given, all jobs in a file will be part |
| 1549 | of the same reporting group, unless separated by a stonewall. |
| 1550 | .TP |
| 1551 | .BI numjobs \fR=\fPint |
| 1552 | Number of clones (processes/threads performing the same workload) of this job. |
| 1553 | Default: 1. |
| 1554 | .TP |
| 1555 | .B group_reporting |
| 1556 | If set, display per-group reports instead of per-job when \fBnumjobs\fR is |
| 1557 | specified. |
| 1558 | .TP |
| 1559 | .B thread |
| 1560 | Use threads created with \fBpthread_create\fR\|(3) instead of processes created |
| 1561 | with \fBfork\fR\|(2). |
| 1562 | .TP |
| 1563 | .BI zonesize \fR=\fPint |
| 1564 | Divide file into zones of the specified size in bytes. See \fBzoneskip\fR. |
| 1565 | .TP |
| 1566 | .BI zonerange \fR=\fPint |
| 1567 | Give size of an IO zone. See \fBzoneskip\fR. |
| 1568 | .TP |
| 1569 | .BI zoneskip \fR=\fPint |
| 1570 | Skip the specified number of bytes when \fBzonesize\fR bytes of data have been |
| 1571 | read. |
| 1572 | .TP |
| 1573 | .BI write_iolog \fR=\fPstr |
| 1574 | Write the issued I/O patterns to the specified file. Specify a separate file |
| 1575 | for each job, otherwise the iologs will be interspersed and the file may be |
| 1576 | corrupt. |
| 1577 | .TP |
| 1578 | .BI read_iolog \fR=\fPstr |
| 1579 | Replay the I/O patterns contained in the specified file generated by |
| 1580 | \fBwrite_iolog\fR, or may be a \fBblktrace\fR binary file. |
| 1581 | .TP |
| 1582 | .BI replay_no_stall \fR=\fPint |
| 1583 | While replaying I/O patterns using \fBread_iolog\fR the default behavior |
| 1584 | attempts to respect timing information between I/Os. Enabling |
| 1585 | \fBreplay_no_stall\fR causes I/Os to be replayed as fast as possible while |
| 1586 | still respecting ordering. |
| 1587 | .TP |
| 1588 | .BI replay_redirect \fR=\fPstr |
| 1589 | While replaying I/O patterns using \fBread_iolog\fR the default behavior |
| 1590 | is to replay the IOPS onto the major/minor device that each IOP was recorded |
| 1591 | from. Setting \fBreplay_redirect\fR causes all IOPS to be replayed onto the |
| 1592 | single specified device regardless of the device it was recorded from. |
| 1593 | .TP |
| 1594 | .BI replay_align \fR=\fPint |
| 1595 | Force alignment of IO offsets and lengths in a trace to this power of 2 value. |
| 1596 | .TP |
| 1597 | .BI replay_scale \fR=\fPint |
| 1598 | Scale sector offsets down by this factor when replaying traces. |
| 1599 | .TP |
| 1600 | .BI per_job_logs \fR=\fPbool |
| 1601 | If set, this generates bw/clat/iops log with per file private filenames. If |
| 1602 | not set, jobs with identical names will share the log filename. Default: true. |
| 1603 | .TP |
| 1604 | .BI write_bw_log \fR=\fPstr |
| 1605 | If given, write a bandwidth log for this job. Can be used to store data of the |
| 1606 | bandwidth of the jobs in their lifetime. The included fio_generate_plots script |
| 1607 | uses gnuplot to turn these text files into nice graphs. See \fBwrite_lat_log\fR |
| 1608 | for behaviour of given filename. For this option, the postfix is _bw.x.log, |
| 1609 | where x is the index of the job (1..N, where N is the number of jobs). If |
| 1610 | \fBper_job_logs\fR is false, then the filename will not include the job index. |
| 1611 | See the \fBLOG FILE FORMATS\fR |
| 1612 | section. |
| 1613 | .TP |
| 1614 | .BI write_lat_log \fR=\fPstr |
| 1615 | Same as \fBwrite_bw_log\fR, but writes I/O completion latencies. If no |
| 1616 | filename is given with this option, the default filename of |
| 1617 | "jobname_type.x.log" is used, where x is the index of the job (1..N, where |
| 1618 | N is the number of jobs). Even if the filename is given, fio will still |
| 1619 | append the type of log. If \fBper_job_logs\fR is false, then the filename will |
| 1620 | not include the job index. See the \fBLOG FILE FORMATS\fR section. |
| 1621 | .TP |
| 1622 | .BI write_hist_log \fR=\fPstr |
| 1623 | Same as \fBwrite_lat_log\fR, but writes I/O completion latency histograms. If |
| 1624 | no filename is given with this option, the default filename of |
| 1625 | "jobname_clat_hist.x.log" is used, where x is the index of the job (1..N, where |
| 1626 | N is the number of jobs). Even if the filename is given, fio will still append |
| 1627 | the type of log. If \fBper_job_logs\fR is false, then the filename will not |
| 1628 | include the job index. See the \fBLOG FILE FORMATS\fR section. |
| 1629 | .TP |
| 1630 | .BI write_iops_log \fR=\fPstr |
| 1631 | Same as \fBwrite_bw_log\fR, but writes IOPS. If no filename is given with this |
| 1632 | option, the default filename of "jobname_type.x.log" is used, where x is the |
| 1633 | index of the job (1..N, where N is the number of jobs). Even if the filename |
| 1634 | is given, fio will still append the type of log. If \fBper_job_logs\fR is false, |
| 1635 | then the filename will not include the job index. See the \fBLOG FILE FORMATS\fR |
| 1636 | section. |
| 1637 | .TP |
| 1638 | .BI log_avg_msec \fR=\fPint |
| 1639 | By default, fio will log an entry in the iops, latency, or bw log for every |
| 1640 | IO that completes. When writing to the disk log, that can quickly grow to a |
| 1641 | very large size. Setting this option makes fio average the each log entry |
| 1642 | over the specified period of time, reducing the resolution of the log. See |
| 1643 | \fBlog_max_value\fR as well. Defaults to 0, logging all entries. |
| 1644 | .TP |
| 1645 | .BI log_max_value \fR=\fPbool |
| 1646 | If \fBlog_avg_msec\fR is set, fio logs the average over that window. If you |
| 1647 | instead want to log the maximum value, set this option to 1. Defaults to |
| 1648 | 0, meaning that averaged values are logged. |
| 1649 | .TP |
| 1650 | .BI log_hist_msec \fR=\fPint |
