1 .TH fio 1 "December 2014" "User Manual"
3 fio \- flexible I/O tester
6 [\fIoptions\fR] [\fIjobfile\fR]...
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.
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.
20 .BI \-\-output \fR=\fPfilename
21 Write output to \fIfilename\fR.
23 .BI \-\-output-format \fR=\fPformat
24 Set the reporting format to \fInormal\fR, \fIterse\fR, or \fIjson\fR.
25 Multiple formats can be selected, separate by a comma.
27 .BI \-\-runtime \fR=\fPruntime
28 Limit run time to \fIruntime\fR seconds.
31 Generate per-job bandwidth logs.
34 Print statistics in a terse, semicolon-delimited format.
37 Print statistics in selected mode AND terse, semicolon-delimited format.
38 Deprecated, use \-\-output-format instead to select multiple formats.
41 Display version information and exit.
43 .BI \-\-terse\-version \fR=\fPversion
44 Set terse version output format (Current version 3, or older version 2).
47 Display usage information and exit.
50 Perform test and validation of internal CPU clock
52 .BI \-\-crctest[\fR=\fPtest]
53 Test the speed of the builtin checksumming functions. If no argument is given,
54 all of them are tested. Or a comma separated list can be passed, in which
55 case the given ones are tested.
57 .BI \-\-cmdhelp \fR=\fPcommand
58 Print help information for \fIcommand\fR. May be `all' for all commands.
60 .BI \-\-enghelp \fR=\fPioengine[,command]
61 List all commands defined by \fIioengine\fR, or print help for \fIcommand\fR defined by \fIioengine\fR.
63 .BI \-\-showcmd \fR=\fPjobfile
64 Convert \fIjobfile\fR to a set of command-line options.
66 .BI \-\-eta \fR=\fPwhen
67 Specifies when real-time ETA estimate should be printed. \fIwhen\fR may
68 be one of `always', `never' or `auto'.
70 .BI \-\-eta\-newline \fR=\fPtime
71 Force an ETA newline for every `time` period passed.
73 .BI \-\-status\-interval \fR=\fPtime
74 Report full output status every `time` period passed.
77 Turn on safety read-only checks, preventing any attempted write.
79 .BI \-\-section \fR=\fPsec
80 Only run section \fIsec\fR from job file. This option can be used multiple times to add more sections to run.
82 .BI \-\-alloc\-size \fR=\fPkb
83 Set the internal smalloc pool size to \fIkb\fP kilobytes.
85 .BI \-\-warnings\-fatal
86 All fio parser warnings are fatal, causing fio to exit with an error.
88 .BI \-\-max\-jobs \fR=\fPnr
89 Set the maximum allowed number of jobs (threads/processes) to support.
91 .BI \-\-server \fR=\fPargs
92 Start a backend server, with \fIargs\fP specifying what to listen to. See client/server section.
94 .BI \-\-daemonize \fR=\fPpidfile
95 Background a fio server, writing the pid to the given pid file.
97 .BI \-\-client \fR=\fPhost
98 Instead of running the jobs locally, send and run them on the given host or set of hosts. See client/server section.
100 .BI \-\-idle\-prof \fR=\fPoption
101 Report cpu idleness on a system or percpu basis (\fIoption\fP=system,percpu) or run unit work calibration only (\fIoption\fP=calibrate).
102 .SH "JOB FILE FORMAT"
103 Job files are in `ini' format. They consist of one or more
104 job definitions, which begin with a job name in square brackets and
105 extend to the next job name. The job name can be any ASCII string
106 except `global', which has a special meaning. Following the job name is
107 a sequence of zero or more parameters, one per line, that define the
108 behavior of the job. Any line starting with a `;' or `#' character is
109 considered a comment and ignored.
111 If \fIjobfile\fR is specified as `-', the job file will be read from
114 The global section contains default parameters for jobs specified in the
115 job file. A job is only affected by global sections residing above it,
116 and there may be any number of global sections. Specific job definitions
117 may override any parameter set in global sections.
120 Some parameters may take arguments of a specific type.
121 Anywhere a numeric value is required, an arithmetic expression may be used,
122 provided it is surrounded by parentheses. Supported operators are:
130 .B multiplication (*)
136 .B exponentiation (^)
140 For time values in expressions, units are microseconds by default. This is
141 different than for time values not in expressions (not enclosed in
142 parentheses). The types used are:
145 String: a sequence of alphanumeric characters.
148 SI integer: a whole number, possibly containing a suffix denoting the base unit
149 of the value. Accepted suffixes are `k', 'M', 'G', 'T', and 'P', denoting
150 kilo (1024), mega (1024^2), giga (1024^3), tera (1024^4), and peta (1024^5)
151 respectively. If prefixed with '0x', the value is assumed to be base 16
152 (hexadecimal). A suffix may include a trailing 'b', for instance 'kb' is
153 identical to 'k'. You can specify a base 10 value by using 'KiB', 'MiB','GiB',
154 etc. This is useful for disk drives where values are often given in base 10
155 values. Specifying '30GiB' will get you 30*1000^3 bytes.
156 When specifying times the default suffix meaning changes, still denoting the
157 base unit of the value, but accepted suffixes are 'D' (days), 'H' (hours), 'M'
158 (minutes), 'S' Seconds, 'ms' (or msec) milli seconds, 'us' (or 'usec') micro
159 seconds. Time values without a unit specify seconds.
160 The suffixes are not case sensitive.
163 Boolean: a true or false value. `0' denotes false, `1' denotes true.
166 Integer range: a range of integers specified in the format
167 \fIlower\fR:\fIupper\fR or \fIlower\fR\-\fIupper\fR. \fIlower\fR and
168 \fIupper\fR may contain a suffix as described above. If an option allows two
169 sets of ranges, they are separated with a `,' or `/' character. For example:
173 List of floating numbers: A list of floating numbers, separated by
178 May be used to override the job name. On the command line, this parameter
179 has the special purpose of signalling the start of a new job.
181 .BI description \fR=\fPstr
182 Human-readable description of the job. It is printed when the job is run, but
183 otherwise has no special purpose.
185 .BI directory \fR=\fPstr
186 Prefix filenames with this directory. Used to place files in a location other
188 You can specify a number of directories by separating the names with a ':'
189 character. These directories will be assigned equally distributed to job clones
190 creates with \fInumjobs\fR as long as they are using generated filenames.
191 If specific \fIfilename(s)\fR are set fio will use the first listed directory,
192 and thereby matching the \fIfilename\fR semantic which generates a file each
193 clone if not specified, but let all clones use the same if set. See
194 \fIfilename\fR for considerations regarding escaping certain characters on
197 .BI filename \fR=\fPstr
199 normally makes up a file name based on the job name, thread number, and file
200 number. If you want to share files between threads in a job or several jobs,
201 specify a \fIfilename\fR for each of them to override the default.
202 If the I/O engine is file-based, you can specify
203 a number of files by separating the names with a `:' character. `\-' is a
204 reserved name, meaning stdin or stdout, depending on the read/write direction
205 set. On Windows, disk devices are accessed as \\.\PhysicalDrive0 for the first
206 device, \\.\PhysicalDrive1 for the second etc. Note: Windows and FreeBSD
207 prevent write access to areas of the disk containing in-use data
208 (e.g. filesystems). If the wanted filename does need to include a colon, then
209 escape that with a '\\' character. For instance, if the filename is
210 "/dev/dsk/foo@3,0:c", then you would use filename="/dev/dsk/foo@3,0\\:c".
212 .BI filename_format \fR=\fPstr
213 If sharing multiple files between jobs, it is usually necessary to have
214 fio generate the exact names that you want. By default, fio will name a file
215 based on the default file format specification of
216 \fBjobname.jobnumber.filenumber\fP. With this option, that can be
217 customized. Fio will recognize and replace the following keywords in this
223 The name of the worker thread or process.
226 The incremental number of the worker thread or process.
229 The incremental number of the file for that worker thread or process.
232 To have dependent jobs share a set of files, this option can be set to
233 have fio generate filenames that are shared between the two. For instance,
234 if \fBtestfiles.$filenum\fR is specified, file number 4 for any job will
235 be named \fBtestfiles.4\fR. The default of \fB$jobname.$jobnum.$filenum\fR
236 will be used if no other format specifier is given.
240 .BI lockfile \fR=\fPstr
241 Fio defaults to not locking any files before it does IO to them. If a file or
242 file descriptor is shared, fio can serialize IO to that file to make the end
243 result consistent. This is usual for emulating real workloads that share files.
249 No locking. This is the default.
252 Only one thread or process may do IO at a time, excluding all others.
255 Read-write locking on the file. Many readers may access the file at the same
256 time, but writes get exclusive access.
260 .BI opendir \fR=\fPstr
261 Recursively open any files below directory \fIstr\fR.
263 .BI readwrite \fR=\fPstr "\fR,\fP rw" \fR=\fPstr
264 Type of I/O pattern. Accepted values are:
275 Sequential trim (Linux block devices only).
284 Random trim (Linux block devices only).
287 Mixed sequential reads and writes.
290 Mixed random reads and writes.
293 Trim and write mixed workload. Blocks will be trimmed first, then the same
294 blocks will be written to.
