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 Print statistics in selected mode AND terse, semicolon-delimited format.
39 Display version information and exit.
41 .BI \-\-terse\-version \fR=\fPversion
42 Set terse version output format (Current version 3, or older version 2).
45 Display usage information and exit.
48 Perform test and validation of internal CPU clock
50 .BI \-\-crctest[\fR=\fPtest]
51 Test the speed of the builtin checksumming functions. If no argument is given,
52 all of them are tested. Or a comma separated list can be passed, in which
53 case the given ones are tested.
55 .BI \-\-cmdhelp \fR=\fPcommand
56 Print help information for \fIcommand\fR. May be `all' for all commands.
58 .BI \-\-enghelp \fR=\fPioengine[,command]
59 List all commands defined by \fIioengine\fR, or print help for \fIcommand\fR defined by \fIioengine\fR.
61 .BI \-\-showcmd \fR=\fPjobfile
62 Convert \fIjobfile\fR to a set of command-line options.
64 .BI \-\-eta \fR=\fPwhen
65 Specifies when real-time ETA estimate should be printed. \fIwhen\fR may
66 be one of `always', `never' or `auto'.
68 .BI \-\-eta\-newline \fR=\fPtime
69 Force an ETA newline for every `time` period passed.
71 .BI \-\-status\-interval \fR=\fPtime
72 Report full output status every `time` period passed.
75 Turn on safety read-only checks, preventing any attempted write.
77 .BI \-\-section \fR=\fPsec
78 Only run section \fIsec\fR from job file. This option can be used multiple times to add more sections to run.
80 .BI \-\-alloc\-size \fR=\fPkb
81 Set the internal smalloc pool size to \fIkb\fP kilobytes.
83 .BI \-\-warnings\-fatal
84 All fio parser warnings are fatal, causing fio to exit with an error.
86 .BI \-\-max\-jobs \fR=\fPnr
87 Set the maximum allowed number of jobs (threads/processes) to support.
89 .BI \-\-server \fR=\fPargs
90 Start a backend server, with \fIargs\fP specifying what to listen to. See client/server section.
92 .BI \-\-daemonize \fR=\fPpidfile
93 Background a fio server, writing the pid to the given pid file.
95 .BI \-\-client \fR=\fPhost
96 Instead of running the jobs locally, send and run them on the given host.
98 .BI \-\-idle\-prof \fR=\fPoption
99 Report cpu idleness on a system or percpu basis (\fIoption\fP=system,percpu) or run unit work calibration only (\fIoption\fP=calibrate).
100 .SH "JOB FILE FORMAT"
101 Job files are in `ini' format. They consist of one or more
102 job definitions, which begin with a job name in square brackets and
103 extend to the next job name. The job name can be any ASCII string
104 except `global', which has a special meaning. Following the job name is
105 a sequence of zero or more parameters, one per line, that define the
106 behavior of the job. Any line starting with a `;' or `#' character is
107 considered a comment and ignored.
109 If \fIjobfile\fR is specified as `-', the job file will be read from
112 The global section contains default parameters for jobs specified in the
113 job file. A job is only affected by global sections residing above it,
114 and there may be any number of global sections. Specific job definitions
115 may override any parameter set in global sections.
118 Some parameters may take arguments of a specific type. The types used are:
121 String: a sequence of alphanumeric characters.
124 SI integer: a whole number, possibly containing a suffix denoting the base unit
125 of the value. Accepted suffixes are `k', 'M', 'G', 'T', and 'P', denoting
126 kilo (1024), mega (1024^2), giga (1024^3), tera (1024^4), and peta (1024^5)
127 respectively. If prefixed with '0x', the value is assumed to be base 16
128 (hexadecimal). A suffix may include a trailing 'b', for instance 'kb' is
129 identical to 'k'. You can specify a base 10 value by using 'KiB', 'MiB','GiB',
130 etc. This is useful for disk drives where values are often given in base 10
131 values. Specifying '30GiB' will get you 30*1000^3 bytes.
132 When specifying times the default suffix meaning changes, still denoting the
133 base unit of the value, but accepted suffixes are 'D' (days), 'H' (hours), 'M'
134 (minutes), 'S' Seconds, 'ms' (or msec) milli seconds, 'us' (or 'usec') micro
135 seconds. Time values without a unit specify seconds.
136 The suffixes are not case sensitive.
139 Boolean: a true or false value. `0' denotes false, `1' denotes true.
142 Integer range: a range of integers specified in the format
143 \fIlower\fR:\fIupper\fR or \fIlower\fR\-\fIupper\fR. \fIlower\fR and
144 \fIupper\fR may contain a suffix as described above. If an option allows two
145 sets of ranges, they are separated with a `,' or `/' character. For example:
149 List of floating numbers: A list of floating numbers, separated by
154 May be used to override the job name. On the command line, this parameter
155 has the special purpose of signalling the start of a new job.
157 .BI description \fR=\fPstr
158 Human-readable description of the job. It is printed when the job is run, but
159 otherwise has no special purpose.
161 .BI directory \fR=\fPstr
162 Prefix filenames with this directory. Used to place files in a location other
164 You can specify a number of directories by separating the names with a ':'
165 character. These directories will be assigned equally distributed to job clones
166 creates with \fInumjobs\fR as long as they are using generated filenames.