| 1651 | Same as \fBlog_avg_msec\fR, but logs entries for completion latency histograms. |
| 1652 | Computing latency percentiles from averages of intervals using \fBlog_avg_msec\fR |
| 1653 | is innacurate. Setting this option makes fio log histogram entries over the |
| 1654 | specified period of time, reducing log sizes for high IOPS devices while |
| 1655 | retaining percentile accuracy. See \fBlog_hist_coarseness\fR as well. Defaults |
| 1656 | to 0, meaning histogram logging is disabled. |
| 1657 | .TP |
| 1658 | .BI log_hist_coarseness \fR=\fPint |
| 1659 | Integer ranging from 0 to 6, defining the coarseness of the resolution of the |
| 1660 | histogram logs enabled with \fBlog_hist_msec\fR. For each increment in |
| 1661 | coarseness, fio outputs half as many bins. Defaults to 0, for which histogram |
| 1662 | logs contain 1216 latency bins. See the \fBLOG FILE FORMATS\fR section. |
| 1663 | .TP |
| 1664 | .BI log_offset \fR=\fPbool |
| 1665 | If this is set, the iolog options will include the byte offset for the IO |
| 1666 | entry as well as the other data values. |
| 1667 | .TP |
| 1668 | .BI log_compression \fR=\fPint |
| 1669 | If this is set, fio will compress the IO logs as it goes, to keep the memory |
| 1670 | footprint lower. When a log reaches the specified size, that chunk is removed |
| 1671 | and compressed in the background. Given that IO logs are fairly highly |
| 1672 | compressible, this yields a nice memory savings for longer runs. The downside |
| 1673 | is that the compression will consume some background CPU cycles, so it may |
| 1674 | impact the run. This, however, is also true if the logging ends up consuming |
| 1675 | most of the system memory. So pick your poison. The IO logs are saved |
| 1676 | normally at the end of a run, by decompressing the chunks and storing them |
| 1677 | in the specified log file. This feature depends on the availability of zlib. |
| 1678 | .TP |
| 1679 | .BI log_compression_cpus \fR=\fPstr |
| 1680 | Define the set of CPUs that are allowed to handle online log compression |
| 1681 | for the IO jobs. This can provide better isolation between performance |
| 1682 | sensitive jobs, and background compression work. |
| 1683 | .TP |
| 1684 | .BI log_store_compressed \fR=\fPbool |
| 1685 | If set, fio will store the log files in a compressed format. They can be |
| 1686 | decompressed with fio, using the \fB\-\-inflate-log\fR command line parameter. |
| 1687 | The files will be stored with a \fB\.fz\fR suffix. |
| 1688 | .TP |
| 1689 | .BI log_unix_epoch \fR=\fPbool |
| 1690 | If set, fio will log Unix timestamps to the log files produced by enabling |
| 1691 | \fBwrite_type_log\fR for each log type, instead of the default zero-based |
| 1692 | timestamps. |
| 1693 | .TP |
| 1694 | .BI block_error_percentiles \fR=\fPbool |
| 1695 | If set, record errors in trim block-sized units from writes and trims and output |
| 1696 | a histogram of how many trims it took to get to errors, and what kind of error |
| 1697 | was encountered. |
| 1698 | .TP |
| 1699 | .BI disable_lat \fR=\fPbool |
| 1700 | Disable measurements of total latency numbers. Useful only for cutting |
| 1701 | back the number of calls to \fBgettimeofday\fR\|(2), as that does impact performance at |
| 1702 | really high IOPS rates. Note that to really get rid of a large amount of these |
| 1703 | calls, this option must be used with disable_slat and disable_bw as well. |
| 1704 | .TP |
| 1705 | .BI disable_clat \fR=\fPbool |
| 1706 | Disable measurements of completion latency numbers. See \fBdisable_lat\fR. |
| 1707 | .TP |
| 1708 | .BI disable_slat \fR=\fPbool |
| 1709 | Disable measurements of submission latency numbers. See \fBdisable_lat\fR. |
| 1710 | .TP |
| 1711 | .BI disable_bw_measurement \fR=\fPbool |
| 1712 | Disable measurements of throughput/bandwidth numbers. See \fBdisable_lat\fR. |
| 1713 | .TP |
| 1714 | .BI lockmem \fR=\fPint |
| 1715 | Pin the specified amount of memory with \fBmlock\fR\|(2). Can be used to |
| 1716 | simulate a smaller amount of memory. The amount specified is per worker. |
| 1717 | .TP |
| 1718 | .BI exec_prerun \fR=\fPstr |
| 1719 | Before running the job, execute the specified command with \fBsystem\fR\|(3). |
| 1720 | .RS |
| 1721 | Output is redirected in a file called \fBjobname.prerun.txt\fR |
| 1722 | .RE |
| 1723 | .TP |
| 1724 | .BI exec_postrun \fR=\fPstr |
| 1725 | Same as \fBexec_prerun\fR, but the command is executed after the job completes. |
| 1726 | .RS |
| 1727 | Output is redirected in a file called \fBjobname.postrun.txt\fR |
| 1728 | .RE |
| 1729 | .TP |
| 1730 | .BI ioscheduler \fR=\fPstr |
| 1731 | Attempt to switch the device hosting the file to the specified I/O scheduler. |
| 1732 | .TP |
| 1733 | .BI disk_util \fR=\fPbool |
| 1734 | Generate disk utilization statistics if the platform supports it. Default: true. |
| 1735 | .TP |
| 1736 | .BI clocksource \fR=\fPstr |
| 1737 | Use the given clocksource as the base of timing. The supported options are: |
| 1738 | .RS |
| 1739 | .TP |
| 1740 | .B gettimeofday |
| 1741 | \fBgettimeofday\fR\|(2) |
| 1742 | .TP |
| 1743 | .B clock_gettime |
| 1744 | \fBclock_gettime\fR\|(2) |
| 1745 | .TP |
| 1746 | .B cpu |
| 1747 | Internal CPU clock source |
| 1748 | .TP |
| 1749 | .RE |
| 1750 | .P |
| 1751 | \fBcpu\fR is the preferred clocksource if it is reliable, as it is very fast |
| 1752 | (and fio is heavy on time calls). Fio will automatically use this clocksource |
| 1753 | if it's supported and considered reliable on the system it is running on, |
| 1754 | unless another clocksource is specifically set. For x86/x86-64 CPUs, this |
| 1755 | means supporting TSC Invariant. |
| 1756 | .TP |
| 1757 | .BI gtod_reduce \fR=\fPbool |