297 For mixed I/O, the default split is 50/50. For certain types of io the result
298 may still be skewed a bit, since the speed may be different. It is possible to
299 specify a number of IO's to do before getting a new offset, this is done by
300 appending a `:\fI<nr>\fR to the end of the string given. For a random read, it
301 would look like \fBrw=randread:8\fR for passing in an offset modifier with a
302 value of 8. If the postfix is used with a sequential IO pattern, then the value
303 specified will be added to the generated offset for each IO. For instance,
304 using \fBrw=write:4k\fR will skip 4k for every write. It turns sequential IO
305 into sequential IO with holes. See the \fBrw_sequencer\fR option.
308 .BI rw_sequencer \fR=\fPstr
309 If an offset modifier is given by appending a number to the \fBrw=<str>\fR line,
310 then this option controls how that number modifies the IO offset being
311 generated. Accepted values are:
316 Generate sequential offset
319 Generate the same offset
322 \fBsequential\fR is only useful for random IO, where fio would normally
323 generate a new random offset for every IO. If you append eg 8 to randread, you
324 would get a new random offset for every 8 IO's. The result would be a seek for
325 only every 8 IO's, instead of for every IO. Use \fBrw=randread:8\fR to specify
326 that. As sequential IO is already sequential, setting \fBsequential\fR for that
327 would not result in any differences. \fBidentical\fR behaves in a similar
328 fashion, except it sends the same offset 8 number of times before generating a
333 .BI kb_base \fR=\fPint
334 The base unit for a kilobyte. The defacto base is 2^10, 1024. Storage
335 manufacturers like to use 10^3 or 1000 as a base ten unit instead, for obvious
336 reasons. Allowed values are 1024 or 1000, with 1024 being the default.
338 .BI unified_rw_reporting \fR=\fPbool
339 Fio normally reports statistics on a per data direction basis, meaning that
340 read, write, and trim are accounted and reported separately. If this option is
341 set fio sums the results and reports them as "mixed" instead.
343 .BI randrepeat \fR=\fPbool
344 Seed the random number generator used for random I/O patterns in a predictable
345 way so the pattern is repeatable across runs. Default: true.
347 .BI allrandrepeat \fR=\fPbool
348 Seed all random number generators in a predictable way so results are
349 repeatable across runs. Default: false.
351 .BI randseed \fR=\fPint
352 Seed the random number generators based on this seed value, to be able to
353 control what sequence of output is being generated. If not set, the random
354 sequence depends on the \fBrandrepeat\fR setting.
356 .BI fallocate \fR=\fPstr
357 Whether pre-allocation is performed when laying down files. Accepted values
363 Do not pre-allocate space.
366 Pre-allocate via \fBposix_fallocate\fR\|(3).
369 Pre-allocate via \fBfallocate\fR\|(2) with FALLOC_FL_KEEP_SIZE set.
372 Backward-compatible alias for 'none'.
375 Backward-compatible alias for 'posix'.
378 May not be available on all supported platforms. 'keep' is only
379 available on Linux. If using ZFS on Solaris this must be set to 'none'
380 because ZFS doesn't support it. Default: 'posix'.
383 .BI fadvise_hint \fR=\fPbool
384 Use \fBposix_fadvise\fR\|(2) to advise the kernel what I/O patterns
385 are likely to be issued. Default: true.
387 .BI fadvise_stream \fR=\fPint
388 Use \fBposix_fadvise\fR\|(2) to advise the kernel what stream ID the
389 writes issued belong to. Only supported on Linux. Note, this option
390 may change going forward.
393 Total size of I/O for this job. \fBfio\fR will run until this many bytes have
394 been transferred, unless limited by other options (\fBruntime\fR, for instance,
395 or increased/descreased by \fBio_size\fR). Unless \fBnrfiles\fR and
396 \fBfilesize\fR options are given, this amount will be divided between the
397 available files for the job. If not set, fio will use the full size of the
398 given files or devices. If the files do not exist, size must be given. It is
399 also possible to give size as a percentage between 1 and 100. If size=20% is
400 given, fio will use 20% of the full size of the given files or devices.
402 .BI io_size \fR=\fPint "\fR,\fB io_limit \fR=\fPint
403 Normally fio operates within the region set by \fBsize\fR, which means that
404 the \fBsize\fR option sets both the region and size of IO to be performed.
405 Sometimes that is not what you want. With this option, it is possible to
406 define just the amount of IO that fio should do. For instance, if \fBsize\fR
407 is set to 20G and \fBio_limit\fR is set to 5G, fio will perform IO within
408 the first 20G but exit when 5G have been done. The opposite is also
409 possible - if \fBsize\fR is set to 20G, and \fBio_size\fR is set to 40G, then
410 fio will do 40G of IO within the 0..20G region.
412 .BI fill_device \fR=\fPbool "\fR,\fB fill_fs" \fR=\fPbool
413 Sets size to something really large and waits for ENOSPC (no space left on
414 device) as the terminating condition. Only makes sense with sequential write.
415 For a read workload, the mount point will be filled first then IO started on
416 the result. This option doesn't make sense if operating on a raw device node,
417 since the size of that is already known by the file system. Additionally,
418 writing beyond end-of-device will not return ENOSPC there.
420 .BI filesize \fR=\fPirange
421 Individual file sizes. May be a range, in which case \fBfio\fR will select sizes
422 for files at random within the given range, limited to \fBsize\fR in total (if
423 that is given). If \fBfilesize\fR is not specified, each created file is the
426 .BI file_append \fR=\fPbool
427 Perform IO after the end of the file. Normally fio will operate within the
428 size of a file. If this option is set, then fio will append to the file
429 instead. This has identical behavior to setting \fRoffset\fP to the size
430 of a file. This option is ignored on non-regular files.
432 .BI blocksize \fR=\fPint[,int] "\fR,\fB bs" \fR=\fPint[,int]
433 Block size for I/O units. Default: 4k. Values for reads, writes, and trims
434 can be specified separately in the format \fIread\fR,\fIwrite\fR,\fItrim\fR
435 either of which may be empty to leave that value at its default. If a trailing
436 comma isn't given, the remainder will inherit the last value set.
438 .BI blocksize_range \fR=\fPirange[,irange] "\fR,\fB bsrange" \fR=\fPirange[,irange]
439 Specify a range of I/O block sizes. The issued I/O unit will always be a
440 multiple of the minimum size, unless \fBblocksize_unaligned\fR is set. Applies
441 to both reads and writes if only one range is given, but can be specified
442 separately with a comma separating the values. Example: bsrange=1k-4k,2k-8k.
443 Also (see \fBblocksize\fR).
445 .BI bssplit \fR=\fPstr
446 This option allows even finer grained control of the block sizes issued,
447 not just even splits between them. With this option, you can weight various
448 block sizes for exact control of the issued IO for a job that has mixed
449 block sizes. The format of the option is bssplit=blocksize/percentage,
450 optionally adding as many definitions as needed separated by a colon.
451 Example: bssplit=4k/10:64k/50:32k/40 would issue 50% 64k blocks, 10% 4k
452 blocks and 40% 32k blocks. \fBbssplit\fR also supports giving separate
453 splits to reads and writes. The format is identical to what the
454 \fBbs\fR option accepts, the read and write parts are separated with a
457 .B blocksize_unaligned\fR,\fP bs_unaligned
458 If set, any size in \fBblocksize_range\fR may be used. This typically won't
459 work with direct I/O, as that normally requires sector alignment.
461 .BI blockalign \fR=\fPint[,int] "\fR,\fB ba" \fR=\fPint[,int]
462 At what boundary to align random IO offsets. Defaults to the same as 'blocksize'
463 the minimum blocksize given. Minimum alignment is typically 512b
464 for using direct IO, though it usually depends on the hardware block size.
465 This option is mutually exclusive with using a random map for files, so it
466 will turn off that option.
468 .BI bs_is_seq_rand \fR=\fPbool
469 If this option is set, fio will use the normal read,write blocksize settings as
470 sequential,random instead. Any random read or write will use the WRITE
471 blocksize settings, and any sequential read or write will use the READ
475 Initialize buffers with all zeros. Default: fill buffers with random data.
478 If this option is given, fio will refill the IO buffers on every submit. The
479 default is to only fill it at init time and reuse that data. Only makes sense
480 if zero_buffers isn't specified, naturally. If data verification is enabled,
481 refill_buffers is also automatically enabled.
483 .BI scramble_buffers \fR=\fPbool
484 If \fBrefill_buffers\fR is too costly and the target is using data
485 deduplication, then setting this option will slightly modify the IO buffer
486 contents to defeat normal de-dupe attempts. This is not enough to defeat
487 more clever block compression attempts, but it will stop naive dedupe
488 of blocks. Default: true.
490 .BI buffer_compress_percentage \fR=\fPint
491 If this is set, then fio will attempt to provide IO buffer content (on WRITEs)
492 that compress to the specified level. Fio does this by providing a mix of
493 random data and a fixed pattern. The fixed pattern is either zeroes, or the
494 pattern specified by \fBbuffer_pattern\fR. If the pattern option is used, it
495 might skew the compression ratio slightly. Note that this is per block size
496 unit, for file/disk wide compression level that matches this setting. Note
497 that this is per block size unit, for file/disk wide compression level that
498 matches this setting, you'll also want to set refill_buffers.