167 If specific \fIfilename(s)\fR are set fio will use the first listed directory,
168 and thereby matching the \fIfilename\fR semantic which generates a file each
169 clone if not specified, but let all clones use the same if set. See
170 \fIfilename\fR for considerations regarding escaping certain characters on
173 .BI filename \fR=\fPstr
175 normally makes up a file name based on the job name, thread number, and file
176 number. If you want to share files between threads in a job or several jobs,
177 specify a \fIfilename\fR for each of them to override the default.
178 If the I/O engine is file-based, you can specify
179 a number of files by separating the names with a `:' character. `\-' is a
180 reserved name, meaning stdin or stdout, depending on the read/write direction
181 set. On Windows, disk devices are accessed as \\.\PhysicalDrive0 for the first
182 device, \\.\PhysicalDrive1 for the second etc. Note: Windows and FreeBSD
183 prevent write access to areas of the disk containing in-use data
184 (e.g. filesystems). If the wanted filename does need to include a colon, then
185 escape that with a '\\' character. For instance, if the filename is
186 "/dev/dsk/foo@3,0:c", then you would use filename="/dev/dsk/foo@3,0\\:c".
188 .BI filename_format \fR=\fPstr
189 If sharing multiple files between jobs, it is usually necessary to have
190 fio generate the exact names that you want. By default, fio will name a file
191 based on the default file format specification of
192 \fBjobname.jobnumber.filenumber\fP. With this option, that can be
193 customized. Fio will recognize and replace the following keywords in this
199 The name of the worker thread or process.
202 The incremental number of the worker thread or process.
205 The incremental number of the file for that worker thread or process.
208 To have dependent jobs share a set of files, this option can be set to
209 have fio generate filenames that are shared between the two. For instance,
210 if \fBtestfiles.$filenum\fR is specified, file number 4 for any job will
211 be named \fBtestfiles.4\fR. The default of \fB$jobname.$jobnum.$filenum\fR
212 will be used if no other format specifier is given.
216 .BI lockfile \fR=\fPstr
217 Fio defaults to not locking any files before it does IO to them. If a file or
218 file descriptor is shared, fio can serialize IO to that file to make the end
219 result consistent. This is usual for emulating real workloads that share files.
225 No locking. This is the default.
228 Only one thread or process may do IO at a time, excluding all others.
231 Read-write locking on the file. Many readers may access the file at the same
232 time, but writes get exclusive access.
236 .BI opendir \fR=\fPstr
237 Recursively open any files below directory \fIstr\fR.
239 .BI readwrite \fR=\fPstr "\fR,\fP rw" \fR=\fPstr
240 Type of I/O pattern. Accepted values are:
251 Sequential trim (Linux block devices only).
260 Random trim (Linux block devices only).
263 Mixed sequential reads and writes.
266 Mixed random reads and writes.
269 For mixed I/O, the default split is 50/50. For certain types of io the result
270 may still be skewed a bit, since the speed may be different. It is possible to
271 specify a number of IO's to do before getting a new offset, this is done by
272 appending a `:\fI<nr>\fR to the end of the string given. For a random read, it
273 would look like \fBrw=randread:8\fR for passing in an offset modifier with a
274 value of 8. If the postfix is used with a sequential IO pattern, then the value
275 specified will be added to the generated offset for each IO. For instance,
276 using \fBrw=write:4k\fR will skip 4k for every write. It turns sequential IO
277 into sequential IO with holes. See the \fBrw_sequencer\fR option.
280 .BI rw_sequencer \fR=\fPstr
281 If an offset modifier is given by appending a number to the \fBrw=<str>\fR line,
282 then this option controls how that number modifies the IO offset being
283 generated. Accepted values are:
288 Generate sequential offset
291 Generate the same offset
294 \fBsequential\fR is only useful for random IO, where fio would normally
295 generate a new random offset for every IO. If you append eg 8 to randread, you
296 would get a new random offset for every 8 IO's. The result would be a seek for
297 only every 8 IO's, instead of for every IO. Use \fBrw=randread:8\fR to specify
298 that. As sequential IO is already sequential, setting \fBsequential\fR for that
299 would not result in any differences. \fBidentical\fR behaves in a similar
300 fashion, except it sends the same offset 8 number of times before generating a
305 .BI kb_base \fR=\fPint
306 The base unit for a kilobyte. The defacto base is 2^10, 1024. Storage
307 manufacturers like to use 10^3 or 1000 as a base ten unit instead, for obvious
308 reasons. Allowed values are 1024 or 1000, with 1024 being the default.
310 .BI unified_rw_reporting \fR=\fPbool
311 Fio normally reports statistics on a per data direction basis, meaning that
312 read, write, and trim are accounted and reported separately. If this option is
313 set fio sums the results and reports them as "mixed" instead.
315 .BI randrepeat \fR=\fPbool
316 Seed the random number generator used for random I/O patterns in a predictable
317 way so the pattern is repeatable across runs. Default: true.
319 .BI allrandrepeat \fR=\fPbool
320 Seed all random number generators in a predictable way so results are
321 repeatable across runs. Default: false.
323 .BI randseed \fR=\fPint
324 Seed the random number generators based on this seed value, to be able to
325 control what sequence of output is being generated. If not set, the random
326 sequence depends on the \fBrandrepeat\fR setting.
328 .BI use_os_rand \fR=\fPbool
329 Fio can either use the random generator supplied by the OS to generate random
330 offsets, or it can use its own internal generator (based on Tausworthe).