| 1758 | Enable all of the \fBgettimeofday\fR\|(2) reducing options (disable_clat, disable_slat, |
| 1759 | disable_bw) plus reduce precision of the timeout somewhat to really shrink the |
| 1760 | \fBgettimeofday\fR\|(2) call count. With this option enabled, we only do about 0.4% of |
| 1761 | the gtod() calls we would have done if all time keeping was enabled. |
| 1762 | .TP |
| 1763 | .BI gtod_cpu \fR=\fPint |
| 1764 | Sometimes it's cheaper to dedicate a single thread of execution to just getting |
| 1765 | the current time. Fio (and databases, for instance) are very intensive on |
| 1766 | \fBgettimeofday\fR\|(2) calls. With this option, you can set one CPU aside for doing |
| 1767 | nothing but logging current time to a shared memory location. Then the other |
| 1768 | threads/processes that run IO workloads need only copy that segment, instead of |
| 1769 | entering the kernel with a \fBgettimeofday\fR\|(2) call. The CPU set aside for doing |
| 1770 | these time calls will be excluded from other uses. Fio will manually clear it |
| 1771 | from the CPU mask of other jobs. |
| 1772 | .TP |
| 1773 | .BI ignore_error \fR=\fPstr |
| 1774 | Sometimes you want to ignore some errors during test in that case you can specify |
| 1775 | error list for each error type. |
| 1776 | .br |
| 1777 | ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST |
| 1778 | .br |
| 1779 | errors for given error type is separated with ':'. |
| 1780 | Error may be symbol ('ENOSPC', 'ENOMEM') or an integer. |
| 1781 | .br |
| 1782 | Example: ignore_error=EAGAIN,ENOSPC:122 . |
| 1783 | .br |
| 1784 | This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from WRITE. |
| 1785 | .TP |
| 1786 | .BI error_dump \fR=\fPbool |
| 1787 | If set dump every error even if it is non fatal, true by default. If disabled |
| 1788 | only fatal error will be dumped |
| 1789 | .TP |
| 1790 | .BI profile \fR=\fPstr |
| 1791 | Select a specific builtin performance test. |
| 1792 | .TP |
| 1793 | .BI cgroup \fR=\fPstr |
| 1794 | Add job to this control group. If it doesn't exist, it will be created. |
| 1795 | The system must have a mounted cgroup blkio mount point for this to work. If |
| 1796 | your system doesn't have it mounted, you can do so with: |
| 1797 | |
| 1798 | # mount \-t cgroup \-o blkio none /cgroup |
| 1799 | .TP |
| 1800 | .BI cgroup_weight \fR=\fPint |
| 1801 | Set the weight of the cgroup to this value. See the documentation that comes |
| 1802 | with the kernel, allowed values are in the range of 100..1000. |
| 1803 | .TP |
| 1804 | .BI cgroup_nodelete \fR=\fPbool |
| 1805 | Normally fio will delete the cgroups it has created after the job completion. |
| 1806 | To override this behavior and to leave cgroups around after the job completion, |
| 1807 | set cgroup_nodelete=1. This can be useful if one wants to inspect various |
| 1808 | cgroup files after job completion. Default: false |
| 1809 | .TP |
| 1810 | .BI uid \fR=\fPint |
| 1811 | Instead of running as the invoking user, set the user ID to this value before |
| 1812 | the thread/process does any work. |
| 1813 | .TP |
| 1814 | .BI gid \fR=\fPint |
| 1815 | Set group ID, see \fBuid\fR. |
| 1816 | .TP |
| 1817 | .BI unit_base \fR=\fPint |
| 1818 | Base unit for reporting. Allowed values are: |
| 1819 | .RS |
| 1820 | .TP |
| 1821 | .B 0 |
| 1822 | Use auto-detection (default). |
| 1823 | .TP |
| 1824 | .B 8 |
| 1825 | Byte based. |
| 1826 | .TP |
| 1827 | .B 1 |
| 1828 | Bit based. |
| 1829 | .RE |
| 1830 | .P |
| 1831 | .TP |
| 1832 | .BI flow_id \fR=\fPint |
| 1833 | The ID of the flow. If not specified, it defaults to being a global flow. See |
| 1834 | \fBflow\fR. |
| 1835 | .TP |
| 1836 | .BI flow \fR=\fPint |
| 1837 | Weight in token-based flow control. If this value is used, then there is a |
| 1838 | \fBflow counter\fR which is used to regulate the proportion of activity between |
| 1839 | two or more jobs. fio attempts to keep this flow counter near zero. The |
| 1840 | \fBflow\fR parameter stands for how much should be added or subtracted to the |
| 1841 | flow counter on each iteration of the main I/O loop. That is, if one job has |
| 1842 | \fBflow=8\fR and another job has \fBflow=-1\fR, then there will be a roughly |
| 1843 | 1:8 ratio in how much one runs vs the other. |
| 1844 | .TP |
| 1845 | .BI flow_watermark \fR=\fPint |
| 1846 | The maximum value that the absolute value of the flow counter is allowed to |
| 1847 | reach before the job must wait for a lower value of the counter. |
| 1848 | .TP |
| 1849 | .BI flow_sleep \fR=\fPint |
| 1850 | The period of time, in microseconds, to wait after the flow watermark has been |
| 1851 | exceeded before retrying operations |
| 1852 | .TP |
| 1853 | .BI clat_percentiles \fR=\fPbool |
| 1854 | Enable the reporting of percentiles of completion latencies. |
| 1855 | .TP |
| 1856 | .BI percentile_list \fR=\fPfloat_list |
| 1857 | Overwrite the default list of percentiles for completion latencies and the |
| 1858 | block error histogram. Each number is a floating number in the range (0,100], |
| 1859 | and the maximum length of the list is 20. Use ':' to separate the |
| 1860 | numbers. For example, \-\-percentile_list=99.5:99.9 will cause fio to |
| 1861 | report the values of completion latency below which 99.5% and 99.9% of |
| 1862 | the observed latencies fell, respectively. |
| 1863 | .SS "Ioengine Parameters List" |
| 1864 | Some parameters are only valid when a specific ioengine is in use. These are |
| 1865 | used identically to normal parameters, with the caveat that when used on the |
| 1866 | command line, they must come after the ioengine. |
| 1867 | .TP |
| 1868 | .BI (cpuio)cpuload \fR=\fPint |
| 1869 | Attempt to use the specified percentage of CPU cycles. |
