1 .TH fio 1 "October 2013" "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 \-\-runtime \fR=\fPruntime
24 Limit run time to \fIruntime\fR seconds.
27 Generate per-job latency logs.
30 Generate per-job bandwidth logs.
33 Print statistics in a terse, semicolon-delimited format.
36 Display version information and exit.
38 .BI \-\-terse\-version \fR=\fPversion
39 Set terse version output format (Current version 3, or older version 2).
42 Display usage information and exit.
45 Perform test and validation of internal CPU clock
47 .BI \-\-crctest[\fR=\fPtest]
48 Test the speed of the builtin checksumming functions. If no argument is given,
49 all of them are tested. Or a comma separated list can be passed, in which
50 case the given ones are tested.
52 .BI \-\-cmdhelp \fR=\fPcommand
53 Print help information for \fIcommand\fR. May be `all' for all commands.
55 .BI \-\-enghelp \fR=\fPioengine[,command]
56 List all commands defined by \fIioengine\fR, or print help for \fIcommand\fR defined by \fIioengine\fR.
58 .BI \-\-showcmd \fR=\fPjobfile
59 Convert \fIjobfile\fR to a set of command-line options.
61 .BI \-\-eta \fR=\fPwhen
62 Specifies when real-time ETA estimate should be printed. \fIwhen\fR may
63 be one of `always', `never' or `auto'.
65 .BI \-\-eta\-newline \fR=\fPtime
66 Force an ETA newline for every `time` period passed.
68 .BI \-\-status\-interval \fR=\fPtime
69 Report full output status every `time` period passed.
72 Turn on safety read-only checks, preventing any attempted write.
74 .BI \-\-section \fR=\fPsec
75 Only run section \fIsec\fR from job file. Multiple of these options can be given, adding more sections to run.
77 .BI \-\-alloc\-size \fR=\fPkb
78 Set the internal smalloc pool size to \fIkb\fP kilobytes.
80 .BI \-\-warnings\-fatal
81 All fio parser warnings are fatal, causing fio to exit with an error.
83 .BI \-\-max\-jobs \fR=\fPnr
84 Set the maximum allowed number of jobs (threads/processes) to support.
86 .BI \-\-server \fR=\fPargs
87 Start a backend server, with \fIargs\fP specifying what to listen to. See client/server section.
89 .BI \-\-daemonize \fR=\fPpidfile
90 Background a fio server, writing the pid to the given pid file.
92 .BI \-\-client \fR=\fPhost
93 Instead of running the jobs locally, send and run them on the given host.
95 .BI \-\-idle\-prof \fR=\fPoption
96 Report cpu idleness on a system or percpu basis (\fIoption\fP=system,percpu) or run unit work calibration only (\fIoption\fP=calibrate).
98 Job files are in `ini' format. They consist of one or more
99 job definitions, which begin with a job name in square brackets and
100 extend to the next job name. The job name can be any ASCII string
101 except `global', which has a special meaning. Following the job name is
102 a sequence of zero or more parameters, one per line, that define the
103 behavior of the job. Any line starting with a `;' or `#' character is
104 considered a comment and ignored.
106 If \fIjobfile\fR is specified as `-', the job file will be read from
109 The global section contains default parameters for jobs specified in the
110 job file. A job is only affected by global sections residing above it,
111 and there may be any number of global sections. Specific job definitions
112 may override any parameter set in global sections.
115 Some parameters may take arguments of a specific type. The types used are:
118 String: a sequence of alphanumeric characters.
121 SI integer: a whole number, possibly containing a suffix denoting the base unit
122 of the value. Accepted suffixes are `k', 'M', 'G', 'T', and 'P', denoting
123 kilo (1024), mega (1024^2), giga (1024^3), tera (1024^4), and peta (1024^5)
124 respectively. If prefixed with '0x', the value is assumed to be base 16
125 (hexadecimal). A suffix may include a trailing 'b', for instance 'kb' is
126 identical to 'k'. You can specify a base 10 value by using 'KiB', 'MiB','GiB',
127 etc. This is useful for disk drives where values are often given in base 10
128 values. Specifying '30GiB' will get you 30*1000^3 bytes.
129 When specifying times the default suffix meaning changes, still denoting the
130 base unit of the value, but accepted suffixes are 'D' (days), 'H' (hours), 'M'
131 (minutes), 'S' Seconds, 'ms' milli seconds. Time values without a unit specify
133 The suffixes are not case sensitive.
136 Boolean: a true or false value. `0' denotes false, `1' denotes true.
139 Integer range: a range of integers specified in the format
140 \fIlower\fR:\fIupper\fR or \fIlower\fR\-\fIupper\fR. \fIlower\fR and
141 \fIupper\fR may contain a suffix as described above. If an option allows two
142 sets of ranges, they are separated with a `,' or `/' character. For example:
146 List of floating numbers: A list of floating numbers, separated by
151 May be used to override the job name. On the command line, this parameter
152 has the special purpose of signalling the start of a new job.
154 .BI description \fR=\fPstr
155 Human-readable description of the job. It is printed when the job is run, but
156 otherwise has no special purpose.
158 .BI directory \fR=\fPstr
159 Prefix filenames with this directory. Used to place files in a location other
161 You can specify a number of directories by separating the names with a ':'
162 character. These directories will be assigned equally distributed to job clones
163 creates with \fInumjobs\fR as long as they are using generated filenames.
164 If specific \fIfilename(s)\fR are set fio will use the first listed directory,
165 and thereby matching the \fIfilename\fR semantic which generates a file each
166 clone if not specified, but let all clones use the same if set.
168 .BI filename \fR=\fPstr
170 normally makes up a file name based on the job name, thread number, and file
171 number. If you want to share files between threads in a job or several jobs,
172 specify a \fIfilename\fR for each of them to override the default.
173 If the I/O engine is file-based, you can specify
174 a number of files by separating the names with a `:' character. `\-' is a
175 reserved name, meaning stdin or stdout, depending on the read/write direction
178 .BI filename_format \fR=\fPstr
179 If sharing multiple files between jobs, it is usually necessary to have
180 fio generate the exact names that you want. By default, fio will name a file
181 based on the default file format specification of
182 \fBjobname.jobnumber.filenumber\fP. With this option, that can be
183 customized. Fio will recognize and replace the following keywords in this
189 The name of the worker thread or process.
192 The incremental number of the worker thread or process.
195 The incremental number of the file for that worker thread or process.
198 To have dependent jobs share a set of files, this option can be set to
199 have fio generate filenames that are shared between the two. For instance,
200 if \fBtestfiles.$filenum\fR is specified, file number 4 for any job will
201 be named \fBtestfiles.4\fR. The default of \fB$jobname.$jobnum.$filenum\fR
202 will be used if no other format specifier is given.
206 .BI lockfile \fR=\fPstr
207 Fio defaults to not locking any files before it does IO to them. If a file or
208 file descriptor is shared, fio can serialize IO to that file to make the end
209 result consistent. This is usual for emulating real workloads that share files.
215 No locking. This is the default.
218 Only one thread or process may do IO at the time, excluding all others.
221 Read-write locking on the file. Many readers may access the file at the same
222 time, but writes get exclusive access.