331 Default is to use the internal generator, which is often of better quality and
332 faster. Default: false.
334 .BI fallocate \fR=\fPstr
335 Whether pre-allocation is performed when laying down files. Accepted values
341 Do not pre-allocate space.
344 Pre-allocate via \fBposix_fallocate\fR\|(3).
347 Pre-allocate via \fBfallocate\fR\|(2) with FALLOC_FL_KEEP_SIZE set.
350 Backward-compatible alias for 'none'.
353 Backward-compatible alias for 'posix'.
356 May not be available on all supported platforms. 'keep' is only
357 available on Linux. If using ZFS on Solaris this must be set to 'none'
358 because ZFS doesn't support it. Default: 'posix'.
361 .BI fadvise_hint \fR=\fPbool
362 Use \fBposix_fadvise\fR\|(2) to advise the kernel what I/O patterns
363 are likely to be issued. Default: true.
366 Total size of I/O for this job. \fBfio\fR will run until this many bytes have
367 been transferred, unless limited by other options (\fBruntime\fR, for instance).
368 Unless \fBnrfiles\fR and \fBfilesize\fR options are given, this amount will be
369 divided between the available files for the job. If not set, fio will use the
370 full size of the given files or devices. If the files do not exist, size
371 must be given. It is also possible to give size as a percentage between 1 and
372 100. If size=20% is given, fio will use 20% of the full size of the given
375 .BI io_limit \fR=\fPint
376 Normally fio operates within the region set by \fBsize\fR, which means that
377 the \fBsize\fR option sets both the region and size of IO to be performed.
378 Sometimes that is not what you want. With this option, it is possible to
379 define just the amount of IO that fio should do. For instance, if \fBsize\fR
380 is set to 20G and \fBio_limit\fR is set to 5G, fio will perform IO within
381 the first 20G but exit when 5G have been done.
383 .BI fill_device \fR=\fPbool "\fR,\fB fill_fs" \fR=\fPbool
384 Sets size to something really large and waits for ENOSPC (no space left on
385 device) as the terminating condition. Only makes sense with sequential write.
386 For a read workload, the mount point will be filled first then IO started on
387 the result. This option doesn't make sense if operating on a raw device node,
388 since the size of that is already known by the file system. Additionally,
389 writing beyond end-of-device will not return ENOSPC there.
391 .BI filesize \fR=\fPirange
392 Individual file sizes. May be a range, in which case \fBfio\fR will select sizes
393 for files at random within the given range, limited to \fBsize\fR in total (if
394 that is given). If \fBfilesize\fR is not specified, each created file is the
397 .BI file_append \fR=\fPbool
398 Perform IO after the end of the file. Normally fio will operate within the
399 size of a file. If this option is set, then fio will append to the file
400 instead. This has identical behavior to setting \fRoffset\fP to the size
401 of a file. This option is ignored on non-regular files.
403 .BI blocksize \fR=\fPint[,int] "\fR,\fB bs" \fR=\fPint[,int]
404 Block size for I/O units. Default: 4k. Values for reads, writes, and trims
405 can be specified separately in the format \fIread\fR,\fIwrite\fR,\fItrim\fR
406 either of which may be empty to leave that value at its default. If a trailing
407 comma isn't given, the remainder will inherit the last value set.
409 .BI blocksize_range \fR=\fPirange[,irange] "\fR,\fB bsrange" \fR=\fPirange[,irange]
410 Specify a range of I/O block sizes. The issued I/O unit will always be a
411 multiple of the minimum size, unless \fBblocksize_unaligned\fR is set. Applies
412 to both reads and writes if only one range is given, but can be specified
413 separately with a comma separating the values. Example: bsrange=1k-4k,2k-8k.
414 Also (see \fBblocksize\fR).
416 .BI bssplit \fR=\fPstr
417 This option allows even finer grained control of the block sizes issued,
418 not just even splits between them. With this option, you can weight various
419 block sizes for exact control of the issued IO for a job that has mixed
420 block sizes. The format of the option is bssplit=blocksize/percentage,
421 optionally adding as many definitions as needed separated by a colon.
422 Example: bssplit=4k/10:64k/50:32k/40 would issue 50% 64k blocks, 10% 4k
423 blocks and 40% 32k blocks. \fBbssplit\fR also supports giving separate
424 splits to reads and writes. The format is identical to what the
425 \fBbs\fR option accepts, the read and write parts are separated with a
428 .B blocksize_unaligned\fR,\fP bs_unaligned
429 If set, any size in \fBblocksize_range\fR may be used. This typically won't
430 work with direct I/O, as that normally requires sector alignment.
432 .BI blockalign \fR=\fPint[,int] "\fR,\fB ba" \fR=\fPint[,int]
433 At what boundary to align random IO offsets. Defaults to the same as 'blocksize'
434 the minimum blocksize given. Minimum alignment is typically 512b
435 for using direct IO, though it usually depends on the hardware block size.
436 This option is mutually exclusive with using a random map for files, so it
437 will turn off that option.
439 .BI bs_is_seq_rand \fR=\fPbool
440 If this option is set, fio will use the normal read,write blocksize settings as
441 sequential,random instead. Any random read or write will use the WRITE
442 blocksize settings, and any sequential read or write will use the READ
446 Initialize buffers with all zeros. Default: fill buffers with random data.
447 The resulting IO buffers will not be completely zeroed, unless
448 \fPscramble_buffers\fR is also turned off.