| 1870 | .TP |
| 1871 | .BI (cpuio)cpuchunks \fR=\fPint |
| 1872 | Split the load into cycles of the given time. In microseconds. |
| 1873 | .TP |
| 1874 | .BI (cpuio)exit_on_io_done \fR=\fPbool |
| 1875 | Detect when IO threads are done, then exit. |
| 1876 | .TP |
| 1877 | .BI (libaio)userspace_reap |
| 1878 | Normally, with the libaio engine in use, fio will use |
| 1879 | the io_getevents system call to reap newly returned events. |
| 1880 | With this flag turned on, the AIO ring will be read directly |
| 1881 | from user-space to reap events. The reaping mode is only |
| 1882 | enabled when polling for a minimum of 0 events (eg when |
| 1883 | iodepth_batch_complete=0). |
| 1884 | .TP |
| 1885 | .BI (pvsync2)hipri |
| 1886 | Set RWF_HIPRI on IO, indicating to the kernel that it's of |
| 1887 | higher priority than normal. |
| 1888 | .TP |
| 1889 | .BI (net,netsplice)hostname \fR=\fPstr |
| 1890 | The host name or IP address to use for TCP or UDP based IO. |
| 1891 | If the job is a TCP listener or UDP reader, the hostname is not |
| 1892 | used and must be omitted unless it is a valid UDP multicast address. |
| 1893 | .TP |
| 1894 | .BI (net,netsplice)port \fR=\fPint |
| 1895 | The TCP or UDP port to bind to or connect to. If this is used with |
| 1896 | \fBnumjobs\fR to spawn multiple instances of the same job type, then |
| 1897 | this will be the starting port number since fio will use a range of ports. |
| 1898 | .TP |
| 1899 | .BI (net,netsplice)interface \fR=\fPstr |
| 1900 | The IP address of the network interface used to send or receive UDP multicast |
| 1901 | packets. |
| 1902 | .TP |
| 1903 | .BI (net,netsplice)ttl \fR=\fPint |
| 1904 | Time-to-live value for outgoing UDP multicast packets. Default: 1 |
| 1905 | .TP |
| 1906 | .BI (net,netsplice)nodelay \fR=\fPbool |
| 1907 | Set TCP_NODELAY on TCP connections. |
| 1908 | .TP |
| 1909 | .BI (net,netsplice)protocol \fR=\fPstr "\fR,\fP proto" \fR=\fPstr |
| 1910 | The network protocol to use. Accepted values are: |
| 1911 | .RS |
| 1912 | .RS |
| 1913 | .TP |
| 1914 | .B tcp |
| 1915 | Transmission control protocol |
| 1916 | .TP |
| 1917 | .B tcpv6 |
| 1918 | Transmission control protocol V6 |
| 1919 | .TP |
| 1920 | .B udp |
| 1921 | User datagram protocol |
| 1922 | .TP |
| 1923 | .B udpv6 |
| 1924 | User datagram protocol V6 |
| 1925 | .TP |
| 1926 | .B unix |
| 1927 | UNIX domain socket |
| 1928 | .RE |
| 1929 | .P |
| 1930 | When the protocol is TCP or UDP, the port must also be given, |
| 1931 | as well as the hostname if the job is a TCP listener or UDP |
| 1932 | reader. For unix sockets, the normal filename option should be |
| 1933 | used and the port is invalid. |
| 1934 | .RE |
| 1935 | .TP |
| 1936 | .BI (net,netsplice)listen |
| 1937 | For TCP network connections, tell fio to listen for incoming |
| 1938 | connections rather than initiating an outgoing connection. The |
| 1939 | hostname must be omitted if this option is used. |
| 1940 | .TP |
| 1941 | .BI (net, pingpong) \fR=\fPbool |
| 1942 | Normally a network writer will just continue writing data, and a network reader |
| 1943 | will just consume packets. If pingpong=1 is set, a writer will send its normal |
| 1944 | payload to the reader, then wait for the reader to send the same payload back. |
| 1945 | This allows fio to measure network latencies. The submission and completion |
| 1946 | latencies then measure local time spent sending or receiving, and the |
| 1947 | completion latency measures how long it took for the other end to receive and |
| 1948 | send back. For UDP multicast traffic pingpong=1 should only be set for a single |
| 1949 | reader when multiple readers are listening to the same address. |
| 1950 | .TP |
| 1951 | .BI (net, window_size) \fR=\fPint |
| 1952 | Set the desired socket buffer size for the connection. |
| 1953 | .TP |
| 1954 | .BI (net, mss) \fR=\fPint |
| 1955 | Set the TCP maximum segment size (TCP_MAXSEG). |
| 1956 | .TP |
| 1957 | .BI (e4defrag,donorname) \fR=\fPstr |
| 1958 | File will be used as a block donor (swap extents between files) |
| 1959 | .TP |
| 1960 | .BI (e4defrag,inplace) \fR=\fPint |
| 1961 | Configure donor file block allocation strategy |
| 1962 | .RS |
| 1963 | .BI 0(default) : |
| 1964 | Preallocate donor's file on init |
| 1965 | .TP |
| 1966 | .BI 1: |
| 1967 | allocate space immediately inside defragment event, and free right after event |
| 1968 | .RE |
| 1969 | .TP |
| 1970 | .BI (rbd)clustername \fR=\fPstr |
| 1971 | Specifies the name of the ceph cluster. |
| 1972 | .TP |
| 1973 | .BI (rbd)rbdname \fR=\fPstr |
| 1974 | Specifies the name of the RBD. |
| 1975 | .TP |
| 1976 | .BI (rbd)pool \fR=\fPstr |
| 1977 | Specifies the name of the Ceph pool containing the RBD. |
| 1978 | .TP |
| 1979 | .BI (rbd)clientname \fR=\fPstr |
| 1980 | Specifies the username (without the 'client.' prefix) used to access the Ceph |
| 1981 | cluster. If the clustername is specified, the clientname shall be the full |
| 1982 | type.id string. If no type. prefix is given, fio will add 'client.' by default. |
| 1983 | .TP |
| 1984 | .BI (mtd)skipbad \fR=\fPbool |
| 1985 | Skip operations against known bad blocks. |
| 1986 | .SH OUTPUT |
| 1987 | While running, \fBfio\fR will display the status of the created jobs. For |
| 1988 | example: |
| 1989 | .RS |
| 1990 | .P |
| 1991 | Jobs: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s] |
| 1992 | .RE |
| 1993 | .P |
| 1994 | The characters in the first set of brackets denote the current status of each |
| 1995 | threads. The possible values are: |
| 1996 | .P |
| 1997 | .PD 0 |
| 1998 | .RS |
| 1999 | .TP |
| 2000 | .B P |
| 2001 | Setup but not started. |
| 2002 | .TP |
| 2003 | .B C |