226 .BI opendir \fR=\fPstr
227 Recursively open any files below directory \fIstr\fR.
229 .BI readwrite \fR=\fPstr "\fR,\fP rw" \fR=\fPstr
230 Type of I/O pattern. Accepted values are:
241 Sequential trim (Linux block devices only).
250 Random trim (Linux block devices only).
253 Mixed sequential reads and writes.
256 Mixed random reads and writes.
259 For mixed I/O, the default split is 50/50. For certain types of io the result
260 may still be skewed a bit, since the speed may be different. It is possible to
261 specify a number of IO's to do before getting a new offset, this is done by
262 appending a `:\fI<nr>\fR to the end of the string given. For a random read, it
263 would look like \fBrw=randread:8\fR for passing in an offset modifier with a
264 value of 8. If the postfix is used with a sequential IO pattern, then the value
265 specified will be added to the generated offset for each IO. For instance,
266 using \fBrw=write:4k\fR will skip 4k for every write. It turns sequential IO
267 into sequential IO with holes. See the \fBrw_sequencer\fR option.
270 .BI rw_sequencer \fR=\fPstr
271 If an offset modifier is given by appending a number to the \fBrw=<str>\fR line,
272 then this option controls how that number modifies the IO offset being
273 generated. Accepted values are:
278 Generate sequential offset
281 Generate the same offset
284 \fBsequential\fR is only useful for random IO, where fio would normally
285 generate a new random offset for every IO. If you append eg 8 to randread, you
286 would get a new random offset for every 8 IO's. The result would be a seek for
287 only every 8 IO's, instead of for every IO. Use \fBrw=randread:8\fR to specify
288 that. As sequential IO is already sequential, setting \fBsequential\fR for that
289 would not result in any differences. \fBidentical\fR behaves in a similar
290 fashion, except it sends the same offset 8 number of times before generating a
295 .BI kb_base \fR=\fPint
296 The base unit for a kilobyte. The defacto base is 2^10, 1024. Storage
297 manufacturers like to use 10^3 or 1000 as a base ten unit instead, for obvious
298 reasons. Allowed values are 1024 or 1000, with 1024 being the default.
300 .BI unified_rw_reporting \fR=\fPbool
301 Fio normally reports statistics on a per data direction basis, meaning that
302 read, write, and trim are accounted and reported separately. If this option is
303 set, the fio will sum the results and report them as "mixed" instead.
305 .BI randrepeat \fR=\fPbool
306 Seed the random number generator used for random I/O patterns in a predictable
307 way so the pattern is repeatable across runs. Default: true.
309 .BI allrandrepeat \fR=\fPbool
310 Seed all random number generators in a predictable way so results are
311 repeatable across runs. Default: false.
313 .BI randseed \fR=\fPint
314 Seed the random number generators based on this seed value, to be able to
315 control what sequence of output is being generated. If not set, the random
316 sequence depends on the \fBrandrepeat\fR setting.
318 .BI use_os_rand \fR=\fPbool
319 Fio can either use the random generator supplied by the OS to generator random
320 offsets, or it can use it's own internal generator (based on Tausworthe).
321 Default is to use the internal generator, which is often of better quality and
322 faster. Default: false.
324 .BI fallocate \fR=\fPstr
325 Whether pre-allocation is performed when laying down files. Accepted values
331 Do not pre-allocate space.
334 Pre-allocate via \fBposix_fallocate\fR\|(3).
337 Pre-allocate via \fBfallocate\fR\|(2) with FALLOC_FL_KEEP_SIZE set.
340 Backward-compatible alias for 'none'.
343 Backward-compatible alias for 'posix'.
346 May not be available on all supported platforms. 'keep' is only
347 available on Linux. If using ZFS on Solaris this must be set to 'none'
348 because ZFS doesn't support it. Default: 'posix'.
351 .BI fadvise_hint \fR=\fPbool
352 Use of \fBposix_fadvise\fR\|(2) to advise the kernel what I/O patterns
353 are likely to be issued. Default: true.
356 Total size of I/O for this job. \fBfio\fR will run until this many bytes have
357 been transferred, unless limited by other options (\fBruntime\fR, for instance).
358 Unless \fBnrfiles\fR and \fBfilesize\fR options are given, this amount will be
359 divided between the available files for the job. If not set, fio will use the
360 full size of the given files or devices. If the files do not exist, size
361 must be given. It is also possible to give size as a percentage between 1 and
362 100. If size=20% is given, fio will use 20% of the full size of the given files
365 .BI fill_device \fR=\fPbool "\fR,\fB fill_fs" \fR=\fPbool
366 Sets size to something really large and waits for ENOSPC (no space left on
367 device) as the terminating condition. Only makes sense with sequential write.
368 For a read workload, the mount point will be filled first then IO started on
369 the result. This option doesn't make sense if operating on a raw device node,
370 since the size of that is already known by the file system. Additionally,
371 writing beyond end-of-device will not return ENOSPC there.
373 .BI filesize \fR=\fPirange
374 Individual file sizes. May be a range, in which case \fBfio\fR will select sizes
375 for files at random within the given range, limited to \fBsize\fR in total (if
376 that is given). If \fBfilesize\fR is not specified, each created file is the
379 .BI blocksize \fR=\fPint[,int] "\fR,\fB bs" \fR=\fPint[,int]
380 Block size for I/O units. Default: 4k. Values for reads, writes, and trims
381 can be specified separately in the format \fIread\fR,\fIwrite\fR,\fItrim\fR
382 either of which may be empty to leave that value at its default. If a trailing
383 comma isn't given, the remainder will inherit the last value set.
385 .BI blocksize_range \fR=\fPirange[,irange] "\fR,\fB bsrange" \fR=\fPirange[,irange]
386 Specify a range of I/O block sizes. The issued I/O unit will always be a
387 multiple of the minimum size, unless \fBblocksize_unaligned\fR is set. Applies
388 to both reads and writes if only one range is given, but can be specified
389 separately with a comma separating the values. Example: bsrange=1k-4k,2k-8k.
390 Also (see \fBblocksize\fR).
392 .BI bssplit \fR=\fPstr
393 This option allows even finer grained control of the block sizes issued,
394 not just even splits between them. With this option, you can weight various
395 block sizes for exact control of the issued IO for a job that has mixed
396 block sizes. The format of the option is bssplit=blocksize/percentage,
397 optionally adding as many definitions as needed separated by a colon.
398 Example: bssplit=4k/10:64k/50:32k/40 would issue 50% 64k blocks, 10% 4k
399 blocks and 40% 32k blocks. \fBbssplit\fR also supports giving separate
400 splits to reads and writes. The format is identical to what the
401 \fBbs\fR option accepts, the read and write parts are separated with a
404 .B blocksize_unaligned\fR,\fP bs_unaligned
405 If set, any size in \fBblocksize_range\fR may be used. This typically won't
406 work with direct I/O, as that normally requires sector alignment.
408 .BI blockalign \fR=\fPint[,int] "\fR,\fB ba" \fR=\fPint[,int]
409 At what boundary to align random IO offsets. Defaults to the same as 'blocksize'
410 the minimum blocksize given. Minimum alignment is typically 512b
411 for using direct IO, though it usually depends on the hardware block size.
412 This option is mutually exclusive with using a random map for files, so it
413 will turn off that option.
415 .BI bs_is_seq_rand \fR=\fPbool
416 If this option is set, fio will use the normal read,write blocksize settings as
417 sequential,random instead. Any random read or write will use the WRITE
418 blocksize settings, and any sequential read or write will use the READ
422 Initialise buffers with all zeros. Default: fill buffers with random data.
425 If this option is given, fio will refill the IO buffers on every submit. The
426 default is to only fill it at init time and reuse that data. Only makes sense
427 if zero_buffers isn't specified, naturally. If data verification is enabled,
428 refill_buffers is also automatically enabled.