451 If this option is given, fio will refill the IO buffers on every submit. The
452 default is to only fill it at init time and reuse that data. Only makes sense
453 if zero_buffers isn't specified, naturally. If data verification is enabled,
454 refill_buffers is also automatically enabled.
456 .BI scramble_buffers \fR=\fPbool
457 If \fBrefill_buffers\fR is too costly and the target is using data
458 deduplication, then setting this option will slightly modify the IO buffer
459 contents to defeat normal de-dupe attempts. This is not enough to defeat
460 more clever block compression attempts, but it will stop naive dedupe
461 of blocks. Default: true.
463 .BI buffer_compress_percentage \fR=\fPint
464 If this is set, then fio will attempt to provide IO buffer content (on WRITEs)
465 that compress to the specified level. Fio does this by providing a mix of
466 random data and zeroes. Note that this is per block size unit, for file/disk
467 wide compression level that matches this setting, you'll also want to set
468 \fBrefill_buffers\fR.
470 .BI buffer_compress_chunk \fR=\fPint
471 See \fBbuffer_compress_percentage\fR. This setting allows fio to manage how
472 big the ranges of random data and zeroed data is. Without this set, fio will
473 provide \fBbuffer_compress_percentage\fR of blocksize random data, followed by
474 the remaining zeroed. With this set to some chunk size smaller than the block
475 size, fio can alternate random and zeroed data throughout the IO buffer.
477 .BI buffer_pattern \fR=\fPstr
478 If set, fio will fill the IO buffers with this pattern. If not set, the contents
479 of IO buffers is defined by the other options related to buffer contents. The
480 setting can be any pattern of bytes, and can be prefixed with 0x for hex
483 .BI nrfiles \fR=\fPint
484 Number of files to use for this job. Default: 1.
486 .BI openfiles \fR=\fPint
487 Number of files to keep open at the same time. Default: \fBnrfiles\fR.
489 .BI file_service_type \fR=\fPstr
490 Defines how files to service are selected. The following types are defined:
495 Choose a file at random.
498 Round robin over opened files (default).
501 Do each file in the set sequentially.
504 The number of I/Os to issue before switching to a new file can be specified by
505 appending `:\fIint\fR' to the service type.
508 .BI ioengine \fR=\fPstr
509 Defines how the job issues I/O. The following types are defined:
514 Basic \fBread\fR\|(2) or \fBwrite\fR\|(2) I/O. \fBfseek\fR\|(2) is used to
515 position the I/O location.
518 Basic \fBpread\fR\|(2) or \fBpwrite\fR\|(2) I/O.
521 Basic \fBreadv\fR\|(2) or \fBwritev\fR\|(2) I/O. Will emulate queuing by
522 coalescing adjacent IOs into a single submission.
525 Basic \fBpreadv\fR\|(2) or \fBpwritev\fR\|(2) I/O.
528 Linux native asynchronous I/O. This ioengine defines engine specific options.
531 POSIX asynchronous I/O using \fBaio_read\fR\|(3) and \fBaio_write\fR\|(3).
534 Solaris native asynchronous I/O.
537 Windows native asynchronous I/O.
540 File is memory mapped with \fBmmap\fR\|(2) and data copied using
544 \fBsplice\fR\|(2) is used to transfer the data and \fBvmsplice\fR\|(2) to
545 transfer data from user-space to the kernel.
548 Use the syslet system calls to make regular read/write asynchronous.
551 SCSI generic sg v3 I/O. May be either synchronous using the SG_IO ioctl, or if
552 the target is an sg character device, we use \fBread\fR\|(2) and
553 \fBwrite\fR\|(2) for asynchronous I/O.
556 Doesn't transfer any data, just pretends to. Mainly used to exercise \fBfio\fR
557 itself and for debugging and testing purposes.
560 Transfer over the network. The protocol to be used can be defined with the
561 \fBprotocol\fR parameter. Depending on the protocol, \fBfilename\fR,
562 \fBhostname\fR, \fBport\fR, or \fBlisten\fR must be specified.
563 This ioengine defines engine specific options.
566 Like \fBnet\fR, but uses \fBsplice\fR\|(2) and \fBvmsplice\fR\|(2) to map data
567 and send/receive. This ioengine defines engine specific options.
570 Doesn't transfer any data, but burns CPU cycles according to \fBcpuload\fR and
571 \fBcpucycles\fR parameters.
574 The GUASI I/O engine is the Generic Userspace Asynchronous Syscall Interface
575 approach to asynchronous I/O.
577 See <http://www.xmailserver.org/guasi\-lib.html>.
580 The RDMA I/O engine supports both RDMA memory semantics (RDMA_WRITE/RDMA_READ)
581 and channel semantics (Send/Recv) for the InfiniBand, RoCE and iWARP protocols.