| 2004 | Thread created. |
| 2005 | .TP |
| 2006 | .B I |
| 2007 | Initialized, waiting. |
| 2008 | .TP |
| 2009 | .B R |
| 2010 | Running, doing sequential reads. |
| 2011 | .TP |
| 2012 | .B r |
| 2013 | Running, doing random reads. |
| 2014 | .TP |
| 2015 | .B W |
| 2016 | Running, doing sequential writes. |
| 2017 | .TP |
| 2018 | .B w |
| 2019 | Running, doing random writes. |
| 2020 | .TP |
| 2021 | .B M |
| 2022 | Running, doing mixed sequential reads/writes. |
| 2023 | .TP |
| 2024 | .B m |
| 2025 | Running, doing mixed random reads/writes. |
| 2026 | .TP |
| 2027 | .B F |
| 2028 | Running, currently waiting for \fBfsync\fR\|(2). |
| 2029 | .TP |
| 2030 | .B V |
| 2031 | Running, verifying written data. |
| 2032 | .TP |
| 2033 | .B E |
| 2034 | Exited, not reaped by main thread. |
| 2035 | .TP |
| 2036 | .B \- |
| 2037 | Exited, thread reaped. |
| 2038 | .RE |
| 2039 | .PD |
| 2040 | .P |
| 2041 | The second set of brackets shows the estimated completion percentage of |
| 2042 | the current group. The third set shows the read and write I/O rate, |
| 2043 | respectively. Finally, the estimated run time of the job is displayed. |
| 2044 | .P |
| 2045 | When \fBfio\fR completes (or is interrupted by Ctrl-C), it will show data |
| 2046 | for each thread, each group of threads, and each disk, in that order. |
| 2047 | .P |
| 2048 | Per-thread statistics first show the threads client number, group-id, and |
| 2049 | error code. The remaining figures are as follows: |
| 2050 | .RS |
| 2051 | .TP |
| 2052 | .B io |
| 2053 | Number of megabytes of I/O performed. |
| 2054 | .TP |
| 2055 | .B bw |
| 2056 | Average data rate (bandwidth). |
| 2057 | .TP |
| 2058 | .B runt |
| 2059 | Threads run time. |
| 2060 | .TP |
| 2061 | .B slat |
| 2062 | Submission latency minimum, maximum, average and standard deviation. This is |
| 2063 | the time it took to submit the I/O. |
| 2064 | .TP |
| 2065 | .B clat |
| 2066 | Completion latency minimum, maximum, average and standard deviation. This |
| 2067 | is the time between submission and completion. |
| 2068 | .TP |
| 2069 | .B bw |
| 2070 | Bandwidth minimum, maximum, percentage of aggregate bandwidth received, average |
| 2071 | and standard deviation. |
| 2072 | .TP |
| 2073 | .B cpu |
| 2074 | CPU usage statistics. Includes user and system time, number of context switches |
| 2075 | this thread went through and number of major and minor page faults. The CPU |
| 2076 | utilization numbers are averages for the jobs in that reporting group, while |
| 2077 | the context and fault counters are summed. |
| 2078 | .TP |
| 2079 | .B IO depths |
| 2080 | Distribution of I/O depths. Each depth includes everything less than (or equal) |
| 2081 | to it, but greater than the previous depth. |
| 2082 | .TP |
| 2083 | .B IO issued |
| 2084 | Number of read/write requests issued, and number of short read/write requests. |
| 2085 | .TP |
| 2086 | .B IO latencies |
| 2087 | Distribution of I/O completion latencies. The numbers follow the same pattern |
| 2088 | as \fBIO depths\fR. |
| 2089 | .RE |
| 2090 | .P |
| 2091 | The group statistics show: |
| 2092 | .PD 0 |
| 2093 | .RS |
| 2094 | .TP |
| 2095 | .B io |
| 2096 | Number of megabytes I/O performed. |
| 2097 | .TP |
| 2098 | .B aggrb |
| 2099 | Aggregate bandwidth of threads in the group. |
| 2100 | .TP |
| 2101 | .B minb |
| 2102 | Minimum average bandwidth a thread saw. |
| 2103 | .TP |
| 2104 | .B maxb |
| 2105 | Maximum average bandwidth a thread saw. |
| 2106 | .TP |
| 2107 | .B mint |
| 2108 | Shortest runtime of threads in the group. |
| 2109 | .TP |
| 2110 | .B maxt |
| 2111 | Longest runtime of threads in the group. |
| 2112 | .RE |
| 2113 | .PD |
| 2114 | .P |
| 2115 | Finally, disk statistics are printed with reads first: |
| 2116 | .PD 0 |
| 2117 | .RS |
| 2118 | .TP |
| 2119 | .B ios |
| 2120 | Number of I/Os performed by all groups. |
| 2121 | .TP |
| 2122 | .B merge |
| 2123 | Number of merges in the I/O scheduler. |
| 2124 | .TP |
| 2125 | .B ticks |
| 2126 | Number of ticks we kept the disk busy. |
| 2127 | .TP |
| 2128 | .B io_queue |
| 2129 | Total time spent in the disk queue. |
| 2130 | .TP |
| 2131 | .B util |
| 2132 | Disk utilization. |
| 2133 | .RE |
| 2134 | .PD |
| 2135 | .P |
| 2136 | It is also possible to get fio to dump the current output while it is |
| 2137 | running, without terminating the job. To do that, send fio the \fBUSR1\fR |
| 2138 | signal. |
| 2139 | .SH TERSE OUTPUT |
| 2140 | If the \fB\-\-minimal\fR / \fB\-\-append-terse\fR options are given, the |
| 2141 | results will be printed/appended in a semicolon-delimited format suitable for |
| 2142 | scripted use. |
| 2143 | A job description (if provided) follows on a new line. Note that the first |
| 2144 | number in the line is the version number. If the output has to be changed |
| 2145 | for some reason, this number will be incremented by 1 to signify that |
| 2146 | change. The fields are: |
| 2147 | .P |
| 2148 | .RS |
| 2149 | .B terse version, fio version, jobname, groupid, error |
| 2150 | .P |
| 2151 | Read status: |
| 2152 | .RS |
| 2153 | .B Total I/O \fR(KiB)\fP, bandwidth \fR(KiB/s)\fP, IOPS, runtime \fR(ms)\fP |
| 2154 | .P |
| 2155 | Submission latency: |
| 2156 | .RS |
| 2157 | .B min, max, mean, standard deviation |
| 2158 | .RE |
| 2159 | Completion latency: |
| 2160 | .RS |
| 2161 | .B min, max, mean, standard deviation |
| 2162 | .RE |
| 2163 | Completion latency percentiles (20 fields): |
| 2164 | .RS |
| 2165 | .B Xth percentile=usec |
| 2166 | .RE |
| 2167 | Total latency: |
| 2168 | .RS |
| 2169 | .B min, max, mean, standard deviation |
| 2170 | .RE |
| 2171 | Bandwidth: |
| 2172 | .RS |
| 2173 | .B min, max, aggregate percentage of total, mean, standard deviation |