500 .BI buffer_compress_chunk \fR=\fPint
501 See \fBbuffer_compress_percentage\fR. This setting allows fio to manage how
502 big the ranges of random data and zeroed data is. Without this set, fio will
503 provide \fBbuffer_compress_percentage\fR of blocksize random data, followed by
504 the remaining zeroed. With this set to some chunk size smaller than the block
505 size, fio can alternate random and zeroed data throughout the IO buffer.
507 .BI buffer_pattern \fR=\fPstr
508 If set, fio will fill the IO buffers with this pattern. If not set, the contents
509 of IO buffers is defined by the other options related to buffer contents. The
510 setting can be any pattern of bytes, and can be prefixed with 0x for hex
511 values. It may also be a string, where the string must then be wrapped with
515 \fBbuffer_pattern\fR="abcd"
519 \fBbuffer_pattern\fR=-12
523 \fBbuffer_pattern\fR=0xdeadface
526 Also you can combine everything together in any order:
529 \fBbuffer_pattern\fR=0xdeadface"abcd"-12
533 .BI dedupe_percentage \fR=\fPint
534 If set, fio will generate this percentage of identical buffers when writing.
535 These buffers will be naturally dedupable. The contents of the buffers depend
536 on what other buffer compression settings have been set. It's possible to have
537 the individual buffers either fully compressible, or not at all. This option
538 only controls the distribution of unique buffers.
540 .BI nrfiles \fR=\fPint
541 Number of files to use for this job. Default: 1.
543 .BI openfiles \fR=\fPint
544 Number of files to keep open at the same time. Default: \fBnrfiles\fR.
546 .BI file_service_type \fR=\fPstr
547 Defines how files to service are selected. The following types are defined:
552 Choose a file at random.
555 Round robin over opened files (default).
558 Do each file in the set sequentially.
561 The number of I/Os to issue before switching to a new file can be specified by
562 appending `:\fIint\fR' to the service type.
565 .BI ioengine \fR=\fPstr
566 Defines how the job issues I/O. The following types are defined:
571 Basic \fBread\fR\|(2) or \fBwrite\fR\|(2) I/O. \fBfseek\fR\|(2) is used to
572 position the I/O location.
575 Basic \fBpread\fR\|(2) or \fBpwrite\fR\|(2) I/O.
578 Basic \fBreadv\fR\|(2) or \fBwritev\fR\|(2) I/O. Will emulate queuing by
579 coalescing adjacent IOs into a single submission.
582 Basic \fBpreadv\fR\|(2) or \fBpwritev\fR\|(2) I/O.
585 Linux native asynchronous I/O. This ioengine defines engine specific options.
588 POSIX asynchronous I/O using \fBaio_read\fR\|(3) and \fBaio_write\fR\|(3).
591 Solaris native asynchronous I/O.
594 Windows native asynchronous I/O.
597 File is memory mapped with \fBmmap\fR\|(2) and data copied using
601 \fBsplice\fR\|(2) is used to transfer the data and \fBvmsplice\fR\|(2) to
602 transfer data from user-space to the kernel.
605 Use the syslet system calls to make regular read/write asynchronous.
608 SCSI generic sg v3 I/O. May be either synchronous using the SG_IO ioctl, or if
609 the target is an sg character device, we use \fBread\fR\|(2) and
610 \fBwrite\fR\|(2) for asynchronous I/O.
613 Doesn't transfer any data, just pretends to. Mainly used to exercise \fBfio\fR
614 itself and for debugging and testing purposes.
617 Transfer over the network. The protocol to be used can be defined with the
618 \fBprotocol\fR parameter. Depending on the protocol, \fBfilename\fR,
619 \fBhostname\fR, \fBport\fR, or \fBlisten\fR must be specified.
620 This ioengine defines engine specific options.
623 Like \fBnet\fR, but uses \fBsplice\fR\|(2) and \fBvmsplice\fR\|(2) to map data
624 and send/receive. This ioengine defines engine specific options.
627 Doesn't transfer any data, but burns CPU cycles according to \fBcpuload\fR and
628 \fBcpucycles\fR parameters.
631 The GUASI I/O engine is the Generic Userspace Asynchronous Syscall Interface
632 approach to asynchronous I/O.
634 See <http://www.xmailserver.org/guasi\-lib.html>.
637 The RDMA I/O engine supports both RDMA memory semantics (RDMA_WRITE/RDMA_READ)
638 and channel semantics (Send/Recv) for the InfiniBand, RoCE and iWARP protocols.
641 Loads an external I/O engine object file. Append the engine filename as
645 IO engine that does regular linux native fallocate call to simulate data
646 transfer as fio ioengine
648 DDIR_READ does fallocate(,mode = FALLOC_FL_KEEP_SIZE,)
650 DIR_WRITE does fallocate(,mode = 0)
652 DDIR_TRIM does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE)
655 IO engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate defragment activity
656 request to DDIR_WRITE event
659 IO engine supporting direct access to Ceph Rados Block Devices (RBD) via librbd
660 without the need to use the kernel rbd driver. This ioengine defines engine specific
664 Using Glusterfs libgfapi sync interface to direct access to Glusterfs volumes without
665 having to go through FUSE. This ioengine defines engine specific
669 Using Glusterfs libgfapi async interface to direct access to Glusterfs volumes without
670 having to go through FUSE. This ioengine defines engine specific
674 Read and write through Hadoop (HDFS). The \fBfilename\fR option is used to
675 specify host,port of the hdfs name-node to connect. This engine interprets
676 offsets a little differently. In HDFS, files once created cannot be modified.
677 So random writes are not possible. To imitate this, libhdfs engine expects
678 bunch of small files to be created over HDFS, and engine will randomly pick a
679 file out of those files based on the offset generated by fio backend. (see the
680 example job file to create such files, use rw=write option). Please note, you
681 might want to set necessary environment variables to work with hdfs/libhdfs
685 Read, write and erase an MTD character device (e.g., /dev/mtd0). Discards are
686 treated as erases. Depending on the underlying device type, the I/O may have
687 to go in a certain pattern, e.g., on NAND, writing sequentially to erase blocks
688 and discarding before overwriting. The writetrim mode works well for this
694 .BI iodepth \fR=\fPint
695 Number of I/O units to keep in flight against the file. Note that increasing
696 iodepth beyond 1 will not affect synchronous ioengines (except for small
697 degress when verify_async is in use). Even async engines may impose OS
698 restrictions causing the desired depth not to be achieved. This may happen on
699 Linux when using libaio and not setting \fBdirect\fR=1, since buffered IO is
700 not async on that OS. Keep an eye on the IO depth distribution in the
701 fio output to verify that the achieved depth is as expected. Default: 1.
703 .BI iodepth_batch \fR=\fPint "\fR,\fP iodepth_batch_submit" \fR=\fPint
704 This defines how many pieces of IO to submit at once. It defaults to 1
705 which means that we submit each IO as soon as it is available, but can
706 be raised to submit bigger batches of IO at the time. If it is set to 0
707 the \fBiodepth\fR value will be used.
709 .BI iodepth_batch_complete_min \fR=\fPint "\fR,\fP iodepth_batch_complete" \fR=\fPint
710 This defines how many pieces of IO to retrieve at once. It defaults to 1 which
711 means that we'll ask for a minimum of 1 IO in the retrieval process from the
712 kernel. The IO retrieval will go on until we hit the limit set by
713 \fBiodepth_low\fR. If this variable is set to 0, then fio will always check for
714 completed events before queuing more IO. This helps reduce IO latency, at the
715 cost of more retrieval system calls.
717 .BI iodepth_batch_complete_max \fR=\fPint
718 This defines maximum pieces of IO to
719 retrieve at once. This variable should be used along with
720 \fBiodepth_batch_complete_min\fR=int variable, specifying the range
721 of min and max amount of IO which should be retrieved. By default
722 it is equal to \fBiodepth_batch_complete_min\fR value.
727 \fBiodepth_batch_complete_min\fR=1
729 \fBiodepth_batch_complete_max\fR=<iodepth>
732 which means that we will retrieve at leat 1 IO and up to the
733 whole submitted queue depth. If none of IO has been completed
738 \fBiodepth_batch_complete_min\fR=0
740 \fBiodepth_batch_complete_max\fR=<iodepth>
743 which means that we can retrieve up to the whole submitted
744 queue depth, but if none of IO has been completed yet, we will
745 NOT wait and immediately exit the system call. In this example
746 we simply do polling.
749 .BI iodepth_low \fR=\fPint
750 Low watermark indicating when to start filling the queue again. Default:
753 .BI io_submit_mode \fR=\fPstr
754 This option controls how fio submits the IO to the IO engine. The default is
755 \fBinline\fR, which means that the fio job threads submit and reap IO directly.