584 Loads an external I/O engine object file. Append the engine filename as
588 IO engine that does regular linux native fallocate call to simulate data
589 transfer as fio ioengine
591 DDIR_READ does fallocate(,mode = FALLOC_FL_KEEP_SIZE,)
593 DIR_WRITE does fallocate(,mode = 0)
595 DDIR_TRIM does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE)
598 IO engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate defragment activity
599 request to DDIR_WRITE event
602 IO engine supporting direct access to Ceph Rados Block Devices (RBD) via librbd
603 without the need to use the kernel rbd driver. This ioengine defines engine specific
607 Using Glusterfs libgfapi sync interface to direct access to Glusterfs volumes without
608 having to go through FUSE. This ioengine defines engine specific
612 Using Glusterfs libgfapi async interface to direct access to Glusterfs volumes without
613 having to go through FUSE. This ioengine defines engine specific
617 Read and write through Hadoop (HDFS)
622 .BI iodepth \fR=\fPint
623 Number of I/O units to keep in flight against the file. Note that increasing
624 iodepth beyond 1 will not affect synchronous ioengines (except for small
625 degress when verify_async is in use). Even async engines may impose OS
626 restrictions causing the desired depth not to be achieved. This may happen on
627 Linux when using libaio and not setting \fBdirect\fR=1, since buffered IO is
628 not async on that OS. Keep an eye on the IO depth distribution in the
629 fio output to verify that the achieved depth is as expected. Default: 1.
631 .BI iodepth_batch \fR=\fPint
632 Number of I/Os to submit at once. Default: \fBiodepth\fR.
634 .BI iodepth_batch_complete \fR=\fPint
635 This defines how many pieces of IO to retrieve at once. It defaults to 1 which
636 means that we'll ask for a minimum of 1 IO in the retrieval process from the
637 kernel. The IO retrieval will go on until we hit the limit set by
638 \fBiodepth_low\fR. If this variable is set to 0, then fio will always check for
639 completed events before queuing more IO. This helps reduce IO latency, at the
640 cost of more retrieval system calls.
642 .BI iodepth_low \fR=\fPint
643 Low watermark indicating when to start filling the queue again. Default:
646 .BI direct \fR=\fPbool
647 If true, use non-buffered I/O (usually O_DIRECT). Default: false.
649 .BI atomic \fR=\fPbool
650 If value is true, attempt to use atomic direct IO. Atomic writes are guaranteed
651 to be stable once acknowledged by the operating system. Only Linux supports
654 .BI buffered \fR=\fPbool
655 If true, use buffered I/O. This is the opposite of the \fBdirect\fR parameter.
658 .BI offset \fR=\fPint
659 Offset in the file to start I/O. Data before the offset will not be touched.
661 .BI offset_increment \fR=\fPint
662 If this is provided, then the real offset becomes the
663 offset + offset_increment * thread_number, where the thread number is a
664 counter that starts at 0 and is incremented for each sub-job (i.e. when
665 numjobs option is specified). This option is useful if there are several jobs
666 which are intended to operate on a file in parallel disjoint segments, with
667 even spacing between the starting points.
669 .BI number_ios \fR=\fPint
670 Fio will normally perform IOs until it has exhausted the size of the region
671 set by \fBsize\fR, or if it exhaust the allocated time (or hits an error
672 condition). With this setting, the range/size can be set independently of
673 the number of IOs to perform. When fio reaches this number, it will exit
674 normally and report status.
677 How many I/Os to perform before issuing an \fBfsync\fR\|(2) of dirty data. If
678 0, don't sync. Default: 0.
680 .BI fdatasync \fR=\fPint
681 Like \fBfsync\fR, but uses \fBfdatasync\fR\|(2) instead to only sync the
682 data parts of the file. Default: 0.
684 .BI write_barrier \fR=\fPint
685 Make every Nth write a barrier write.
687 .BI sync_file_range \fR=\fPstr:int
688 Use \fBsync_file_range\fR\|(2) for every \fRval\fP number of write operations. Fio will
689 track range of writes that have happened since the last \fBsync_file_range\fR\|(2) call.
690 \fRstr\fP can currently be one or more of:
694 SYNC_FILE_RANGE_WAIT_BEFORE
697 SYNC_FILE_RANGE_WRITE
700 SYNC_FILE_RANGE_WRITE
704 So if you do sync_file_range=wait_before,write:8, fio would use
705 \fBSYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE\fP for every 8 writes.
706 Also see the \fBsync_file_range\fR\|(2) man page. This option is Linux specific.
708 .BI overwrite \fR=\fPbool
709 If writing, setup the file first and do overwrites. Default: false.
711 .BI end_fsync \fR=\fPbool
712 Sync file contents when a write stage has completed. Default: false.
714 .BI fsync_on_close \fR=\fPbool
715 If true, sync file contents on close. This differs from \fBend_fsync\fR in that
716 it will happen on every close, not just at the end of the job. Default: false.
718 .BI rwmixread \fR=\fPint
719 Percentage of a mixed workload that should be reads. Default: 50.
721 .BI rwmixwrite \fR=\fPint
722 Percentage of a mixed workload that should be writes. If \fBrwmixread\fR and
723 \fBrwmixwrite\fR are given and do not sum to 100%, the latter of the two
724 overrides the first. This may interfere with a given rate setting, if fio is
725 asked to limit reads or writes to a certain rate. If that is the case, then
726 the distribution may be skewed. Default: 50.
728 .BI random_distribution \fR=\fPstr:float
729 By default, fio will use a completely uniform random distribution when asked
730 to perform random IO. Sometimes it is useful to skew the distribution in
731 specific ways, ensuring that some parts of the data is more hot than others.