430 .BI scramble_buffers \fR=\fPbool
431 If \fBrefill_buffers\fR is too costly and the target is using data
432 deduplication, then setting this option will slightly modify the IO buffer
433 contents to defeat normal de-dupe attempts. This is not enough to defeat
434 more clever block compression attempts, but it will stop naive dedupe
435 of blocks. Default: true.
437 .BI buffer_compress_percentage \fR=\fPint
438 If this is set, then fio will attempt to provide IO buffer content (on WRITEs)
439 that compress to the specified level. Fio does this by providing a mix of
440 random data and zeroes. Note that this is per block size unit, for file/disk
441 wide compression level that matches this setting, you'll also want to set
442 \fBrefill_buffers\fR.
444 .BI buffer_compress_chunk \fR=\fPint
445 See \fBbuffer_compress_percentage\fR. This setting allows fio to manage how
446 big the ranges of random data and zeroed data is. Without this set, fio will
447 provide \fBbuffer_compress_percentage\fR of blocksize random data, followed by
448 the remaining zeroed. With this set to some chunk size smaller than the block
449 size, fio can alternate random and zeroed data throughout the IO buffer.
451 .BI buffer_pattern \fR=\fPstr
452 If set, fio will fill the io buffers with this pattern. If not set, the contents
453 of io buffers is defined by the other options related to buffer contents. The
454 setting can be any pattern of bytes, and can be prefixed with 0x for hex
457 .BI nrfiles \fR=\fPint
458 Number of files to use for this job. Default: 1.
460 .BI openfiles \fR=\fPint
461 Number of files to keep open at the same time. Default: \fBnrfiles\fR.
463 .BI file_service_type \fR=\fPstr
464 Defines how files to service are selected. The following types are defined:
469 Choose a file at random.
472 Round robin over open files (default).
475 Do each file in the set sequentially.
478 The number of I/Os to issue before switching a new file can be specified by
479 appending `:\fIint\fR' to the service type.
482 .BI ioengine \fR=\fPstr
483 Defines how the job issues I/O. The following types are defined:
488 Basic \fBread\fR\|(2) or \fBwrite\fR\|(2) I/O. \fBfseek\fR\|(2) is used to
489 position the I/O location.
492 Basic \fBpread\fR\|(2) or \fBpwrite\fR\|(2) I/O.
495 Basic \fBreadv\fR\|(2) or \fBwritev\fR\|(2) I/O. Will emulate queuing by
496 coalescing adjacent IOs into a single submission.
499 Basic \fBpreadv\fR\|(2) or \fBpwritev\fR\|(2) I/O.
502 Linux native asynchronous I/O. This ioengine defines engine specific options.
505 POSIX asynchronous I/O using \fBaio_read\fR\|(3) and \fBaio_write\fR\|(3).
508 Solaris native asynchronous I/O.
511 Windows native asynchronous I/O.
514 File is memory mapped with \fBmmap\fR\|(2) and data copied using
518 \fBsplice\fR\|(2) is used to transfer the data and \fBvmsplice\fR\|(2) to
519 transfer data from user-space to the kernel.
522 Use the syslet system calls to make regular read/write asynchronous.
525 SCSI generic sg v3 I/O. May be either synchronous using the SG_IO ioctl, or if
526 the target is an sg character device, we use \fBread\fR\|(2) and
527 \fBwrite\fR\|(2) for asynchronous I/O.
530 Doesn't transfer any data, just pretends to. Mainly used to exercise \fBfio\fR
531 itself and for debugging and testing purposes.
534 Transfer over the network. The protocol to be used can be defined with the
535 \fBprotocol\fR parameter. Depending on the protocol, \fBfilename\fR,
536 \fBhostname\fR, \fBport\fR, or \fBlisten\fR must be specified.
537 This ioengine defines engine specific options.
540 Like \fBnet\fR, but uses \fBsplice\fR\|(2) and \fBvmsplice\fR\|(2) to map data
541 and send/receive. This ioengine defines engine specific options.
544 Doesn't transfer any data, but burns CPU cycles according to \fBcpuload\fR and
545 \fBcpucycles\fR parameters.
548 The GUASI I/O engine is the Generic Userspace Asynchronous Syscall Interface
549 approach to asynchronous I/O.
551 See <http://www.xmailserver.org/guasi\-lib.html>.
554 The RDMA I/O engine supports both RDMA memory semantics (RDMA_WRITE/RDMA_READ)
555 and channel semantics (Send/Recv) for the InfiniBand, RoCE and iWARP protocols.
558 Loads an external I/O engine object file. Append the engine filename as
562 IO engine that does regular linux native fallocate call to simulate data
563 transfer as fio ioengine
565 DDIR_READ does fallocate(,mode = FALLOC_FL_KEEP_SIZE,)
567 DIR_WRITE does fallocate(,mode = 0)
569 DDIR_TRIM does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE)
572 IO engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate defragment activity
573 request to DDIR_WRITE event
576 IO engine supporting direct access to Ceph Rados Block Devices (RBD) via librbd
577 without the need to use the kernel rbd driver. This ioengine defines engine specific
583 .BI iodepth \fR=\fPint
584 Number of I/O units to keep in flight against the file. Note that increasing
585 iodepth beyond 1 will not affect synchronous ioengines (except for small
586 degress when verify_async is in use). Even async engines my impose OS
587 restrictions causing the desired depth not to be achieved. This may happen on
588 Linux when using libaio and not setting \fBdirect\fR=1, since buffered IO is
589 not async on that OS. Keep an eye on the IO depth distribution in the
590 fio output to verify that the achieved depth is as expected. Default: 1.
592 .BI iodepth_batch \fR=\fPint
593 Number of I/Os to submit at once. Default: \fBiodepth\fR.
595 .BI iodepth_batch_complete \fR=\fPint
596 This defines how many pieces of IO to retrieve at once. It defaults to 1 which
597 means that we'll ask for a minimum of 1 IO in the retrieval process from the
598 kernel. The IO retrieval will go on until we hit the limit set by
599 \fBiodepth_low\fR. If this variable is set to 0, then fio will always check for
600 completed events before queuing more IO. This helps reduce IO latency, at the
601 cost of more retrieval system calls.
603 .BI iodepth_low \fR=\fPint
604 Low watermark indicating when to start filling the queue again. Default:
607 .BI direct \fR=\fPbool
608 If true, use non-buffered I/O (usually O_DIRECT). Default: false.
610 .BI atomic \fR=\fPbool
611 If value is true, attempt to use atomic direct IO. Atomic writes are guaranteed
612 to be stable once acknowledged by the operating system. Only Linux supports
615 .BI buffered \fR=\fPbool
616 If true, use buffered I/O. This is the opposite of the \fBdirect\fR parameter.
619 .BI offset \fR=\fPint
620 Offset in the file to start I/O. Data before the offset will not be touched.
622 .BI offset_increment \fR=\fPint
623 If this is provided, then the real offset becomes the
624 offset + offset_increment * thread_number, where the thread number is a counter
625 that starts at 0 and is incremented for each job. This option is useful if
626 there are several jobs which are intended to operate on a file in parallel in
627 disjoint segments, with even spacing between the starting points.
629 .BI number_ios \fR=\fPint
630 Fio will normally perform IOs until it has exhausted the size of the region
631 set by \fBsize\fR, or if it exhaust the allocated time (or hits an error
632 condition). With this setting, the range/size can be set independently of
633 the number of IOs to perform. When fio reaches this number, it will exit
634 normally and report status.
637 How many I/Os to perform before issuing an \fBfsync\fR\|(2) of dirty data. If
638 0, don't sync. Default: 0.