756 If set to \fBoffload\fR, the job threads will offload IO submission to a
757 dedicated pool of IO threads. This requires some coordination and thus has a
758 bit of extra overhead, especially for lower queue depth IO where it can
759 increase latencies. The benefit is that fio can manage submission rates
760 independently of the device completion rates. This avoids skewed latency
761 reporting if IO gets back up on the device side (the coordinated omission
764 .BI direct \fR=\fPbool
765 If true, use non-buffered I/O (usually O_DIRECT). Default: false.
767 .BI atomic \fR=\fPbool
768 If value is true, attempt to use atomic direct IO. Atomic writes are guaranteed
769 to be stable once acknowledged by the operating system. Only Linux supports
772 .BI buffered \fR=\fPbool
773 If true, use buffered I/O. This is the opposite of the \fBdirect\fR parameter.
776 .BI offset \fR=\fPint
777 Offset in the file to start I/O. Data before the offset will not be touched.
779 .BI offset_increment \fR=\fPint
780 If this is provided, then the real offset becomes the
781 offset + offset_increment * thread_number, where the thread number is a
782 counter that starts at 0 and is incremented for each sub-job (i.e. when
783 numjobs option is specified). This option is useful if there are several jobs
784 which are intended to operate on a file in parallel disjoint segments, with
785 even spacing between the starting points.
787 .BI number_ios \fR=\fPint
788 Fio will normally perform IOs until it has exhausted the size of the region
789 set by \fBsize\fR, or if it exhaust the allocated time (or hits an error
790 condition). With this setting, the range/size can be set independently of
791 the number of IOs to perform. When fio reaches this number, it will exit
792 normally and report status. Note that this does not extend the amount
793 of IO that will be done, it will only stop fio if this condition is met
794 before other end-of-job criteria.
797 How many I/Os to perform before issuing an \fBfsync\fR\|(2) of dirty data. If
798 0, don't sync. Default: 0.
800 .BI fdatasync \fR=\fPint
801 Like \fBfsync\fR, but uses \fBfdatasync\fR\|(2) instead to only sync the
802 data parts of the file. Default: 0.
804 .BI write_barrier \fR=\fPint
805 Make every Nth write a barrier write.
807 .BI sync_file_range \fR=\fPstr:int
808 Use \fBsync_file_range\fR\|(2) for every \fRval\fP number of write operations. Fio will
809 track range of writes that have happened since the last \fBsync_file_range\fR\|(2) call.
810 \fRstr\fP can currently be one or more of:
814 SYNC_FILE_RANGE_WAIT_BEFORE
817 SYNC_FILE_RANGE_WRITE
820 SYNC_FILE_RANGE_WRITE
824 So if you do sync_file_range=wait_before,write:8, fio would use
825 \fBSYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE\fP for every 8 writes.
826 Also see the \fBsync_file_range\fR\|(2) man page. This option is Linux specific.
828 .BI overwrite \fR=\fPbool
829 If writing, setup the file first and do overwrites. Default: false.
831 .BI end_fsync \fR=\fPbool
832 Sync file contents when a write stage has completed. Default: false.
834 .BI fsync_on_close \fR=\fPbool
835 If true, sync file contents on close. This differs from \fBend_fsync\fR in that
836 it will happen on every close, not just at the end of the job. Default: false.
838 .BI rwmixread \fR=\fPint
839 Percentage of a mixed workload that should be reads. Default: 50.
841 .BI rwmixwrite \fR=\fPint
842 Percentage of a mixed workload that should be writes. If \fBrwmixread\fR and
843 \fBrwmixwrite\fR are given and do not sum to 100%, the latter of the two
844 overrides the first. This may interfere with a given rate setting, if fio is
845 asked to limit reads or writes to a certain rate. If that is the case, then
846 the distribution may be skewed. Default: 50.
848 .BI random_distribution \fR=\fPstr:float
849 By default, fio will use a completely uniform random distribution when asked
850 to perform random IO. Sometimes it is useful to skew the distribution in
851 specific ways, ensuring that some parts of the data is more hot than others.
852 Fio includes the following distribution models:
856 Uniform random distribution
866 When using a zipf or pareto distribution, an input value is also needed to
867 define the access pattern. For zipf, this is the zipf theta. For pareto,
868 it's the pareto power. Fio includes a test program, genzipf, that can be
869 used visualize what the given input values will yield in terms of hit rates.
870 If you wanted to use zipf with a theta of 1.2, you would use
871 random_distribution=zipf:1.2 as the option. If a non-uniform model is used,
872 fio will disable use of the random map.
874 .BI percentage_random \fR=\fPint
875 For a random workload, set how big a percentage should be random. This defaults
876 to 100%, in which case the workload is fully random. It can be set from
877 anywhere from 0 to 100. Setting it to 0 would make the workload fully
878 sequential. It is possible to set different values for reads, writes, and
879 trim. To do so, simply use a comma separated list. See \fBblocksize\fR.
882 Normally \fBfio\fR will cover every block of the file when doing random I/O. If
883 this parameter is given, a new offset will be chosen without looking at past
884 I/O history. This parameter is mutually exclusive with \fBverify\fR.
886 .BI softrandommap \fR=\fPbool
887 See \fBnorandommap\fR. If fio runs with the random block map enabled and it
888 fails to allocate the map, if this option is set it will continue without a
889 random block map. As coverage will not be as complete as with random maps, this
890 option is disabled by default.
892 .BI random_generator \fR=\fPstr
893 Fio supports the following engines for generating IO offsets for random IO:
897 Strong 2^88 cycle random number generator
900 Linear feedback shift register generator
903 Strong 64-bit 2^258 cycle random number generator
907 Tausworthe is a strong random number generator, but it requires tracking on the
908 side if we want to ensure that blocks are only read or written once. LFSR
909 guarantees that we never generate the same offset twice, and it's also less
910 computationally expensive. It's not a true random generator, however, though
911 for IO purposes it's typically good enough. LFSR only works with single block
912 sizes, not with workloads that use multiple block sizes. If used with such a
913 workload, fio may read or write some blocks multiple times.
916 Run job with given nice value. See \fBnice\fR\|(2).
919 Set I/O priority value of this job between 0 (highest) and 7 (lowest). See
922 .BI prioclass \fR=\fPint
923 Set I/O priority class. See \fBionice\fR\|(1).
925 .BI thinktime \fR=\fPint
926 Stall job for given number of microseconds between issuing I/Os.
928 .BI thinktime_spin \fR=\fPint
929 Pretend to spend CPU time for given number of microseconds, sleeping the rest
930 of the time specified by \fBthinktime\fR. Only valid if \fBthinktime\fR is set.
932 .BI thinktime_blocks \fR=\fPint
933 Only valid if thinktime is set - control how many blocks to issue, before
934 waiting \fBthinktime\fR microseconds. If not set, defaults to 1 which will
935 make fio wait \fBthinktime\fR microseconds after every block. This
936 effectively makes any queue depth setting redundant, since no more than 1 IO
937 will be queued before we have to complete it and do our thinktime. In other
938 words, this setting effectively caps the queue depth if the latter is larger.
942 Cap bandwidth used by this job. The number is in bytes/sec, the normal postfix
943 rules apply. You can use \fBrate\fR=500k to limit reads and writes to 500k each,
944 or you can specify read and writes separately. Using \fBrate\fR=1m,500k would
945 limit reads to 1MB/sec and writes to 500KB/sec. Capping only reads or writes
946 can be done with \fBrate\fR=,500k or \fBrate\fR=500k,. The former will only
947 limit writes (to 500KB/sec), the latter will only limit reads.
949 .BI ratemin \fR=\fPint
950 Tell \fBfio\fR to do whatever it can to maintain at least the given bandwidth.
951 Failing to meet this requirement will cause the job to exit. The same format
952 as \fBrate\fR is used for read vs write separation.
954 .BI rate_iops \fR=\fPint
955 Cap the bandwidth to this number of IOPS. Basically the same as rate, just
956 specified independently of bandwidth. The same format as \fBrate\fR is used for
957 read vs write separation. If \fBblocksize\fR is a range, the smallest block
958 size is used as the metric.
960 .BI rate_iops_min \fR=\fPint
961 If this rate of I/O is not met, the job will exit. The same format as \fBrate\fR
962 is used for read vs write separation.
964 .BI ratecycle \fR=\fPint
965 Average bandwidth for \fBrate\fR and \fBratemin\fR over this number of
966 milliseconds. Default: 1000ms.
968 .BI latency_target \fR=\fPint
969 If set, fio will attempt to find the max performance point that the given
970 workload will run at while maintaining a latency below this target. The
971 values is given in microseconds. See \fBlatency_window\fR and
972 \fBlatency_percentile\fR.
974 .BI latency_window \fR=\fPint
975 Used with \fBlatency_target\fR to specify the sample window that the job
976 is run at varying queue depths to test the performance. The value is given
979 .BI latency_percentile \fR=\fPfloat
980 The percentage of IOs that must fall within the criteria specified by
981 \fBlatency_target\fR and \fBlatency_window\fR. If not set, this defaults
982 to 100.0, meaning that all IOs must be equal or below to the value set
983 by \fBlatency_target\fR.
985 .BI max_latency \fR=\fPint
986 If set, fio will exit the job if it exceeds this maximum latency. It will exit
989 .BI cpumask \fR=\fPint
990 Set CPU affinity for this job. \fIint\fR is a bitmask of allowed CPUs the job
991 may run on. See \fBsched_setaffinity\fR\|(2).