| 2174 | .RE |
| 2175 | .RE |
| 2176 | .P |
| 2177 | Write status: |
| 2178 | .RS |
| 2179 | .B Total I/O \fR(KiB)\fP, bandwidth \fR(KiB/s)\fP, IOPS, runtime \fR(ms)\fP |
| 2180 | .P |
| 2181 | Submission latency: |
| 2182 | .RS |
| 2183 | .B min, max, mean, standard deviation |
| 2184 | .RE |
| 2185 | Completion latency: |
| 2186 | .RS |
| 2187 | .B min, max, mean, standard deviation |
| 2188 | .RE |
| 2189 | Completion latency percentiles (20 fields): |
| 2190 | .RS |
| 2191 | .B Xth percentile=usec |
| 2192 | .RE |
| 2193 | Total latency: |
| 2194 | .RS |
| 2195 | .B min, max, mean, standard deviation |
| 2196 | .RE |
| 2197 | Bandwidth: |
| 2198 | .RS |
| 2199 | .B min, max, aggregate percentage of total, mean, standard deviation |
| 2200 | .RE |
| 2201 | .RE |
| 2202 | .P |
| 2203 | CPU usage: |
| 2204 | .RS |
| 2205 | .B user, system, context switches, major page faults, minor page faults |
| 2206 | .RE |
| 2207 | .P |
| 2208 | IO depth distribution: |
| 2209 | .RS |
| 2210 | .B <=1, 2, 4, 8, 16, 32, >=64 |
| 2211 | .RE |
| 2212 | .P |
| 2213 | IO latency distribution: |
| 2214 | .RS |
| 2215 | Microseconds: |
| 2216 | .RS |
| 2217 | .B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000 |
| 2218 | .RE |
| 2219 | Milliseconds: |
| 2220 | .RS |
| 2221 | .B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000 |
| 2222 | .RE |
| 2223 | .RE |
| 2224 | .P |
| 2225 | Disk utilization (1 for each disk used): |
| 2226 | .RS |
| 2227 | .B name, read ios, write ios, read merges, write merges, read ticks, write ticks, read in-queue time, write in-queue time, disk utilization percentage |
| 2228 | .RE |
| 2229 | .P |
| 2230 | Error Info (dependent on continue_on_error, default off): |
| 2231 | .RS |
| 2232 | .B total # errors, first error code |
| 2233 | .RE |
| 2234 | .P |
| 2235 | .B text description (if provided in config - appears on newline) |
| 2236 | .RE |
| 2237 | .SH TRACE FILE FORMAT |
| 2238 | There are two trace file format that you can encounter. The older (v1) format |
| 2239 | is unsupported since version 1.20-rc3 (March 2008). It will still be described |
| 2240 | below in case that you get an old trace and want to understand it. |
| 2241 | |
| 2242 | In any case the trace is a simple text file with a single action per line. |
| 2243 | |
| 2244 | .P |
| 2245 | .B Trace file format v1 |
| 2246 | .RS |
| 2247 | Each line represents a single io action in the following format: |
| 2248 | |
| 2249 | rw, offset, length |
| 2250 | |
| 2251 | where rw=0/1 for read/write, and the offset and length entries being in bytes. |
| 2252 | |
| 2253 | This format is not supported in Fio versions => 1.20-rc3. |
| 2254 | |
| 2255 | .RE |
| 2256 | .P |
| 2257 | .B Trace file format v2 |
| 2258 | .RS |
| 2259 | The second version of the trace file format was added in Fio version 1.17. |
| 2260 | It allows one to access more then one file per trace and has a bigger set of |
| 2261 | possible file actions. |
| 2262 | |
| 2263 | The first line of the trace file has to be: |
| 2264 | |
| 2265 | \fBfio version 2 iolog\fR |
| 2266 | |
| 2267 | Following this can be lines in two different formats, which are described below. |
| 2268 | The file management format: |
| 2269 | |
| 2270 | \fBfilename action\fR |
| 2271 | |
| 2272 | The filename is given as an absolute path. The action can be one of these: |
| 2273 | |
| 2274 | .P |
| 2275 | .PD 0 |
| 2276 | .RS |
| 2277 | .TP |
| 2278 | .B add |
| 2279 | Add the given filename to the trace |
| 2280 | .TP |
| 2281 | .B open |
| 2282 | Open the file with the given filename. The filename has to have been previously |
| 2283 | added with the \fBadd\fR action. |
| 2284 | .TP |
| 2285 | .B close |
| 2286 | Close the file with the given filename. The file must have previously been |
| 2287 | opened. |
| 2288 | .RE |
| 2289 | .PD |
| 2290 | .P |
| 2291 | |
| 2292 | The file io action format: |
| 2293 | |
| 2294 | \fBfilename action offset length\fR |
| 2295 | |
| 2296 | The filename is given as an absolute path, and has to have been added and opened |
| 2297 | before it can be used with this format. The offset and length are given in |
| 2298 | bytes. The action can be one of these: |
| 2299 | |
| 2300 | .P |
| 2301 | .PD 0 |
| 2302 | .RS |
| 2303 | .TP |
| 2304 | .B wait |
| 2305 | Wait for 'offset' microseconds. Everything below 100 is discarded. The time is |
| 2306 | relative to the previous wait statement. |
| 2307 | .TP |
| 2308 | .B read |
| 2309 | Read \fBlength\fR bytes beginning from \fBoffset\fR |
| 2310 | .TP |
| 2311 | .B write |
| 2312 | Write \fBlength\fR bytes beginning from \fBoffset\fR |
| 2313 | .TP |
| 2314 | .B sync |
| 2315 | fsync() the file |
| 2316 | .TP |
| 2317 | .B datasync |
| 2318 | fdatasync() the file |
| 2319 | .TP |
| 2320 | .B trim |
| 2321 | trim the given file from the given \fBoffset\fR for \fBlength\fR bytes |
| 2322 | .RE |
| 2323 | .PD |
| 2324 | .P |
| 2325 | |
| 2326 | .SH CPU IDLENESS PROFILING |
| 2327 | In some cases, we want to understand CPU overhead in a test. For example, |
| 2328 | we test patches for the specific goodness of whether they reduce CPU usage. |
| 2329 | fio implements a balloon approach to create a thread per CPU that runs at |
| 2330 | idle priority, meaning that it only runs when nobody else needs the cpu. |
| 2331 | By measuring the amount of work completed by the thread, idleness of each |
| 2332 | CPU can be derived accordingly. |
| 2333 | |
| 2334 | An unit work is defined as touching a full page of unsigned characters. Mean |
| 2335 | and standard deviation of time to complete an unit work is reported in "unit |
| 2336 | work" section. Options can be chosen to report detailed percpu idleness or |
| 2337 | overall system idleness by aggregating percpu stats. |