732 Fio includes the following distribution models:
736 Uniform random distribution
746 When using a zipf or pareto distribution, an input value is also needed to
747 define the access pattern. For zipf, this is the zipf theta. For pareto,
748 it's the pareto power. Fio includes a test program, genzipf, that can be
749 used visualize what the given input values will yield in terms of hit rates.
750 If you wanted to use zipf with a theta of 1.2, you would use
751 random_distribution=zipf:1.2 as the option. If a non-uniform model is used,
752 fio will disable use of the random map.
754 .BI percentage_random \fR=\fPint
755 For a random workload, set how big a percentage should be random. This defaults
756 to 100%, in which case the workload is fully random. It can be set from
757 anywhere from 0 to 100. Setting it to 0 would make the workload fully
758 sequential. It is possible to set different values for reads, writes, and
759 trim. To do so, simply use a comma separated list. See \fBblocksize\fR.
762 Normally \fBfio\fR will cover every block of the file when doing random I/O. If
763 this parameter is given, a new offset will be chosen without looking at past
764 I/O history. This parameter is mutually exclusive with \fBverify\fR.
766 .BI softrandommap \fR=\fPbool
767 See \fBnorandommap\fR. If fio runs with the random block map enabled and it
768 fails to allocate the map, if this option is set it will continue without a
769 random block map. As coverage will not be as complete as with random maps, this
770 option is disabled by default.
772 .BI random_generator \fR=\fPstr
773 Fio supports the following engines for generating IO offsets for random IO:
777 Strong 2^88 cycle random number generator
780 Linear feedback shift register generator
784 Tausworthe is a strong random number generator, but it requires tracking on the
785 side if we want to ensure that blocks are only read or written once. LFSR
786 guarantees that we never generate the same offset twice, and it's also less
787 computationally expensive. It's not a true random generator, however, though
788 for IO purposes it's typically good enough. LFSR only works with single block
789 sizes, not with workloads that use multiple block sizes. If used with such a
790 workload, fio may read or write some blocks multiple times.
793 Run job with given nice value. See \fBnice\fR\|(2).
796 Set I/O priority value of this job between 0 (highest) and 7 (lowest). See
799 .BI prioclass \fR=\fPint
800 Set I/O priority class. See \fBionice\fR\|(1).
802 .BI thinktime \fR=\fPint
803 Stall job for given number of microseconds between issuing I/Os.
805 .BI thinktime_spin \fR=\fPint
806 Pretend to spend CPU time for given number of microseconds, sleeping the rest
807 of the time specified by \fBthinktime\fR. Only valid if \fBthinktime\fR is set.
809 .BI thinktime_blocks \fR=\fPint
810 Only valid if thinktime is set - control how many blocks to issue, before
811 waiting \fBthinktime\fR microseconds. If not set, defaults to 1 which will
812 make fio wait \fBthinktime\fR microseconds after every block. This
813 effectively makes any queue depth setting redundant, since no more than 1 IO
814 will be queued before we have to complete it and do our thinktime. In other
815 words, this setting effectively caps the queue depth if the latter is larger.
819 Cap bandwidth used by this job. The number is in bytes/sec, the normal postfix
820 rules apply. You can use \fBrate\fR=500k to limit reads and writes to 500k each,
821 or you can specify read and writes separately. Using \fBrate\fR=1m,500k would
822 limit reads to 1MB/sec and writes to 500KB/sec. Capping only reads or writes
823 can be done with \fBrate\fR=,500k or \fBrate\fR=500k,. The former will only
824 limit writes (to 500KB/sec), the latter will only limit reads.
826 .BI ratemin \fR=\fPint
827 Tell \fBfio\fR to do whatever it can to maintain at least the given bandwidth.
828 Failing to meet this requirement will cause the job to exit. The same format
829 as \fBrate\fR is used for read vs write separation.
831 .BI rate_iops \fR=\fPint
832 Cap the bandwidth to this number of IOPS. Basically the same as rate, just
833 specified independently of bandwidth. The same format as \fBrate\fR is used for
834 read vs write separation. If \fBblocksize\fR is a range, the smallest block
835 size is used as the metric.
837 .BI rate_iops_min \fR=\fPint
838 If this rate of I/O is not met, the job will exit. The same format as \fBrate\fR
839 is used for read vs write separation.
841 .BI ratecycle \fR=\fPint
842 Average bandwidth for \fBrate\fR and \fBratemin\fR over this number of
843 milliseconds. Default: 1000ms.
845 .BI latency_target \fR=\fPint
846 If set, fio will attempt to find the max performance point that the given
847 workload will run at while maintaining a latency below this target. The
848 values is given in microseconds. See \fBlatency_window\fR and
849 \fBlatency_percentile\fR.
851 .BI latency_window \fR=\fPint
852 Used with \fBlatency_target\fR to specify the sample window that the job
853 is run at varying queue depths to test the performance. The value is given
856 .BI latency_percentile \fR=\fPfloat
857 The percentage of IOs that must fall within the criteria specified by
858 \fBlatency_target\fR and \fBlatency_window\fR. If not set, this defaults
859 to 100.0, meaning that all IOs must be equal or below to the value set
860 by \fBlatency_target\fR.
862 .BI max_latency \fR=\fPint
863 If set, fio will exit the job if it exceeds this maximum latency. It will exit
866 .BI cpumask \fR=\fPint
867 Set CPU affinity for this job. \fIint\fR is a bitmask of allowed CPUs the job
868 may run on. See \fBsched_setaffinity\fR\|(2).