640 .BI fdatasync \fR=\fPint
641 Like \fBfsync\fR, but uses \fBfdatasync\fR\|(2) instead to only sync the
642 data parts of the file. Default: 0.
644 .BI write_barrier \fR=\fPint
645 Make every Nth write a barrier write.
647 .BI sync_file_range \fR=\fPstr:int
648 Use \fBsync_file_range\fR\|(2) for every \fRval\fP number of write operations. Fio will
649 track range of writes that have happened since the last \fBsync_file_range\fR\|(2) call.
650 \fRstr\fP can currently be one or more of:
654 SYNC_FILE_RANGE_WAIT_BEFORE
657 SYNC_FILE_RANGE_WRITE
660 SYNC_FILE_RANGE_WRITE
664 So if you do sync_file_range=wait_before,write:8, fio would use
665 \fBSYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE\fP for every 8 writes.
666 Also see the \fBsync_file_range\fR\|(2) man page. This option is Linux specific.
668 .BI overwrite \fR=\fPbool
669 If writing, setup the file first and do overwrites. Default: false.
671 .BI end_fsync \fR=\fPbool
672 Sync file contents when a write stage has completed. Default: false.
674 .BI fsync_on_close \fR=\fPbool
675 If true, sync file contents on close. This differs from \fBend_fsync\fR in that
676 it will happen on every close, not just at the end of the job. Default: false.
678 .BI rwmixread \fR=\fPint
679 Percentage of a mixed workload that should be reads. Default: 50.
681 .BI rwmixwrite \fR=\fPint
682 Percentage of a mixed workload that should be writes. If \fBrwmixread\fR and
683 \fBrwmixwrite\fR are given and do not sum to 100%, the latter of the two
684 overrides the first. This may interfere with a given rate setting, if fio is
685 asked to limit reads or writes to a certain rate. If that is the case, then
686 the distribution may be skewed. Default: 50.
688 .BI random_distribution \fR=\fPstr:float
689 By default, fio will use a completely uniform random distribution when asked
690 to perform random IO. Sometimes it is useful to skew the distribution in
691 specific ways, ensuring that some parts of the data is more hot than others.
692 Fio includes the following distribution models:
696 Uniform random distribution
706 When using a zipf or pareto distribution, an input value is also needed to
707 define the access pattern. For zipf, this is the zipf theta. For pareto,
708 it's the pareto power. Fio includes a test program, genzipf, that can be
709 used visualize what the given input values will yield in terms of hit rates.
710 If you wanted to use zipf with a theta of 1.2, you would use
711 random_distribution=zipf:1.2 as the option. If a non-uniform model is used,
712 fio will disable use of the random map.
714 .BI percentage_random \fR=\fPint
715 For a random workload, set how big a percentage should be random. This defaults
716 to 100%, in which case the workload is fully random. It can be set from
717 anywhere from 0 to 100. Setting it to 0 would make the workload fully
718 sequential. It is possible to set different values for reads, writes, and
719 trim. To do so, simply use a comma separated list. See \fBblocksize\fR.
722 Normally \fBfio\fR will cover every block of the file when doing random I/O. If
723 this parameter is given, a new offset will be chosen without looking at past
724 I/O history. This parameter is mutually exclusive with \fBverify\fR.
726 .BI softrandommap \fR=\fPbool
727 See \fBnorandommap\fR. If fio runs with the random block map enabled and it
728 fails to allocate the map, if this option is set it will continue without a
729 random block map. As coverage will not be as complete as with random maps, this
730 option is disabled by default.
732 .BI random_generator \fR=\fPstr
733 Fio supports the following engines for generating IO offsets for random IO:
737 Strong 2^88 cycle random number generator
740 Linear feedback shift register generator
744 Tausworthe is a strong random number generator, but it requires tracking on the
745 side if we want to ensure that blocks are only read or written once. LFSR
746 guarantees that we never generate the same offset twice, and it's also less
747 computationally expensive. It's not a true random generator, however, though
748 for IO purposes it's typically good enough. LFSR only works with single block
749 sizes, not with workloads that use multiple block sizes. If used with such a
750 workload, fio may read or write some blocks multiple times.
753 Run job with given nice value. See \fBnice\fR\|(2).
756 Set I/O priority value of this job between 0 (highest) and 7 (lowest). See
759 .BI prioclass \fR=\fPint
760 Set I/O priority class. See \fBionice\fR\|(1).
762 .BI thinktime \fR=\fPint
763 Stall job for given number of microseconds between issuing I/Os.
765 .BI thinktime_spin \fR=\fPint
766 Pretend to spend CPU time for given number of microseconds, sleeping the rest
767 of the time specified by \fBthinktime\fR. Only valid if \fBthinktime\fR is set.
769 .BI thinktime_blocks \fR=\fPint
770 Only valid if thinktime is set - control how many blocks to issue, before
771 waiting \fBthinktime\fR microseconds. If not set, defaults to 1 which will
772 make fio wait \fBthinktime\fR microseconds after every block. This
773 effectively makes any queue depth setting redundant, since no more than 1 IO
774 will be queued before we have to complete it and do our thinktime. In other
775 words, this setting effectively caps the queue depth if the latter is larger.
779 Cap bandwidth used by this job. The number is in bytes/sec, the normal postfix
780 rules apply. You can use \fBrate\fR=500k to limit reads and writes to 500k each,
781 or you can specify read and writes separately. Using \fBrate\fR=1m,500k would
782 limit reads to 1MB/sec and writes to 500KB/sec. Capping only reads or writes
783 can be done with \fBrate\fR=,500k or \fBrate\fR=500k,. The former will only
784 limit writes (to 500KB/sec), the latter will only limit reads.
786 .BI ratemin \fR=\fPint
787 Tell \fBfio\fR to do whatever it can to maintain at least the given bandwidth.
788 Failing to meet this requirement will cause the job to exit. The same format
789 as \fBrate\fR is used for read vs write separation.
791 .BI rate_iops \fR=\fPint
792 Cap the bandwidth to this number of IOPS. Basically the same as rate, just
793 specified independently of bandwidth. The same format as \fBrate\fR is used for
794 read vs write separation. If \fBblocksize\fR is a range, the smallest block
795 size is used as the metric.
797 .BI rate_iops_min \fR=\fPint
798 If this rate of I/O is not met, the job will exit. The same format as \fBrate\fR
799 is used for read vs write separation.
801 .BI ratecycle \fR=\fPint
802 Average bandwidth for \fBrate\fR and \fBratemin\fR over this number of
803 milliseconds. Default: 1000ms.
805 .BI latency_target \fR=\fPint
806 If set, fio will attempt to find the max performance point that the given
807 workload will run at while maintaining a latency below this target. The
808 values is given in microseconds. See \fBlatency_window\fR and
809 \fBlatency_percentile\fR.
811 .BI latency_window \fR=\fPint
812 Used with \fBlatency_target\fR to specify the sample window that the job
813 is run at varying queue depths to test the performance. The value is given
816 .BI latency_percentile \fR=\fPfloat
817 The percentage of IOs that must fall within the criteria specified by
818 \fBlatency_target\fR and \fBlatency_window\fR. If not set, this defaults
819 to 100.0, meaning that all IOs must be equal or below to the value set
820 by \fBlatency_target\fR.
822 .BI max_latency \fR=\fPint
823 If set, fio will exit the job if it exceeds this maximum latency. It will exit
826 .BI cpumask \fR=\fPint
827 Set CPU affinity for this job. \fIint\fR is a bitmask of allowed CPUs the job
828 may run on. See \fBsched_setaffinity\fR\|(2).