993 .BI cpus_allowed \fR=\fPstr
994 Same as \fBcpumask\fR, but allows a comma-delimited list of CPU numbers.
996 .BI cpus_allowed_policy \fR=\fPstr
997 Set the policy of how fio distributes the CPUs specified by \fBcpus_allowed\fR
998 or \fBcpumask\fR. Two policies are supported:
1003 All jobs will share the CPU set specified.
1006 Each job will get a unique CPU from the CPU set.
1009 \fBshared\fR is the default behaviour, if the option isn't specified. If
1010 \fBsplit\fR is specified, then fio will assign one cpu per job. If not enough
1011 CPUs are given for the jobs listed, then fio will roundrobin the CPUs in
1016 .BI numa_cpu_nodes \fR=\fPstr
1017 Set this job running on specified NUMA nodes' CPUs. The arguments allow
1018 comma delimited list of cpu numbers, A-B ranges, or 'all'.
1020 .BI numa_mem_policy \fR=\fPstr
1021 Set this job's memory policy and corresponding NUMA nodes. Format of
1025 .B <mode>[:<nodelist>]
1028 is one of the following memory policy:
1030 .B default, prefer, bind, interleave, local
1033 For \fBdefault\fR and \fBlocal\fR memory policy, no \fBnodelist\fR is
1034 needed to be specified. For \fBprefer\fR, only one node is
1035 allowed. For \fBbind\fR and \fBinterleave\fR, \fBnodelist\fR allows
1036 comma delimited list of numbers, A-B ranges, or 'all'.
1038 .BI startdelay \fR=\fPirange
1039 Delay start of job for the specified number of seconds. Supports all time
1040 suffixes to allow specification of hours, minutes, seconds and
1041 milliseconds - seconds are the default if a unit is omitted.
1042 Can be given as a range which causes each thread to choose randomly out of the
1045 .BI runtime \fR=\fPint
1046 Terminate processing after the specified number of seconds.
1049 If given, run for the specified \fBruntime\fR duration even if the files are
1050 completely read or written. The same workload will be repeated as many times
1051 as \fBruntime\fR allows.
1053 .BI ramp_time \fR=\fPint
1054 If set, fio will run the specified workload for this amount of time before
1055 logging any performance numbers. Useful for letting performance settle before
1056 logging results, thus minimizing the runtime required for stable results. Note
1057 that the \fBramp_time\fR is considered lead in time for a job, thus it will
1058 increase the total runtime if a special timeout or runtime is specified.
1060 .BI invalidate \fR=\fPbool
1061 Invalidate buffer-cache for the file prior to starting I/O. Default: true.
1063 .BI sync \fR=\fPbool
1064 Use synchronous I/O for buffered writes. For the majority of I/O engines,
1065 this means using O_SYNC. Default: false.
1067 .BI iomem \fR=\fPstr "\fR,\fP mem" \fR=\fPstr
1068 Allocation method for I/O unit buffer. Allowed values are:
1073 Allocate memory with \fBmalloc\fR\|(3).
1076 Use shared memory buffers allocated through \fBshmget\fR\|(2).
1079 Same as \fBshm\fR, but use huge pages as backing.
1082 Use \fBmmap\fR\|(2) for allocation. Uses anonymous memory unless a filename
1083 is given after the option in the format `:\fIfile\fR'.
1086 Same as \fBmmap\fR, but use huge files as backing.
1089 The amount of memory allocated is the maximum allowed \fBblocksize\fR for the
1090 job multiplied by \fBiodepth\fR. For \fBshmhuge\fR or \fBmmaphuge\fR to work,
1091 the system must have free huge pages allocated. \fBmmaphuge\fR also needs to
1092 have hugetlbfs mounted, and \fIfile\fR must point there. At least on Linux,
1093 huge pages must be manually allocated. See \fB/proc/sys/vm/nr_hugehages\fR
1094 and the documentation for that. Normally you just need to echo an appropriate
1095 number, eg echoing 8 will ensure that the OS has 8 huge pages ready for
1099 .BI iomem_align \fR=\fPint "\fR,\fP mem_align" \fR=\fPint
1100 This indicates the memory alignment of the IO memory buffers. Note that the
1101 given alignment is applied to the first IO unit buffer, if using \fBiodepth\fR
1102 the alignment of the following buffers are given by the \fBbs\fR used. In
1103 other words, if using a \fBbs\fR that is a multiple of the page sized in the
1104 system, all buffers will be aligned to this value. If using a \fBbs\fR that
1105 is not page aligned, the alignment of subsequent IO memory buffers is the
1106 sum of the \fBiomem_align\fR and \fBbs\fR used.
1108 .BI hugepage\-size \fR=\fPint
1109 Defines the size of a huge page. Must be at least equal to the system setting.
1110 Should be a multiple of 1MB. Default: 4MB.
1113 Terminate all jobs when one finishes. Default: wait for each job to finish.
1115 .BI bwavgtime \fR=\fPint
1116 Average bandwidth calculations over the given time in milliseconds. Default:
1119 .BI iopsavgtime \fR=\fPint
1120 Average IOPS calculations over the given time in milliseconds. Default:
1123 .BI create_serialize \fR=\fPbool
1124 If true, serialize file creation for the jobs. Default: true.
1126 .BI create_fsync \fR=\fPbool
1127 \fBfsync\fR\|(2) data file after creation. Default: true.
1129 .BI create_on_open \fR=\fPbool
1130 If true, the files are not created until they are opened for IO by the job.
1132 .BI create_only \fR=\fPbool
1133 If true, fio will only run the setup phase of the job. If files need to be
1134 laid out or updated on disk, only that will be done. The actual job contents
1137 .BI allow_file_create \fR=\fPbool
1138 If true, fio is permitted to create files as part of its workload. This is
1139 the default behavior. If this option is false, then fio will error out if the
1140 files it needs to use don't already exist. Default: true.
1142 .BI allow_mounted_write \fR=\fPbool
1143 If this isn't set, fio will abort jobs that are destructive (eg that write)
1144 to what appears to be a mounted device or partition. This should help catch
1145 creating inadvertently destructive tests, not realizing that the test will
1146 destroy data on the mounted file system. Default: false.
1148 .BI pre_read \fR=\fPbool
1149 If this is given, files will be pre-read into memory before starting the given
1150 IO operation. This will also clear the \fR \fBinvalidate\fR flag, since it is
1151 pointless to pre-read and then drop the cache. This will only work for IO
1152 engines that are seekable, since they allow you to read the same data
1153 multiple times. Thus it will not work on eg network or splice IO.
1155 .BI unlink \fR=\fPbool
1156 Unlink job files when done. Default: false.
1158 .BI loops \fR=\fPint
1159 Specifies the number of iterations (runs of the same workload) of this job.
1162 .BI verify_only \fR=\fPbool
1163 Do not perform the specified workload, only verify data still matches previous
1164 invocation of this workload. This option allows one to check data multiple
1165 times at a later date without overwriting it. This option makes sense only for
1166 workloads that write data, and does not support workloads with the
1167 \fBtime_based\fR option set.
1169 .BI do_verify \fR=\fPbool
1170 Run the verify phase after a write phase. Only valid if \fBverify\fR is set.
1173 .BI verify \fR=\fPstr
1174 Method of verifying file contents after each iteration of the job. Each
1175 verification method also implies verification of special header, which is
1176 written to the beginning of each block. This header also includes meta
1177 information, like offset of the block, block number, timestamp when block
1178 was written, etc. \fBverify\fR=str can be combined with \fBverify_pattern\fR=str
1179 option. The allowed values are:
1183 .B md5 crc16 crc32 crc32c crc32c-intel crc64 crc7 sha256 sha512 sha1 xxhash
1184 Store appropriate checksum in the header of each block. crc32c-intel is
1185 hardware accelerated SSE4.2 driven, falls back to regular crc32c if
1186 not supported by the system.
1189 This option is deprecated, since now meta information is included in generic
1190 verification header and meta verification happens by default. For detailed
1191 information see the description of the \fBverify\fR=str setting. This option
1192 is kept because of compatibility's sake with old configurations. Do not use it.
1195 Verify a strict pattern. Normally fio includes a header with some basic
1196 information and checksumming, but if this option is set, only the
1197 specific pattern set with \fBverify_pattern\fR is verified.
1200 Pretend to verify. Used for testing internals.
1203 This option can be used for repeated burn-in tests of a system to make sure
1204 that the written data is also correctly read back. If the data direction given
1205 is a read or random read, fio will assume that it should verify a previously
1206 written file. If the data direction includes any form of write, the verify will
1207 be of the newly written data.
1210 .BI verifysort \fR=\fPbool
1211 If true, written verify blocks are sorted if \fBfio\fR deems it to be faster to
1212 read them back in a sorted manner. Default: true.
1214 .BI verifysort_nr \fR=\fPint
1215 Pre-load and sort verify blocks for a read workload.
1217 .BI verify_offset \fR=\fPint
1218 Swap the verification header with data somewhere else in the block before
1219 writing. It is swapped back before verifying.
1221 .BI verify_interval \fR=\fPint
1222 Write the verification header for this number of bytes, which should divide
1223 \fBblocksize\fR. Default: \fBblocksize\fR.