870 .BI cpus_allowed \fR=\fPstr
871 Same as \fBcpumask\fR, but allows a comma-delimited list of CPU numbers.
873 .BI cpus_allowed_policy \fR=\fPstr
874 Set the policy of how fio distributes the CPUs specified by \fBcpus_allowed\fR
875 or \fBcpumask\fR. Two policies are supported:
880 All jobs will share the CPU set specified.
883 Each job will get a unique CPU from the CPU set.
886 \fBshared\fR is the default behaviour, if the option isn't specified. If
887 \fBsplit\fR is specified, then fio will assign one cpu per job. If not enough
888 CPUs are given for the jobs listed, then fio will roundrobin the CPUs in
893 .BI numa_cpu_nodes \fR=\fPstr
894 Set this job running on specified NUMA nodes' CPUs. The arguments allow
895 comma delimited list of cpu numbers, A-B ranges, or 'all'.
897 .BI numa_mem_policy \fR=\fPstr
898 Set this job's memory policy and corresponding NUMA nodes. Format of
902 .B <mode>[:<nodelist>]
905 is one of the following memory policy:
907 .B default, prefer, bind, interleave, local
910 For \fBdefault\fR and \fBlocal\fR memory policy, no \fBnodelist\fR is
911 needed to be specified. For \fBprefer\fR, only one node is
912 allowed. For \fBbind\fR and \fBinterleave\fR, \fBnodelist\fR allows
913 comma delimited list of numbers, A-B ranges, or 'all'.
915 .BI startdelay \fR=\fPirange
916 Delay start of job for the specified number of seconds. Supports all time
917 suffixes to allow specification of hours, minutes, seconds and
918 milliseconds - seconds are the default if a unit is ommited.
919 Can be given as a range which causes each thread to choose randomly out of the
922 .BI runtime \fR=\fPint
923 Terminate processing after the specified number of seconds.
926 If given, run for the specified \fBruntime\fR duration even if the files are
927 completely read or written. The same workload will be repeated as many times
928 as \fBruntime\fR allows.
930 .BI ramp_time \fR=\fPint
931 If set, fio will run the specified workload for this amount of time before
932 logging any performance numbers. Useful for letting performance settle before
933 logging results, thus minimizing the runtime required for stable results. Note
934 that the \fBramp_time\fR is considered lead in time for a job, thus it will
935 increase the total runtime if a special timeout or runtime is specified.
937 .BI invalidate \fR=\fPbool
938 Invalidate buffer-cache for the file prior to starting I/O. Default: true.
941 Use synchronous I/O for buffered writes. For the majority of I/O engines,
942 this means using O_SYNC. Default: false.
944 .BI iomem \fR=\fPstr "\fR,\fP mem" \fR=\fPstr
945 Allocation method for I/O unit buffer. Allowed values are:
950 Allocate memory with \fBmalloc\fR\|(3).
953 Use shared memory buffers allocated through \fBshmget\fR\|(2).
956 Same as \fBshm\fR, but use huge pages as backing.
959 Use \fBmmap\fR\|(2) for allocation. Uses anonymous memory unless a filename
960 is given after the option in the format `:\fIfile\fR'.
963 Same as \fBmmap\fR, but use huge files as backing.
966 The amount of memory allocated is the maximum allowed \fBblocksize\fR for the
967 job multiplied by \fBiodepth\fR. For \fBshmhuge\fR or \fBmmaphuge\fR to work,
968 the system must have free huge pages allocated. \fBmmaphuge\fR also needs to
969 have hugetlbfs mounted, and \fIfile\fR must point there. At least on Linux,
970 huge pages must be manually allocated. See \fB/proc/sys/vm/nr_hugehages\fR
971 and the documentation for that. Normally you just need to echo an appropriate
972 number, eg echoing 8 will ensure that the OS has 8 huge pages ready for
976 .BI iomem_align \fR=\fPint "\fR,\fP mem_align" \fR=\fPint
977 This indicates the memory alignment of the IO memory buffers. Note that the
978 given alignment is applied to the first IO unit buffer, if using \fBiodepth\fR
979 the alignment of the following buffers are given by the \fBbs\fR used. In
980 other words, if using a \fBbs\fR that is a multiple of the page sized in the
981 system, all buffers will be aligned to this value. If using a \fBbs\fR that
982 is not page aligned, the alignment of subsequent IO memory buffers is the
983 sum of the \fBiomem_align\fR and \fBbs\fR used.
985 .BI hugepage\-size \fR=\fPint
986 Defines the size of a huge page. Must be at least equal to the system setting.
987 Should be a multiple of 1MB. Default: 4MB.
990 Terminate all jobs when one finishes. Default: wait for each job to finish.
992 .BI bwavgtime \fR=\fPint
993 Average bandwidth calculations over the given time in milliseconds. Default:
996 .BI iopsavgtime \fR=\fPint
997 Average IOPS calculations over the given time in milliseconds. Default:
1000 .BI create_serialize \fR=\fPbool
1001 If true, serialize file creation for the jobs. Default: true.
1003 .BI create_fsync \fR=\fPbool
1004 \fBfsync\fR\|(2) data file after creation. Default: true.