830 .BI cpus_allowed \fR=\fPstr
831 Same as \fBcpumask\fR, but allows a comma-delimited list of CPU numbers.
833 .BI numa_cpu_nodes \fR=\fPstr
834 Set this job running on specified NUMA nodes' CPUs. The arguments allow
835 comma delimited list of cpu numbers, A-B ranges, or 'all'.
837 .BI numa_mem_policy \fR=\fPstr
838 Set this job's memory policy and corresponding NUMA nodes. Format of
842 .B <mode>[:<nodelist>]
845 is one of the following memory policy:
847 .B default, prefer, bind, interleave, local
850 For \fBdefault\fR and \fBlocal\fR memory policy, no \fBnodelist\fR is
851 needed to be specified. For \fBprefer\fR, only one node is
852 allowed. For \fBbind\fR and \fBinterleave\fR, \fBnodelist\fR allows
853 comma delimited list of numbers, A-B ranges, or 'all'.
855 .BI startdelay \fR=\fPirange
856 Delay start of job for the specified number of seconds. Supports all time
857 suffixes to allow specification of hours, minutes, seconds and
858 milliseconds - seconds are the default if a unit is ommited.
859 Can be given as a range which causes each thread to choose randomly out of the
862 .BI runtime \fR=\fPint
863 Terminate processing after the specified number of seconds.
866 If given, run for the specified \fBruntime\fR duration even if the files are
867 completely read or written. The same workload will be repeated as many times
868 as \fBruntime\fR allows.
870 .BI ramp_time \fR=\fPint
871 If set, fio will run the specified workload for this amount of time before
872 logging any performance numbers. Useful for letting performance settle before
873 logging results, thus minimizing the runtime required for stable results. Note
874 that the \fBramp_time\fR is considered lead in time for a job, thus it will
875 increase the total runtime if a special timeout or runtime is specified.
877 .BI invalidate \fR=\fPbool
878 Invalidate buffer-cache for the file prior to starting I/O. Default: true.
881 Use synchronous I/O for buffered writes. For the majority of I/O engines,
882 this means using O_SYNC. Default: false.
884 .BI iomem \fR=\fPstr "\fR,\fP mem" \fR=\fPstr
885 Allocation method for I/O unit buffer. Allowed values are:
890 Allocate memory with \fBmalloc\fR\|(3).
893 Use shared memory buffers allocated through \fBshmget\fR\|(2).
896 Same as \fBshm\fR, but use huge pages as backing.
899 Use \fBmmap\fR\|(2) for allocation. Uses anonymous memory unless a filename
900 is given after the option in the format `:\fIfile\fR'.
903 Same as \fBmmap\fR, but use huge files as backing.
906 The amount of memory allocated is the maximum allowed \fBblocksize\fR for the
907 job multiplied by \fBiodepth\fR. For \fBshmhuge\fR or \fBmmaphuge\fR to work,
908 the system must have free huge pages allocated. \fBmmaphuge\fR also needs to
909 have hugetlbfs mounted, and \fIfile\fR must point there. At least on Linux,
910 huge pages must be manually allocated. See \fB/proc/sys/vm/nr_hugehages\fR
911 and the documentation for that. Normally you just need to echo an appropriate
912 number, eg echoing 8 will ensure that the OS has 8 huge pages ready for
916 .BI iomem_align \fR=\fPint "\fR,\fP mem_align" \fR=\fPint
917 This indicates the memory alignment of the IO memory buffers. Note that the
918 given alignment is applied to the first IO unit buffer, if using \fBiodepth\fR
919 the alignment of the following buffers are given by the \fBbs\fR used. In
920 other words, if using a \fBbs\fR that is a multiple of the page sized in the
921 system, all buffers will be aligned to this value. If using a \fBbs\fR that
922 is not page aligned, the alignment of subsequent IO memory buffers is the
923 sum of the \fBiomem_align\fR and \fBbs\fR used.
925 .BI hugepage\-size \fR=\fPint
926 Defines the size of a huge page. Must be at least equal to the system setting.
927 Should be a multiple of 1MB. Default: 4MB.
930 Terminate all jobs when one finishes. Default: wait for each job to finish.
932 .BI bwavgtime \fR=\fPint
933 Average bandwidth calculations over the given time in milliseconds. Default:
936 .BI iopsavgtime \fR=\fPint
937 Average IOPS calculations over the given time in milliseconds. Default:
940 .BI create_serialize \fR=\fPbool
941 If true, serialize file creation for the jobs. Default: true.
943 .BI create_fsync \fR=\fPbool
944 \fBfsync\fR\|(2) data file after creation. Default: true.
946 .BI create_on_open \fR=\fPbool
947 If true, the files are not created until they are opened for IO by the job.
949 .BI create_only \fR=\fPbool
950 If true, fio will only run the setup phase of the job. If files need to be
951 laid out or updated on disk, only that will be done. The actual job contents
954 .BI pre_read \fR=\fPbool
955 If this is given, files will be pre-read into memory before starting the given
956 IO operation. This will also clear the \fR \fBinvalidate\fR flag, since it is
957 pointless to pre-read and then drop the cache. This will only work for IO
958 engines that are seekable, since they allow you to read the same data
959 multiple times. Thus it will not work on eg network or splice IO.
961 .BI unlink \fR=\fPbool
962 Unlink job files when done. Default: false.
965 Specifies the number of iterations (runs of the same workload) of this job.
968 .BI verify_only \fR=\fPbool
969 Do not perform the specified workload, only verify data still matches previous
970 invocation of this workload. This option allows one to check data multiple
971 times at a later date without overwriting it. This option makes sense only for
972 workloads that write data, and does not support workloads with the
973 \fBtime_based\fR option set.
975 .BI do_verify \fR=\fPbool
976 Run the verify phase after a write phase. Only valid if \fBverify\fR is set.
979 .BI verify \fR=\fPstr
980 Method of verifying file contents after each iteration of the job. Allowed
985 .B md5 crc16 crc32 crc32c crc32c-intel crc64 crc7 sha256 sha512 sha1
986 Store appropriate checksum in the header of each block. crc32c-intel is
987 hardware accelerated SSE4.2 driven, falls back to regular crc32c if
988 not supported by the system.
991 Write extra information about each I/O (timestamp, block number, etc.). The
992 block number is verified. See \fBverify_pattern\fR as well.
995 Pretend to verify. Used for testing internals.
998 This option can be used for repeated burn-in tests of a system to make sure
999 that the written data is also correctly read back. If the data direction given
1000 is a read or random read, fio will assume that it should verify a previously
1001 written file. If the data direction includes any form of write, the verify will
1002 be of the newly written data.
1005 .BI verifysort \fR=\fPbool
1006 If true, written verify blocks are sorted if \fBfio\fR deems it to be faster to
1007 read them back in a sorted manner. Default: true.
1009 .BI verifysort_nr \fR=\fPint
1010 Pre-load and sort verify blocks for a read workload.
1012 .BI verify_offset \fR=\fPint
1013 Swap the verification header with data somewhere else in the block before
1014 writing. It is swapped back before verifying.
1016 .BI verify_interval \fR=\fPint
1017 Write the verification header for this number of bytes, which should divide
1018 \fBblocksize\fR. Default: \fBblocksize\fR.