1225 .BI verify_pattern \fR=\fPstr
1226 If set, fio will fill the io buffers with this pattern. Fio defaults to filling
1227 with totally random bytes, but sometimes it's interesting to fill with a known
1228 pattern for io verification purposes. Depending on the width of the pattern,
1229 fio will fill 1/2/3/4 bytes of the buffer at the time(it can be either a
1230 decimal or a hex number). The verify_pattern if larger than a 32-bit quantity
1231 has to be a hex number that starts with either "0x" or "0X". Use with
1232 \fBverify\fP=str. Also, verify_pattern supports %o format, which means that for
1233 each block offset will be written and then verifyied back, e.g.:
1236 \fBverify_pattern\fR=%o
1238 Or use combination of everything:
1241 \fBverify_pattern\fR=0xff%o"abcd"-21
1245 .BI verify_fatal \fR=\fPbool
1246 If true, exit the job on the first observed verification failure. Default:
1249 .BI verify_dump \fR=\fPbool
1250 If set, dump the contents of both the original data block and the data block we
1251 read off disk to files. This allows later analysis to inspect just what kind of
1252 data corruption occurred. Off by default.
1254 .BI verify_async \fR=\fPint
1255 Fio will normally verify IO inline from the submitting thread. This option
1256 takes an integer describing how many async offload threads to create for IO
1257 verification instead, causing fio to offload the duty of verifying IO contents
1258 to one or more separate threads. If using this offload option, even sync IO
1259 engines can benefit from using an \fBiodepth\fR setting higher than 1, as it
1260 allows them to have IO in flight while verifies are running.
1262 .BI verify_async_cpus \fR=\fPstr
1263 Tell fio to set the given CPU affinity on the async IO verification threads.
1264 See \fBcpus_allowed\fP for the format used.
1266 .BI verify_backlog \fR=\fPint
1267 Fio will normally verify the written contents of a job that utilizes verify
1268 once that job has completed. In other words, everything is written then
1269 everything is read back and verified. You may want to verify continually
1270 instead for a variety of reasons. Fio stores the meta data associated with an
1271 IO block in memory, so for large verify workloads, quite a bit of memory would
1272 be used up holding this meta data. If this option is enabled, fio will write
1273 only N blocks before verifying these blocks.
1275 .BI verify_backlog_batch \fR=\fPint
1276 Control how many blocks fio will verify if verify_backlog is set. If not set,
1277 will default to the value of \fBverify_backlog\fR (meaning the entire queue is
1278 read back and verified). If \fBverify_backlog_batch\fR is less than
1279 \fBverify_backlog\fR then not all blocks will be verified, if
1280 \fBverify_backlog_batch\fR is larger than \fBverify_backlog\fR, some blocks
1281 will be verified more than once.
1283 .BI trim_percentage \fR=\fPint
1284 Number of verify blocks to discard/trim.
1286 .BI trim_verify_zero \fR=\fPbool
1287 Verify that trim/discarded blocks are returned as zeroes.
1289 .BI trim_backlog \fR=\fPint
1290 Trim after this number of blocks are written.
1292 .BI trim_backlog_batch \fR=\fPint
1293 Trim this number of IO blocks.
1295 .BI experimental_verify \fR=\fPbool
1296 Enable experimental verification.
1298 .BI verify_state_save \fR=\fPbool
1299 When a job exits during the write phase of a verify workload, save its
1300 current state. This allows fio to replay up until that point, if the
1301 verify state is loaded for the verify read phase.
1303 .BI verify_state_load \fR=\fPbool
1304 If a verify termination trigger was used, fio stores the current write
1305 state of each thread. This can be used at verification time so that fio
1306 knows how far it should verify. Without this information, fio will run
1307 a full verification pass, according to the settings in the job file used.
1309 .B stonewall "\fR,\fP wait_for_previous"
1310 Wait for preceding jobs in the job file to exit before starting this one.
1311 \fBstonewall\fR implies \fBnew_group\fR.
1314 Start a new reporting group. If not given, all jobs in a file will be part
1315 of the same reporting group, unless separated by a stonewall.
1317 .BI numjobs \fR=\fPint
1318 Number of clones (processes/threads performing the same workload) of this job.
1322 If set, display per-group reports instead of per-job when \fBnumjobs\fR is
1326 Use threads created with \fBpthread_create\fR\|(3) instead of processes created
1327 with \fBfork\fR\|(2).
1329 .BI zonesize \fR=\fPint
1330 Divide file into zones of the specified size in bytes. See \fBzoneskip\fR.
1332 .BI zonerange \fR=\fPint
1333 Give size of an IO zone. See \fBzoneskip\fR.
1335 .BI zoneskip \fR=\fPint
1336 Skip the specified number of bytes when \fBzonesize\fR bytes of data have been
1339 .BI write_iolog \fR=\fPstr
1340 Write the issued I/O patterns to the specified file. Specify a separate file
1341 for each job, otherwise the iologs will be interspersed and the file may be
1344 .BI read_iolog \fR=\fPstr
1345 Replay the I/O patterns contained in the specified file generated by
1346 \fBwrite_iolog\fR, or may be a \fBblktrace\fR binary file.
1348 .BI replay_no_stall \fR=\fPint
1349 While replaying I/O patterns using \fBread_iolog\fR the default behavior
1350 attempts to respect timing information between I/Os. Enabling
1351 \fBreplay_no_stall\fR causes I/Os to be replayed as fast as possible while
1352 still respecting ordering.
1354 .BI replay_redirect \fR=\fPstr
1355 While replaying I/O patterns using \fBread_iolog\fR the default behavior
1356 is to replay the IOPS onto the major/minor device that each IOP was recorded
1357 from. Setting \fBreplay_redirect\fR causes all IOPS to be replayed onto the
1358 single specified device regardless of the device it was recorded from.
1360 .BI replay_align \fR=\fPint
1361 Force alignment of IO offsets and lengths in a trace to this power of 2 value.
1363 .BI replay_scale \fR=\fPint
1364 Scale sector offsets down by this factor when replaying traces.
1366 .BI per_job_logs \fR=\fPbool
1367 If set, this generates bw/clat/iops log with per file private filenames. If
1368 not set, jobs with identical names will share the log filename. Default: true.
1370 .BI write_bw_log \fR=\fPstr
1371 If given, write a bandwidth log of the jobs in this job file. Can be used to
1372 store data of the bandwidth of the jobs in their lifetime. The included
1373 fio_generate_plots script uses gnuplot to turn these text files into nice
1374 graphs. See \fBwrite_lat_log\fR for behaviour of given filename. For this
1375 option, the postfix is _bw.x.log, where x is the index of the job (1..N,
1376 where N is the number of jobs). If \fBper_job_logs\fR is false, then the
1377 filename will not include the job index.
1379 .BI write_lat_log \fR=\fPstr
1380 Same as \fBwrite_bw_log\fR, but writes I/O completion latencies. If no
1381 filename is given with this option, the default filename of
1382 "jobname_type.x.log" is used, where x is the index of the job (1..N, where
1383 N is the number of jobs). Even if the filename is given, fio will still
1384 append the type of log. If \fBper_job_logs\fR is false, then the filename will
1385 not include the job index.
1387 .BI write_iops_log \fR=\fPstr
1388 Same as \fBwrite_bw_log\fR, but writes IOPS. If no filename is given with this
1389 option, the default filename of "jobname_type.x.log" is used, where x is the
1390 index of the job (1..N, where N is the number of jobs). Even if the filename
1391 is given, fio will still append the type of log. If \fBper_job_logs\fR is false,
1392 then the filename will not include the job index.
1394 .BI log_avg_msec \fR=\fPint
1395 By default, fio will log an entry in the iops, latency, or bw log for every
1396 IO that completes. When writing to the disk log, that can quickly grow to a
1397 very large size. Setting this option makes fio average the each log entry
1398 over the specified period of time, reducing the resolution of the log.
1401 .BI log_offset \fR=\fPbool
1402 If this is set, the iolog options will include the byte offset for the IO
1403 entry as well as the other data values.
1405 .BI log_compression \fR=\fPint
1406 If this is set, fio will compress the IO logs as it goes, to keep the memory
1407 footprint lower. When a log reaches the specified size, that chunk is removed
1408 and compressed in the background. Given that IO logs are fairly highly
1409 compressible, this yields a nice memory savings for longer runs. The downside
1410 is that the compression will consume some background CPU cycles, so it may
1411 impact the run. This, however, is also true if the logging ends up consuming
1412 most of the system memory. So pick your poison. The IO logs are saved
1413 normally at the end of a run, by decompressing the chunks and storing them
1414 in the specified log file. This feature depends on the availability of zlib.
1416 .BI log_store_compressed \fR=\fPbool
1417 If set, and \fBlog\fR_compression is also set, fio will store the log files in
1418 a compressed format. They can be decompressed with fio, using the
1419 \fB\-\-inflate-log\fR command line parameter. The files will be stored with a
1422 .BI block_error_percentiles \fR=\fPbool
1423 If set, record errors in trim block-sized units from writes and trims and output
1424 a histogram of how many trims it took to get to errors, and what kind of error
1427 .BI disable_lat \fR=\fPbool
1428 Disable measurements of total latency numbers. Useful only for cutting
1429 back the number of calls to \fBgettimeofday\fR\|(2), as that does impact performance at
1430 really high IOPS rates. Note that to really get rid of a large amount of these
1431 calls, this option must be used with disable_slat and disable_bw as well.