| 2338 | |
| 2339 | .SH VERIFICATION AND TRIGGERS |
| 2340 | Fio is usually run in one of two ways, when data verification is done. The |
| 2341 | first is a normal write job of some sort with verify enabled. When the |
| 2342 | write phase has completed, fio switches to reads and verifies everything |
| 2343 | it wrote. The second model is running just the write phase, and then later |
| 2344 | on running the same job (but with reads instead of writes) to repeat the |
| 2345 | same IO patterns and verify the contents. Both of these methods depend |
| 2346 | on the write phase being completed, as fio otherwise has no idea how much |
| 2347 | data was written. |
| 2348 | |
| 2349 | With verification triggers, fio supports dumping the current write state |
| 2350 | to local files. Then a subsequent read verify workload can load this state |
| 2351 | and know exactly where to stop. This is useful for testing cases where |
| 2352 | power is cut to a server in a managed fashion, for instance. |
| 2353 | |
| 2354 | A verification trigger consists of two things: |
| 2355 | |
| 2356 | .RS |
| 2357 | Storing the write state of each job |
| 2358 | .LP |
| 2359 | Executing a trigger command |
| 2360 | .RE |
| 2361 | |
| 2362 | The write state is relatively small, on the order of hundreds of bytes |
| 2363 | to single kilobytes. It contains information on the number of completions |
| 2364 | done, the last X completions, etc. |
| 2365 | |
| 2366 | A trigger is invoked either through creation (\fBtouch\fR) of a specified |
| 2367 | file in the system, or through a timeout setting. If fio is run with |
| 2368 | \fB\-\-trigger\-file=/tmp/trigger-file\fR, then it will continually check for |
| 2369 | the existence of /tmp/trigger-file. When it sees this file, it will |
| 2370 | fire off the trigger (thus saving state, and executing the trigger |
| 2371 | command). |
| 2372 | |
| 2373 | For client/server runs, there's both a local and remote trigger. If |
| 2374 | fio is running as a server backend, it will send the job states back |
| 2375 | to the client for safe storage, then execute the remote trigger, if |
| 2376 | specified. If a local trigger is specified, the server will still send |
| 2377 | back the write state, but the client will then execute the trigger. |
| 2378 | |
| 2379 | .RE |
| 2380 | .P |
| 2381 | .B Verification trigger example |
| 2382 | .RS |
| 2383 | |
| 2384 | Lets say we want to run a powercut test on the remote machine 'server'. |
| 2385 | Our write workload is in write-test.fio. We want to cut power to 'server' |
| 2386 | at some point during the run, and we'll run this test from the safety |
| 2387 | or our local machine, 'localbox'. On the server, we'll start the fio |
| 2388 | backend normally: |
| 2389 | |
| 2390 | server# \fBfio \-\-server\fR |
| 2391 | |
| 2392 | and on the client, we'll fire off the workload: |
| 2393 | |
| 2394 | localbox$ \fBfio \-\-client=server \-\-trigger\-file=/tmp/my\-trigger \-\-trigger-remote="bash \-c "echo b > /proc/sysrq-triger""\fR |
| 2395 | |
| 2396 | We set \fB/tmp/my-trigger\fR as the trigger file, and we tell fio to execute |
| 2397 | |
| 2398 | \fBecho b > /proc/sysrq-trigger\fR |
| 2399 | |
| 2400 | on the server once it has received the trigger and sent us the write |
| 2401 | state. This will work, but it's not \fIreally\fR cutting power to the server, |
| 2402 | it's merely abruptly rebooting it. If we have a remote way of cutting |
| 2403 | power to the server through IPMI or similar, we could do that through |
| 2404 | a local trigger command instead. Lets assume we have a script that does |
| 2405 | IPMI reboot of a given hostname, ipmi-reboot. On localbox, we could |
| 2406 | then have run fio with a local trigger instead: |
| 2407 | |
| 2408 | localbox$ \fBfio \-\-client=server \-\-trigger\-file=/tmp/my\-trigger \-\-trigger="ipmi-reboot server"\fR |
| 2409 | |
| 2410 | For this case, fio would wait for the server to send us the write state, |
| 2411 | then execute 'ipmi-reboot server' when that happened. |
| 2412 | |
| 2413 | .RE |
| 2414 | .P |
| 2415 | .B Loading verify state |
| 2416 | .RS |
| 2417 | To load store write state, read verification job file must contain |
| 2418 | the verify_state_load option. If that is set, fio will load the previously |
| 2419 | stored state. For a local fio run this is done by loading the files directly, |
| 2420 | and on a client/server run, the server backend will ask the client to send |
| 2421 | the files over and load them from there. |
| 2422 | |
| 2423 | .RE |
| 2424 | |
| 2425 | .SH LOG FILE FORMATS |
| 2426 | |
| 2427 | Fio supports a variety of log file formats, for logging latencies, bandwidth, |
| 2428 | and IOPS. The logs share a common format, which looks like this: |
| 2429 | |
| 2430 | .B time (msec), value, data direction, offset |
| 2431 | |
| 2432 | Time for the log entry is always in milliseconds. The value logged depends |
| 2433 | on the type of log, it will be one of the following: |
| 2434 | |
| 2435 | .P |
| 2436 | .PD 0 |
| 2437 | .TP |
| 2438 | .B Latency log |
| 2439 | Value is in latency in usecs |
| 2440 | .TP |
| 2441 | .B Bandwidth log |
| 2442 | Value is in KiB/sec |
| 2443 | .TP |
| 2444 | .B IOPS log |
| 2445 | Value is in IOPS |
| 2446 | .PD |
| 2447 | .P |
| 2448 | |
| 2449 | Data direction is one of the following: |
| 2450 | |
| 2451 | .P |
| 2452 | .PD 0 |
| 2453 | .TP |
| 2454 | .B 0 |
| 2455 | IO is a READ |
| 2456 | .TP |
| 2457 | .B 1 |
| 2458 | IO is a WRITE |
| 2459 | .TP |
| 2460 | .B 2 |
| 2461 | IO is a TRIM |
| 2462 | .PD |
| 2463 | .P |
| 2464 | |
| 2465 | The \fIoffset\fR is the offset, in bytes, from the start of the file, for that |
| 2466 | particular IO. The logging of the offset can be toggled with \fBlog_offset\fR. |