1020 .BI verify_pattern \fR=\fPstr
1021 If set, fio will fill the io buffers with this pattern. Fio defaults to filling
1022 with totally random bytes, but sometimes it's interesting to fill with a known
1023 pattern for io verification purposes. Depending on the width of the pattern,
1024 fio will fill 1/2/3/4 bytes of the buffer at the time(it can be either a
1025 decimal or a hex number). The verify_pattern if larger than a 32-bit quantity
1026 has to be a hex number that starts with either "0x" or "0X". Use with
1029 .BI verify_fatal \fR=\fPbool
1030 If true, exit the job on the first observed verification failure. Default:
1033 .BI verify_dump \fR=\fPbool
1034 If set, dump the contents of both the original data block and the data block we
1035 read off disk to files. This allows later analysis to inspect just what kind of
1036 data corruption occurred. Off by default.
1038 .BI verify_async \fR=\fPint
1039 Fio will normally verify IO inline from the submitting thread. This option
1040 takes an integer describing how many async offload threads to create for IO
1041 verification instead, causing fio to offload the duty of verifying IO contents
1042 to one or more separate threads. If using this offload option, even sync IO
1043 engines can benefit from using an \fBiodepth\fR setting higher than 1, as it
1044 allows them to have IO in flight while verifies are running.
1046 .BI verify_async_cpus \fR=\fPstr
1047 Tell fio to set the given CPU affinity on the async IO verification threads.
1048 See \fBcpus_allowed\fP for the format used.
1050 .BI verify_backlog \fR=\fPint
1051 Fio will normally verify the written contents of a job that utilizes verify
1052 once that job has completed. In other words, everything is written then
1053 everything is read back and verified. You may want to verify continually
1054 instead for a variety of reasons. Fio stores the meta data associated with an
1055 IO block in memory, so for large verify workloads, quite a bit of memory would
1056 be used up holding this meta data. If this option is enabled, fio will write
1057 only N blocks before verifying these blocks.
1059 .BI verify_backlog_batch \fR=\fPint
1060 Control how many blocks fio will verify if verify_backlog is set. If not set,
1061 will default to the value of \fBverify_backlog\fR (meaning the entire queue is
1062 read back and verified). If \fBverify_backlog_batch\fR is less than
1063 \fBverify_backlog\fR then not all blocks will be verified, if
1064 \fBverify_backlog_batch\fR is larger than \fBverify_backlog\fR, some blocks
1065 will be verified more than once.
1067 .BI trim_percentage \fR=\fPint
1068 Number of verify blocks to discard/trim.
1070 .BI trim_verify_zero \fR=\fPbool
1071 Verify that trim/discarded blocks are returned as zeroes.
1073 .BI trim_backlog \fR=\fPint
1074 Trim after this number of blocks are written.
1076 .BI trim_backlog_batch \fR=\fPint
1077 Trim this number of IO blocks.
1079 .BI experimental_verify \fR=\fPbool
1080 Enable experimental verification.
1082 .B stonewall "\fR,\fP wait_for_previous"
1083 Wait for preceding jobs in the job file to exit before starting this one.
1084 \fBstonewall\fR implies \fBnew_group\fR.
1087 Start a new reporting group. If not given, all jobs in a file will be part
1088 of the same reporting group, unless separated by a stonewall.
1090 .BI numjobs \fR=\fPint
1091 Number of clones (processes/threads performing the same workload) of this job.
1095 If set, display per-group reports instead of per-job when \fBnumjobs\fR is
1099 Use threads created with \fBpthread_create\fR\|(3) instead of processes created
1100 with \fBfork\fR\|(2).
1102 .BI zonesize \fR=\fPint
1103 Divide file into zones of the specified size in bytes. See \fBzoneskip\fR.
1105 .BI zonerange \fR=\fPint
1106 Give size of an IO zone. See \fBzoneskip\fR.
1108 .BI zoneskip \fR=\fPint
1109 Skip the specified number of bytes when \fBzonesize\fR bytes of data have been
1112 .BI write_iolog \fR=\fPstr
1113 Write the issued I/O patterns to the specified file. Specify a separate file
1114 for each job, otherwise the iologs will be interspersed and the file may be
1117 .BI read_iolog \fR=\fPstr
1118 Replay the I/O patterns contained in the specified file generated by
1119 \fBwrite_iolog\fR, or may be a \fBblktrace\fR binary file.
1121 .BI replay_no_stall \fR=\fPint
1122 While replaying I/O patterns using \fBread_iolog\fR the default behavior
1123 attempts to respect timing information between I/Os. Enabling
1124 \fBreplay_no_stall\fR causes I/Os to be replayed as fast as possible while
1125 still respecting ordering.
1127 .BI replay_redirect \fR=\fPstr
1128 While replaying I/O patterns using \fBread_iolog\fR the default behavior
1129 is to replay the IOPS onto the major/minor device that each IOP was recorded
1130 from. Setting \fBreplay_redirect\fR causes all IOPS to be replayed onto the
1131 single specified device regardless of the device it was recorded from.
1133 .BI write_bw_log \fR=\fPstr
1134 If given, write a bandwidth log of the jobs in this job file. Can be used to
1135 store data of the bandwidth of the jobs in their lifetime. The included
1136 fio_generate_plots script uses gnuplot to turn these text files into nice
1137 graphs. See \fBwrite_lat_log\fR for behaviour of given filename. For this
1138 option, the postfix is _bw.log.
1140 .BI write_lat_log \fR=\fPstr
1141 Same as \fBwrite_bw_log\fR, but writes I/O completion latencies. If no
1142 filename is given with this option, the default filename of "jobname_type.log"
1143 is used. Even if the filename is given, fio will still append the type of log.
1145 .BI write_iops_log \fR=\fPstr
1146 Same as \fBwrite_bw_log\fR, but writes IOPS. If no filename is given with this
1147 option, the default filename of "jobname_type.log" is used. Even if the
1148 filename is given, fio will still append the type of log.
1150 .BI log_avg_msec \fR=\fPint
1151 By default, fio will log an entry in the iops, latency, or bw log for every
1152 IO that completes. When writing to the disk log, that can quickly grow to a
1153 very large size. Setting this option makes fio average the each log entry
1154 over the specified period of time, reducing the resolution of the log.
1157 .BI disable_lat \fR=\fPbool
1158 Disable measurements of total latency numbers. Useful only for cutting
1159 back the number of calls to \fBgettimeofday\fR\|(2), as that does impact performance at
1160 really high IOPS rates. Note that to really get rid of a large amount of these
1161 calls, this option must be used with disable_slat and disable_bw as well.
1163 .BI disable_clat \fR=\fPbool
1164 Disable measurements of completion latency numbers. See \fBdisable_lat\fR.
1166 .BI disable_slat \fR=\fPbool
1167 Disable measurements of submission latency numbers. See \fBdisable_lat\fR.
1169 .BI disable_bw_measurement \fR=\fPbool
1170 Disable measurements of throughput/bandwidth numbers. See \fBdisable_lat\fR.
1172 .BI lockmem \fR=\fPint
1173 Pin the specified amount of memory with \fBmlock\fR\|(2). Can be used to
1174 simulate a smaller amount of memory. The amount specified is per worker.
1176 .BI exec_prerun \fR=\fPstr
1177 Before running the job, execute the specified command with \fBsystem\fR\|(3).
1179 Output is redirected in a file called \fBjobname.prerun.txt\fR
1182 .BI exec_postrun \fR=\fPstr
1183 Same as \fBexec_prerun\fR, but the command is executed after the job completes.
1185 Output is redirected in a file called \fBjobname.postrun.txt\fR
1188 .BI ioscheduler \fR=\fPstr
1189 Attempt to switch the device hosting the file to the specified I/O scheduler.
1191 .BI cpuload \fR=\fPint
1192 If the job is a CPU cycle-eater, attempt to use the specified percentage of
1195 .BI cpuchunks \fR=\fPint
1196 If the job is a CPU cycle-eater, split the load into cycles of the
1197 given time in milliseconds.