1433 .BI disable_clat \fR=\fPbool
1434 Disable measurements of completion latency numbers. See \fBdisable_lat\fR.
1436 .BI disable_slat \fR=\fPbool
1437 Disable measurements of submission latency numbers. See \fBdisable_lat\fR.
1439 .BI disable_bw_measurement \fR=\fPbool
1440 Disable measurements of throughput/bandwidth numbers. See \fBdisable_lat\fR.
1442 .BI lockmem \fR=\fPint
1443 Pin the specified amount of memory with \fBmlock\fR\|(2). Can be used to
1444 simulate a smaller amount of memory. The amount specified is per worker.
1446 .BI exec_prerun \fR=\fPstr
1447 Before running the job, execute the specified command with \fBsystem\fR\|(3).
1449 Output is redirected in a file called \fBjobname.prerun.txt\fR
1452 .BI exec_postrun \fR=\fPstr
1453 Same as \fBexec_prerun\fR, but the command is executed after the job completes.
1455 Output is redirected in a file called \fBjobname.postrun.txt\fR
1458 .BI ioscheduler \fR=\fPstr
1459 Attempt to switch the device hosting the file to the specified I/O scheduler.
1461 .BI disk_util \fR=\fPbool
1462 Generate disk utilization statistics if the platform supports it. Default: true.
1464 .BI clocksource \fR=\fPstr
1465 Use the given clocksource as the base of timing. The supported options are:
1469 \fBgettimeofday\fR\|(2)
1472 \fBclock_gettime\fR\|(2)
1475 Internal CPU clock source
1479 \fBcpu\fR is the preferred clocksource if it is reliable, as it is very fast
1480 (and fio is heavy on time calls). Fio will automatically use this clocksource
1481 if it's supported and considered reliable on the system it is running on,
1482 unless another clocksource is specifically set. For x86/x86-64 CPUs, this
1483 means supporting TSC Invariant.
1485 .BI gtod_reduce \fR=\fPbool
1486 Enable all of the \fBgettimeofday\fR\|(2) reducing options (disable_clat, disable_slat,
1487 disable_bw) plus reduce precision of the timeout somewhat to really shrink the
1488 \fBgettimeofday\fR\|(2) call count. With this option enabled, we only do about 0.4% of
1489 the gtod() calls we would have done if all time keeping was enabled.
1491 .BI gtod_cpu \fR=\fPint
1492 Sometimes it's cheaper to dedicate a single thread of execution to just getting
1493 the current time. Fio (and databases, for instance) are very intensive on
1494 \fBgettimeofday\fR\|(2) calls. With this option, you can set one CPU aside for doing
1495 nothing but logging current time to a shared memory location. Then the other
1496 threads/processes that run IO workloads need only copy that segment, instead of
1497 entering the kernel with a \fBgettimeofday\fR\|(2) call. The CPU set aside for doing
1498 these time calls will be excluded from other uses. Fio will manually clear it
1499 from the CPU mask of other jobs.
1501 .BI ignore_error \fR=\fPstr
1502 Sometimes you want to ignore some errors during test in that case you can specify
1503 error list for each error type.
1505 ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST
1507 errors for given error type is separated with ':'.
1508 Error may be symbol ('ENOSPC', 'ENOMEM') or an integer.
1510 Example: ignore_error=EAGAIN,ENOSPC:122 .
1512 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from WRITE.
1514 .BI error_dump \fR=\fPbool
1515 If set dump every error even if it is non fatal, true by default. If disabled
1516 only fatal error will be dumped
1518 .BI profile \fR=\fPstr
1519 Select a specific builtin performance test.
1521 .BI cgroup \fR=\fPstr
1522 Add job to this control group. If it doesn't exist, it will be created.
1523 The system must have a mounted cgroup blkio mount point for this to work. If
1524 your system doesn't have it mounted, you can do so with:
1526 # mount \-t cgroup \-o blkio none /cgroup
1528 .BI cgroup_weight \fR=\fPint
1529 Set the weight of the cgroup to this value. See the documentation that comes
1530 with the kernel, allowed values are in the range of 100..1000.
1532 .BI cgroup_nodelete \fR=\fPbool
1533 Normally fio will delete the cgroups it has created after the job completion.
1534 To override this behavior and to leave cgroups around after the job completion,
1535 set cgroup_nodelete=1. This can be useful if one wants to inspect various
1536 cgroup files after job completion. Default: false
1539 Instead of running as the invoking user, set the user ID to this value before
1540 the thread/process does any work.
1543 Set group ID, see \fBuid\fR.
1545 .BI unit_base \fR=\fPint
1546 Base unit for reporting. Allowed values are:
1550 Use auto-detection (default).
1560 .BI flow_id \fR=\fPint
1561 The ID of the flow. If not specified, it defaults to being a global flow. See
1565 Weight in token-based flow control. If this value is used, then there is a
1566 \fBflow counter\fR which is used to regulate the proportion of activity between
1567 two or more jobs. fio attempts to keep this flow counter near zero. The
1568 \fBflow\fR parameter stands for how much should be added or subtracted to the
1569 flow counter on each iteration of the main I/O loop. That is, if one job has
1570 \fBflow=8\fR and another job has \fBflow=-1\fR, then there will be a roughly
1571 1:8 ratio in how much one runs vs the other.
1573 .BI flow_watermark \fR=\fPint
1574 The maximum value that the absolute value of the flow counter is allowed to
1575 reach before the job must wait for a lower value of the counter.
1577 .BI flow_sleep \fR=\fPint
1578 The period of time, in microseconds, to wait after the flow watermark has been
1579 exceeded before retrying operations
1581 .BI clat_percentiles \fR=\fPbool
1582 Enable the reporting of percentiles of completion latencies.
1584 .BI percentile_list \fR=\fPfloat_list
1585 Overwrite the default list of percentiles for completion latencies and the
1586 block error histogram. Each number is a floating number in the range (0,100],
1587 and the maximum length of the list is 20. Use ':' to separate the
1588 numbers. For example, \-\-percentile_list=99.5:99.9 will cause fio to
1589 report the values of completion latency below which 99.5% and 99.9% of
1590 the observed latencies fell, respectively.
1591 .SS "Ioengine Parameters List"
1592 Some parameters are only valid when a specific ioengine is in use. These are
1593 used identically to normal parameters, with the caveat that when used on the
1594 command line, they must come after the ioengine.
1596 .BI (cpu)cpuload \fR=\fPint
1597 Attempt to use the specified percentage of CPU cycles.
1599 .BI (cpu)cpuchunks \fR=\fPint
1600 Split the load into cycles of the given time. In microseconds.
1602 .BI (cpu)exit_on_io_done \fR=\fPbool
1603 Detect when IO threads are done, then exit.
1605 .BI (libaio)userspace_reap
1606 Normally, with the libaio engine in use, fio will use
1607 the io_getevents system call to reap newly returned events.
1608 With this flag turned on, the AIO ring will be read directly
1609 from user-space to reap events. The reaping mode is only
1610 enabled when polling for a minimum of 0 events (eg when
1611 iodepth_batch_complete=0).
1613 .BI (net,netsplice)hostname \fR=\fPstr
1614 The host name or IP address to use for TCP or UDP based IO.
1615 If the job is a TCP listener or UDP reader, the hostname is not
1616 used and must be omitted unless it is a valid UDP multicast address.
1618 .BI (net,netsplice)port \fR=\fPint
1619 The TCP or UDP port to bind to or connect to. If this is used with
1620 \fBnumjobs\fR to spawn multiple instances of the same job type, then
1621 this will be the starting port number since fio will use a range of ports.
1623 .BI (net,netsplice)interface \fR=\fPstr
1624 The IP address of the network interface used to send or receive UDP multicast
1627 .BI (net,netsplice)ttl \fR=\fPint
1628 Time-to-live value for outgoing UDP multicast packets. Default: 1
1630 .BI (net,netsplice)nodelay \fR=\fPbool
1631 Set TCP_NODELAY on TCP connections.
1633 .BI (net,netsplice)protocol \fR=\fPstr "\fR,\fP proto" \fR=\fPstr
1634 The network protocol to use. Accepted values are:
1639 Transmission control protocol
1642 Transmission control protocol V6
1645 User datagram protocol
1648 User datagram protocol V6
1654 When the protocol is TCP or UDP, the port must also be given,
1655 as well as the hostname if the job is a TCP listener or UDP
1656 reader. For unix sockets, the normal filename option should be
1657 used and the port is invalid.
1660 .BI (net,netsplice)listen
1661 For TCP network connections, tell fio to listen for incoming
1662 connections rather than initiating an outgoing connection. The
1663 hostname must be omitted if this option is used.
1665 .BI (net, pingpong) \fR=\fPbool
1666 Normally a network writer will just continue writing data, and a network reader
1667 will just consume packets. If pingpong=1 is set, a writer will send its normal
1668 payload to the reader, then wait for the reader to send the same payload back.