| 2467 | |
| 2468 | If windowed logging is enabled through \fBlog_avg_msec\fR, then fio doesn't log |
| 2469 | individual IOs. Instead of logs the average values over the specified |
| 2470 | period of time. Since \fIdata direction\fR and \fIoffset\fR are per-IO values, |
| 2471 | they aren't applicable if windowed logging is enabled. If windowed logging |
| 2472 | is enabled and \fBlog_max_value\fR is set, then fio logs maximum values in |
| 2473 | that window instead of averages. |
| 2474 | |
| 2475 | For histogram logging the logs look like this: |
| 2476 | |
| 2477 | .B time (msec), data direction, block-size, bin 0, bin 1, ..., bin 1215 |
| 2478 | |
| 2479 | Where 'bin i' gives the frequency of IO requests with a latency falling in |
| 2480 | the i-th bin. See \fBlog_hist_coarseness\fR for logging fewer bins. |
| 2481 | |
| 2482 | .RE |
| 2483 | |
| 2484 | .SH CLIENT / SERVER |
| 2485 | Normally you would run fio as a stand-alone application on the machine |
| 2486 | where the IO workload should be generated. However, it is also possible to |
| 2487 | run the frontend and backend of fio separately. This makes it possible to |
| 2488 | have a fio server running on the machine(s) where the IO workload should |
| 2489 | be running, while controlling it from another machine. |
| 2490 | |
| 2491 | To start the server, you would do: |
| 2492 | |
| 2493 | \fBfio \-\-server=args\fR |
| 2494 | |
| 2495 | on that machine, where args defines what fio listens to. The arguments |
| 2496 | are of the form 'type:hostname or IP:port'. 'type' is either 'ip' (or ip4) |
| 2497 | for TCP/IP v4, 'ip6' for TCP/IP v6, or 'sock' for a local unix domain |
| 2498 | socket. 'hostname' is either a hostname or IP address, and 'port' is the port to |
| 2499 | listen to (only valid for TCP/IP, not a local socket). Some examples: |
| 2500 | |
| 2501 | 1) \fBfio \-\-server\fR |
| 2502 | |
| 2503 | Start a fio server, listening on all interfaces on the default port (8765). |
| 2504 | |
| 2505 | 2) \fBfio \-\-server=ip:hostname,4444\fR |
| 2506 | |
| 2507 | Start a fio server, listening on IP belonging to hostname and on port 4444. |
| 2508 | |
| 2509 | 3) \fBfio \-\-server=ip6:::1,4444\fR |
| 2510 | |
| 2511 | Start a fio server, listening on IPv6 localhost ::1 and on port 4444. |
| 2512 | |
| 2513 | 4) \fBfio \-\-server=,4444\fR |
| 2514 | |
| 2515 | Start a fio server, listening on all interfaces on port 4444. |
| 2516 | |
| 2517 | 5) \fBfio \-\-server=1.2.3.4\fR |
| 2518 | |
| 2519 | Start a fio server, listening on IP 1.2.3.4 on the default port. |
| 2520 | |
| 2521 | 6) \fBfio \-\-server=sock:/tmp/fio.sock\fR |
| 2522 | |
| 2523 | Start a fio server, listening on the local socket /tmp/fio.sock. |
| 2524 | |
| 2525 | When a server is running, you can connect to it from a client. The client |
| 2526 | is run with: |
| 2527 | |
| 2528 | \fBfio \-\-local-args \-\-client=server \-\-remote-args <job file(s)>\fR |
| 2529 | |
| 2530 | where \-\-local-args are arguments that are local to the client where it is |
| 2531 | running, 'server' is the connect string, and \-\-remote-args and <job file(s)> |
| 2532 | are sent to the server. The 'server' string follows the same format as it |
| 2533 | does on the server side, to allow IP/hostname/socket and port strings. |
| 2534 | You can connect to multiple clients as well, to do that you could run: |
| 2535 | |
| 2536 | \fBfio \-\-client=server2 \-\-client=server2 <job file(s)>\fR |
| 2537 | |
| 2538 | If the job file is located on the fio server, then you can tell the server |
| 2539 | to load a local file as well. This is done by using \-\-remote-config: |
| 2540 | |
| 2541 | \fBfio \-\-client=server \-\-remote-config /path/to/file.fio\fR |
| 2542 | |
| 2543 | Then fio will open this local (to the server) job file instead |
| 2544 | of being passed one from the client. |
| 2545 | |
| 2546 | If you have many servers (example: 100 VMs/containers), you can input a pathname |
| 2547 | of a file containing host IPs/names as the parameter value for the \-\-client option. |
| 2548 | For example, here is an example "host.list" file containing 2 hostnames: |
| 2549 | |
| 2550 | host1.your.dns.domain |
| 2551 | .br |
| 2552 | host2.your.dns.domain |
| 2553 | |
| 2554 | The fio command would then be: |
| 2555 | |
| 2556 | \fBfio \-\-client=host.list <job file>\fR |
| 2557 | |
| 2558 | In this mode, you cannot input server-specific parameters or job files, and all |
| 2559 | servers receive the same job file. |
| 2560 | |
| 2561 | In order to enable fio \-\-client runs utilizing a shared filesystem from multiple hosts, |
| 2562 | fio \-\-client now prepends the IP address of the server to the filename. For example, |
| 2563 | if fio is using directory /mnt/nfs/fio and is writing filename fileio.tmp, |
| 2564 | with a \-\-client hostfile |
| 2565 | containing two hostnames h1 and h2 with IP addresses 192.168.10.120 and 192.168.10.121, then |
| 2566 | fio will create two files: |
| 2567 | |
| 2568 | /mnt/nfs/fio/192.168.10.120.fileio.tmp |
| 2569 | .br |
| 2570 | /mnt/nfs/fio/192.168.10.121.fileio.tmp |
| 2571 | |
| 2572 | .SH AUTHORS |
| 2573 | |
| 2574 | .B fio |
| 2575 | was written by Jens Axboe <jens.axboe@oracle.com>, |
| 2576 | now Jens Axboe <axboe@fb.com>. |
| 2577 | .br |
| 2578 | This man page was written by Aaron Carroll <aaronc@cse.unsw.edu.au> based |
| 2579 | on documentation by Jens Axboe. |
| 2580 | .SH "REPORTING BUGS" |
| 2581 | Report bugs to the \fBfio\fR mailing list <fio@vger.kernel.org>. |
| 2582 | See \fBREADME\fR. |
| 2583 | .SH "SEE ALSO" |
| 2584 | For further documentation see \fBHOWTO\fR and \fBREADME\fR. |
| 2585 | .br |
| 2586 | Sample jobfiles are available in the \fBexamples\fR directory. |