1199 .BI disk_util \fR=\fPbool
1200 Generate disk utilization statistics if the platform supports it. Default: true.
1202 .BI clocksource \fR=\fPstr
1203 Use the given clocksource as the base of timing. The supported options are:
1207 \fBgettimeofday\fR\|(2)
1210 \fBclock_gettime\fR\|(2)
1213 Internal CPU clock source
1217 \fBcpu\fR is the preferred clocksource if it is reliable, as it is very fast
1218 (and fio is heavy on time calls). Fio will automatically use this clocksource
1219 if it's supported and considered reliable on the system it is running on,
1220 unless another clocksource is specifically set. For x86/x86-64 CPUs, this
1221 means supporting TSC Invariant.
1223 .BI gtod_reduce \fR=\fPbool
1224 Enable all of the \fBgettimeofday\fR\|(2) reducing options (disable_clat, disable_slat,
1225 disable_bw) plus reduce precision of the timeout somewhat to really shrink the
1226 \fBgettimeofday\fR\|(2) call count. With this option enabled, we only do about 0.4% of
1227 the gtod() calls we would have done if all time keeping was enabled.
1229 .BI gtod_cpu \fR=\fPint
1230 Sometimes it's cheaper to dedicate a single thread of execution to just getting
1231 the current time. Fio (and databases, for instance) are very intensive on
1232 \fBgettimeofday\fR\|(2) calls. With this option, you can set one CPU aside for doing
1233 nothing but logging current time to a shared memory location. Then the other
1234 threads/processes that run IO workloads need only copy that segment, instead of
1235 entering the kernel with a \fBgettimeofday\fR\|(2) call. The CPU set aside for doing
1236 these time calls will be excluded from other uses. Fio will manually clear it
1237 from the CPU mask of other jobs.
1239 .BI ignore_error \fR=\fPstr
1240 Sometimes you want to ignore some errors during test in that case you can specify
1241 error list for each error type.
1243 ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST
1245 errors for given error type is separated with ':'.
1246 Error may be symbol ('ENOSPC', 'ENOMEM') or an integer.
1248 Example: ignore_error=EAGAIN,ENOSPC:122 .
1250 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from WRITE.
1252 .BI error_dump \fR=\fPbool
1253 If set dump every error even if it is non fatal, true by default. If disabled
1254 only fatal error will be dumped
1256 .BI profile \fR=\fPstr
1257 Select a specific builtin performance test.
1259 .BI cgroup \fR=\fPstr
1260 Add job to this control group. If it doesn't exist, it will be created.
1261 The system must have a mounted cgroup blkio mount point for this to work. If
1262 your system doesn't have it mounted, you can do so with:
1264 # mount \-t cgroup \-o blkio none /cgroup
1266 .BI cgroup_weight \fR=\fPint
1267 Set the weight of the cgroup to this value. See the documentation that comes
1268 with the kernel, allowed values are in the range of 100..1000.
1270 .BI cgroup_nodelete \fR=\fPbool
1271 Normally fio will delete the cgroups it has created after the job completion.
1272 To override this behavior and to leave cgroups around after the job completion,
1273 set cgroup_nodelete=1. This can be useful if one wants to inspect various
1274 cgroup files after job completion. Default: false
1277 Instead of running as the invoking user, set the user ID to this value before
1278 the thread/process does any work.
1281 Set group ID, see \fBuid\fR.
1283 .BI unit_base \fR=\fPint
1284 Base unit for reporting. Allowed values are:
1288 Use auto-detection (default).
1298 .BI flow_id \fR=\fPint
1299 The ID of the flow. If not specified, it defaults to being a global flow. See
1303 Weight in token-based flow control. If this value is used, then there is a
1304 \fBflow counter\fR which is used to regulate the proportion of activity between
1305 two or more jobs. fio attempts to keep this flow counter near zero. The
1306 \fBflow\fR parameter stands for how much should be added or subtracted to the
1307 flow counter on each iteration of the main I/O loop. That is, if one job has
1308 \fBflow=8\fR and another job has \fBflow=-1\fR, then there will be a roughly
1309 1:8 ratio in how much one runs vs the other.
1311 .BI flow_watermark \fR=\fPint
1312 The maximum value that the absolute value of the flow counter is allowed to
1313 reach before the job must wait for a lower value of the counter.
1315 .BI flow_sleep \fR=\fPint
1316 The period of time, in microseconds, to wait after the flow watermark has been
1317 exceeded before retrying operations
1319 .BI clat_percentiles \fR=\fPbool
1320 Enable the reporting of percentiles of completion latencies.
1322 .BI percentile_list \fR=\fPfloat_list
1323 Overwrite the default list of percentiles for completion
1324 latencies. Each number is a floating number in the range (0,100], and
1325 the maximum length of the list is 20. Use ':' to separate the
1326 numbers. For example, \-\-percentile_list=99.5:99.9 will cause fio to
1327 report the values of completion latency below which 99.5% and 99.9% of
1328 the observed latencies fell, respectively.
1329 .SS "Ioengine Parameters List"
1330 Some parameters are only valid when a specific ioengine is in use. These are
1331 used identically to normal parameters, with the caveat that when used on the
1332 command line, the must come after the ioengine that defines them is selected.
1334 .BI (cpu)cpuload \fR=\fPint
1335 Attempt to use the specified percentage of CPU cycles.
1337 .BI (cpu)cpuchunks \fR=\fPint
1338 Split the load into cycles of the given time. In microseconds.
1340 .BI (libaio)userspace_reap
1341 Normally, with the libaio engine in use, fio will use
1342 the io_getevents system call to reap newly returned events.
1343 With this flag turned on, the AIO ring will be read directly
1344 from user-space to reap events. The reaping mode is only
1345 enabled when polling for a minimum of 0 events (eg when
1346 iodepth_batch_complete=0).
1348 .BI (net,netsplice)hostname \fR=\fPstr
1349 The host name or IP address to use for TCP or UDP based IO.
1350 If the job is a TCP listener or UDP reader, the hostname is not
1351 used and must be omitted unless it is a valid UDP multicast address.
1353 .BI (net,netsplice)port \fR=\fPint
1354 The TCP or UDP port to bind to or connect to.
1356 .BI (net,netsplice)interface \fR=\fPstr
1357 The IP address of the network interface used to send or receive UDP multicast
1360 .BI (net,netsplice)ttl \fR=\fPint
1361 Time-to-live value for outgoing UDP multicast packets. Default: 1
1363 .BI (net,netsplice)nodelay \fR=\fPbool
1364 Set TCP_NODELAY on TCP connections.
1366 .BI (net,netsplice)protocol \fR=\fPstr "\fR,\fP proto" \fR=\fPstr
1367 The network protocol to use. Accepted values are:
1372 Transmission control protocol
1375 Transmission control protocol V6
1378 User datagram protocol
1381 User datagram protocol V6
1387 When the protocol is TCP or UDP, the port must also be given,
1388 as well as the hostname if the job is a TCP listener or UDP
1389 reader. For unix sockets, the normal filename option should be
1390 used and the port is invalid.
1393 .BI (net,netsplice)listen
1394 For TCP network connections, tell fio to listen for incoming
1395 connections rather than initiating an outgoing connection. The
1396 hostname must be omitted if this option is used.
1398 .BI (net, pingpong) \fR=\fPbool
1399 Normally a network writer will just continue writing data, and a network reader
1400 will just consume packages. If pingpong=1 is set, a writer will send its normal
1401 payload to the reader, then wait for the reader to send the same payload back.