1669 This allows fio to measure network latencies. The submission and completion
1670 latencies then measure local time spent sending or receiving, and the
1671 completion latency measures how long it took for the other end to receive and
1672 send back. For UDP multicast traffic pingpong=1 should only be set for a single
1673 reader when multiple readers are listening to the same address.
1675 .BI (net, window_size) \fR=\fPint
1676 Set the desired socket buffer size for the connection.
1678 .BI (net, mss) \fR=\fPint
1679 Set the TCP maximum segment size (TCP_MAXSEG).
1681 .BI (e4defrag,donorname) \fR=\fPstr
1682 File will be used as a block donor (swap extents between files)
1684 .BI (e4defrag,inplace) \fR=\fPint
1685 Configure donor file block allocation strategy
1688 Preallocate donor's file on init
1691 allocate space immediately inside defragment event, and free right after event
1694 .BI (rbd)rbdname \fR=\fPstr
1695 Specifies the name of the RBD.
1697 .BI (rbd)pool \fR=\fPstr
1698 Specifies the name of the Ceph pool containing the RBD.
1700 .BI (rbd)clientname \fR=\fPstr
1701 Specifies the username (without the 'client.' prefix) used to access the Ceph cluster.
1703 .BI (mtd)skipbad \fR=\fPbool
1704 Skip operations against known bad blocks.
1706 While running, \fBfio\fR will display the status of the created jobs. For
1710 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
1713 The characters in the first set of brackets denote the current status of each
1714 threads. The possible values are:
1720 Setup but not started.
1726 Initialized, waiting.
1729 Running, doing sequential reads.
1732 Running, doing random reads.
1735 Running, doing sequential writes.
1738 Running, doing random writes.
1741 Running, doing mixed sequential reads/writes.
1744 Running, doing mixed random reads/writes.
1747 Running, currently waiting for \fBfsync\fR\|(2).
1750 Running, verifying written data.
1753 Exited, not reaped by main thread.
1756 Exited, thread reaped.
1760 The second set of brackets shows the estimated completion percentage of
1761 the current group. The third set shows the read and write I/O rate,
1762 respectively. Finally, the estimated run time of the job is displayed.
1764 When \fBfio\fR completes (or is interrupted by Ctrl-C), it will show data
1765 for each thread, each group of threads, and each disk, in that order.
1767 Per-thread statistics first show the threads client number, group-id, and
1768 error code. The remaining figures are as follows:
1772 Number of megabytes of I/O performed.
1775 Average data rate (bandwidth).
1781 Submission latency minimum, maximum, average and standard deviation. This is
1782 the time it took to submit the I/O.
1785 Completion latency minimum, maximum, average and standard deviation. This
1786 is the time between submission and completion.
1789 Bandwidth minimum, maximum, percentage of aggregate bandwidth received, average
1790 and standard deviation.
1793 CPU usage statistics. Includes user and system time, number of context switches
1794 this thread went through and number of major and minor page faults.
1797 Distribution of I/O depths. Each depth includes everything less than (or equal)
1798 to it, but greater than the previous depth.
1801 Number of read/write requests issued, and number of short read/write requests.
1804 Distribution of I/O completion latencies. The numbers follow the same pattern
1808 The group statistics show:
1813 Number of megabytes I/O performed.
1816 Aggregate bandwidth of threads in the group.
1819 Minimum average bandwidth a thread saw.
1822 Maximum average bandwidth a thread saw.
1825 Shortest runtime of threads in the group.
1828 Longest runtime of threads in the group.
1832 Finally, disk statistics are printed with reads first:
1837 Number of I/Os performed by all groups.
1840 Number of merges in the I/O scheduler.
1843 Number of ticks we kept the disk busy.
1846 Total time spent in the disk queue.
1853 It is also possible to get fio to dump the current output while it is
1854 running, without terminating the job. To do that, send fio the \fBUSR1\fR
1857 If the \fB\-\-minimal\fR / \fB\-\-append-terse\fR options are given, the
1858 results will be printed/appended in a semicolon-delimited format suitable for
1860 A job description (if provided) follows on a new line. Note that the first
1861 number in the line is the version number. If the output has to be changed
1862 for some reason, this number will be incremented by 1 to signify that
1863 change. The fields are:
1866 .B terse version, fio version, jobname, groupid, error
1870 .B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, IOPS, runtime \fR(ms)\fP
1874 .B min, max, mean, standard deviation
1878 .B min, max, mean, standard deviation
1880 Completion latency percentiles (20 fields):
1882 .B Xth percentile=usec
1886 .B min, max, mean, standard deviation
1890 .B min, max, aggregate percentage of total, mean, standard deviation
1896 .B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, IOPS, runtime \fR(ms)\fP
1900 .B min, max, mean, standard deviation
1904 .B min, max, mean, standard deviation
1906 Completion latency percentiles (20 fields):
1908 .B Xth percentile=usec
1912 .B min, max, mean, standard deviation
1916 .B min, max, aggregate percentage of total, mean, standard deviation
1922 .B user, system, context switches, major page faults, minor page faults
1925 IO depth distribution:
1927 .B <=1, 2, 4, 8, 16, 32, >=64
1930 IO latency distribution:
1934 .B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
1938 .B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
1942 Disk utilization (1 for each disk used):
1944 .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
1947 Error Info (dependent on continue_on_error, default off):
1949 .B total # errors, first error code
1952 .B text description (if provided in config - appears on newline)
1955 Normally you would run fio as a stand-alone application on the machine
1956 where the IO workload should be generated. However, it is also possible to
1957 run the frontend and backend of fio separately. This makes it possible to
1958 have a fio server running on the machine(s) where the IO workload should
1959 be running, while controlling it from another machine.
1961 To start the server, you would do:
1963 \fBfio \-\-server=args\fR
1965 on that machine, where args defines what fio listens to. The arguments
1966 are of the form 'type:hostname or IP:port'. 'type' is either 'ip' (or ip4)
1967 for TCP/IP v4, 'ip6' for TCP/IP v6, or 'sock' for a local unix domain
1968 socket. 'hostname' is either a hostname or IP address, and 'port' is the port to
1969 listen to (only valid for TCP/IP, not a local socket). Some examples:
1973 Start a fio server, listening on all interfaces on the default port (8765).
1975 2) fio \-\-server=ip:hostname,4444
1977 Start a fio server, listening on IP belonging to hostname and on port 4444.
1979 3) fio \-\-server=ip6:::1,4444
1981 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
1983 4) fio \-\-server=,4444
1985 Start a fio server, listening on all interfaces on port 4444.
1987 5) fio \-\-server=1.2.3.4
1989 Start a fio server, listening on IP 1.2.3.4 on the default port.
1991 6) fio \-\-server=sock:/tmp/fio.sock
1993 Start a fio server, listening on the local socket /tmp/fio.sock.
1995 When a server is running, you can connect to it from a client. The client
1998 fio \-\-local-args \-\-client=server \-\-remote-args <job file(s)>
2000 where \-\-local-args are arguments that are local to the client where it is
2001 running, 'server' is the connect string, and \-\-remote-args and <job file(s)>
2002 are sent to the server. The 'server' string follows the same format as it
2003 does on the server side, to allow IP/hostname/socket and port strings.
2004 You can connect to multiple clients as well, to do that you could run:
2006 fio \-\-client=server2 \-\-client=server2 <job file(s)>
2008 If the job file is located on the fio server, then you can tell the server
2009 to load a local file as well. This is done by using \-\-remote-config:
2011 fio \-\-client=server \-\-remote-config /path/to/file.fio
2013 Then fio will open this local (to the server) job file instead
2014 of being passed one from the client.
2016 If you have many servers (example: 100 VMs/containers), you can input a pathname
2017 of a file containing host IPs/names as the parameter value for the \-\-client option.
2018 For example, here is an example "host.list" file containing 2 hostnames:
2020 host1.your.dns.domain
2022 host2.your.dns.domain
2024 The fio command would then be:
2026 fio \-\-client=host.list <job file>
2028 In this mode, you cannot input server-specific parameters or job files, and all
2029 servers receive the same job file.
2031 In order to enable fio \-\-client runs utilizing a shared filesystem from multiple hosts,
2032 fio \-\-client now prepends the IP address of the server to the filename. For example,
2033 if fio is using directory /mnt/nfs/fio and is writing filename fileio.tmp,
2034 with a \-\-client hostfile
2035 containing two hostnames h1 and h2 with IP addresses 192.168.10.120 and 192.168.10.121, then
2036 fio will create two files:
2038 /mnt/nfs/fio/192.168.10.120.fileio.tmp
2040 /mnt/nfs/fio/192.168.10.121.fileio.tmp
2045 was written by Jens Axboe <jens.axboe@oracle.com>,
2046 now Jens Axboe <axboe@fb.com>.
2048 This man page was written by Aaron Carroll <aaronc@cse.unsw.edu.au> based
2049 on documentation by Jens Axboe.
2050 .SH "REPORTING BUGS"
2051 Report bugs to the \fBfio\fR mailing list <fio@vger.kernel.org>.
2054 For further documentation see \fBHOWTO\fR and \fBREADME\fR.
2056 Sample jobfiles are available in the \fBexamples\fR directory.
projects / fio.git / blob