1402 This allows fio to measure network latencies. The submission and completion
1403 latencies then measure local time spent sending or receiving, and the
1404 completion latency measures how long it took for the other end to receive and
1405 send back. For UDP multicast traffic pingpong=1 should only be set for a single
1406 reader when multiple readers are listening to the same address.
1408 .BI (e4defrag,donorname) \fR=\fPstr
1409 File will be used as a block donor (swap extents between files)
1411 .BI (e4defrag,inplace) \fR=\fPint
1412 Configure donor file block allocation strategy
1415 Preallocate donor's file on init
1418 allocate space immediately inside defragment event, and free right after event
1421 .BI (rbd)rbdname \fR=\fPstr
1422 Specifies the name of the RBD.
1424 .BI (rbd)pool \fR=\fPstr
1425 Specifies the name of the Ceph pool containing the RBD.
1427 .BI (rbd)clientname \fR=\fPstr
1428 Specifies the username (without the 'client.' prefix) used to access the Ceph cluster.
1430 While running, \fBfio\fR will display the status of the created jobs. For
1434 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
1437 The characters in the first set of brackets denote the current status of each
1438 threads. The possible values are:
1444 Setup but not started.
1450 Initialized, waiting.
1453 Running, doing sequential reads.
1456 Running, doing random reads.
1459 Running, doing sequential writes.
1462 Running, doing random writes.
1465 Running, doing mixed sequential reads/writes.
1468 Running, doing mixed random reads/writes.
1471 Running, currently waiting for \fBfsync\fR\|(2).
1474 Running, verifying written data.
1477 Exited, not reaped by main thread.
1480 Exited, thread reaped.
1484 The second set of brackets shows the estimated completion percentage of
1485 the current group. The third set shows the read and write I/O rate,
1486 respectively. Finally, the estimated run time of the job is displayed.
1488 When \fBfio\fR completes (or is interrupted by Ctrl-C), it will show data
1489 for each thread, each group of threads, and each disk, in that order.
1491 Per-thread statistics first show the threads client number, group-id, and
1492 error code. The remaining figures are as follows:
1496 Number of megabytes of I/O performed.
1499 Average data rate (bandwidth).
1505 Submission latency minimum, maximum, average and standard deviation. This is
1506 the time it took to submit the I/O.
1509 Completion latency minimum, maximum, average and standard deviation. This
1510 is the time between submission and completion.
1513 Bandwidth minimum, maximum, percentage of aggregate bandwidth received, average
1514 and standard deviation.
1517 CPU usage statistics. Includes user and system time, number of context switches
1518 this thread went through and number of major and minor page faults.
1521 Distribution of I/O depths. Each depth includes everything less than (or equal)
1522 to it, but greater than the previous depth.
1525 Number of read/write requests issued, and number of short read/write requests.
1528 Distribution of I/O completion latencies. The numbers follow the same pattern
1532 The group statistics show:
1537 Number of megabytes I/O performed.
1540 Aggregate bandwidth of threads in the group.
1543 Minimum average bandwidth a thread saw.
1546 Maximum average bandwidth a thread saw.
1549 Shortest runtime of threads in the group.
1552 Longest runtime of threads in the group.
1556 Finally, disk statistics are printed with reads first:
1561 Number of I/Os performed by all groups.
1564 Number of merges in the I/O scheduler.
1567 Number of ticks we kept the disk busy.
1570 Total time spent in the disk queue.
1577 It is also possible to get fio to dump the current output while it is
1578 running, without terminating the job. To do that, send fio the \fBUSR1\fR
1581 If the \fB\-\-minimal\fR option is given, the results will be printed in a
1582 semicolon-delimited format suitable for scripted use - a job description
1583 (if provided) follows on a new line. Note that the first
1584 number in the line is the version number. If the output has to be changed
1585 for some reason, this number will be incremented by 1 to signify that
1586 change. The fields are:
1589 .B terse version, fio version, jobname, groupid, error
1593 .B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, IOPS, runtime \fR(ms)\fP
1597 .B min, max, mean, standard deviation
1601 .B min, max, mean, standard deviation
1603 Completion latency percentiles (20 fields):
1605 .B Xth percentile=usec
1609 .B min, max, mean, standard deviation
1613 .B min, max, aggregate percentage of total, mean, standard deviation
1619 .B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, IOPS, runtime \fR(ms)\fP
1623 .B min, max, mean, standard deviation
1627 .B min, max, mean, standard deviation
1629 Completion latency percentiles (20 fields):
1631 .B Xth percentile=usec
1635 .B min, max, mean, standard deviation
1639 .B min, max, aggregate percentage of total, mean, standard deviation
1645 .B user, system, context switches, major page faults, minor page faults
1648 IO depth distribution:
1650 .B <=1, 2, 4, 8, 16, 32, >=64
1653 IO latency distribution:
1657 .B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
1661 .B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
1665 Disk utilization (1 for each disk used):
1667 .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
1670 Error Info (dependent on continue_on_error, default off):
1672 .B total # errors, first error code
1675 .B text description (if provided in config - appears on newline)
1678 Normally you would run fio as a stand-alone application on the machine
1679 where the IO workload should be generated. However, it is also possible to
1680 run the frontend and backend of fio separately. This makes it possible to
1681 have a fio server running on the machine(s) where the IO workload should
1682 be running, while controlling it from another machine.
1684 To start the server, you would do:
1686 \fBfio \-\-server=args\fR
1688 on that machine, where args defines what fio listens to. The arguments
1689 are of the form 'type:hostname or IP:port'. 'type' is either 'ip' (or ip4)
1690 for TCP/IP v4, 'ip6' for TCP/IP v6, or 'sock' for a local unix domain
1691 socket. 'hostname' is either a hostname or IP address, and 'port' is the port to
1692 listen to (only valid for TCP/IP, not a local socket). Some examples:
1696 Start a fio server, listening on all interfaces on the default port (8765).
1698 2) fio \-\-server=ip:hostname,4444
1700 Start a fio server, listening on IP belonging to hostname and on port 4444.
1702 3) fio \-\-server=ip6:::1,4444
1704 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
1706 4) fio \-\-server=,4444
1708 Start a fio server, listening on all interfaces on port 4444.
1710 5) fio \-\-server=1.2.3.4
1712 Start a fio server, listening on IP 1.2.3.4 on the default port.
1714 6) fio \-\-server=sock:/tmp/fio.sock
1716 Start a fio server, listening on the local socket /tmp/fio.sock.
1718 When a server is running, you can connect to it from a client. The client
1721 fio \-\-local-args \-\-client=server \-\-remote-args <job file(s)>
1723 where \-\-local-args are arguments that are local to the client where it is
1724 running, 'server' is the connect string, and \-\-remote-args and <job file(s)>
1725 are sent to the server. The 'server' string follows the same format as it
1726 does on the server side, to allow IP/hostname/socket and port strings.
1727 You can connect to multiple clients as well, to do that you could run:
1729 fio \-\-client=server2 \-\-client=server2 <job file(s)>
1733 was written by Jens Axboe <jens.axboe@oracle.com>,
1734 now Jens Axboe <jaxboe@fusionio.com>.
1736 This man page was written by Aaron Carroll <aaronc@cse.unsw.edu.au> based
1737 on documentation by Jens Axboe.
1738 .SH "REPORTING BUGS"
1739 Report bugs to the \fBfio\fR mailing list <fio@vger.kernel.org>.
1742 For further documentation see \fBHOWTO\fR and \fBREADME\fR.
1744 Sample jobfiles are available in the \fBexamples\fR directory.