1006 .BI create_on_open \fR=\fPbool
1007 If true, the files are not created until they are opened for IO by the job.
1009 .BI create_only \fR=\fPbool
1010 If true, fio will only run the setup phase of the job. If files need to be
1011 laid out or updated on disk, only that will be done. The actual job contents
1014 .BI pre_read \fR=\fPbool
1015 If this is given, files will be pre-read into memory before starting the given
1016 IO operation. This will also clear the \fR \fBinvalidate\fR flag, since it is
1017 pointless to pre-read and then drop the cache. This will only work for IO
1018 engines that are seekable, since they allow you to read the same data
1019 multiple times. Thus it will not work on eg network or splice IO.
1021 .BI unlink \fR=\fPbool
1022 Unlink job files when done. Default: false.
1024 .BI loops \fR=\fPint
1025 Specifies the number of iterations (runs of the same workload) of this job.
1028 .BI verify_only \fR=\fPbool
1029 Do not perform the specified workload, only verify data still matches previous
1030 invocation of this workload. This option allows one to check data multiple
1031 times at a later date without overwriting it. This option makes sense only for
1032 workloads that write data, and does not support workloads with the
1033 \fBtime_based\fR option set.
1035 .BI do_verify \fR=\fPbool
1036 Run the verify phase after a write phase. Only valid if \fBverify\fR is set.
1039 .BI verify \fR=\fPstr
1040 Method of verifying file contents after each iteration of the job. Allowed
1045 .B md5 crc16 crc32 crc32c crc32c-intel crc64 crc7 sha256 sha512 sha1 xxhash
1046 Store appropriate checksum in the header of each block. crc32c-intel is
1047 hardware accelerated SSE4.2 driven, falls back to regular crc32c if
1048 not supported by the system.
1051 Write extra information about each I/O (timestamp, block number, etc.). The
1052 block number is verified. See \fBverify_pattern\fR as well.
1055 Pretend to verify. Used for testing internals.
1058 This option can be used for repeated burn-in tests of a system to make sure
1059 that the written data is also correctly read back. If the data direction given
1060 is a read or random read, fio will assume that it should verify a previously
1061 written file. If the data direction includes any form of write, the verify will
1062 be of the newly written data.
1065 .BI verifysort \fR=\fPbool
1066 If true, written verify blocks are sorted if \fBfio\fR deems it to be faster to
1067 read them back in a sorted manner. Default: true.
1069 .BI verifysort_nr \fR=\fPint
1070 Pre-load and sort verify blocks for a read workload.
1072 .BI verify_offset \fR=\fPint
1073 Swap the verification header with data somewhere else in the block before
1074 writing. It is swapped back before verifying.
1076 .BI verify_interval \fR=\fPint
1077 Write the verification header for this number of bytes, which should divide
1078 \fBblocksize\fR. Default: \fBblocksize\fR.
1080 .BI verify_pattern \fR=\fPstr
1081 If set, fio will fill the io buffers with this pattern. Fio defaults to filling
1082 with totally random bytes, but sometimes it's interesting to fill with a known
1083 pattern for io verification purposes. Depending on the width of the pattern,
1084 fio will fill 1/2/3/4 bytes of the buffer at the time(it can be either a
1085 decimal or a hex number). The verify_pattern if larger than a 32-bit quantity
1086 has to be a hex number that starts with either "0x" or "0X". Use with
1089 .BI verify_fatal \fR=\fPbool
1090 If true, exit the job on the first observed verification failure. Default:
1093 .BI verify_dump \fR=\fPbool
1094 If set, dump the contents of both the original data block and the data block we
1095 read off disk to files. This allows later analysis to inspect just what kind of
1096 data corruption occurred. Off by default.
1098 .BI verify_async \fR=\fPint
1099 Fio will normally verify IO inline from the submitting thread. This option
1100 takes an integer describing how many async offload threads to create for IO
1101 verification instead, causing fio to offload the duty of verifying IO contents
1102 to one or more separate threads. If using this offload option, even sync IO
1103 engines can benefit from using an \fBiodepth\fR setting higher than 1, as it
1104 allows them to have IO in flight while verifies are running.
1106 .BI verify_async_cpus \fR=\fPstr
1107 Tell fio to set the given CPU affinity on the async IO verification threads.
1108 See \fBcpus_allowed\fP for the format used.
1110 .BI verify_backlog \fR=\fPint
1111 Fio will normally verify the written contents of a job that utilizes verify
1112 once that job has completed. In other words, everything is written then
1113 everything is read back and verified. You may want to verify continually
1114 instead for a variety of reasons. Fio stores the meta data associated with an
1115 IO block in memory, so for large verify workloads, quite a bit of memory would
1116 be used up holding this meta data. If this option is enabled, fio will write
1117 only N blocks before verifying these blocks.
1119 .BI verify_backlog_batch \fR=\fPint
1120 Control how many blocks fio will verify if verify_backlog is set. If not set,
1121 will default to the value of \fBverify_backlog\fR (meaning the entire queue is
1122 read back and verified). If \fBverify_backlog_batch\fR is less than
1123 \fBverify_backlog\fR then not all blocks will be verified, if
1124 \fBverify_backlog_batch\fR is larger than \fBverify_backlog\fR, some blocks
1125 will be verified more than once.
1127 .BI trim_percentage \fR=\fPint
1128 Number of verify blocks to discard/trim.
1130 .BI trim_verify_zero \fR=\fPbool
1131 Verify that trim/discarded blocks are returned as zeroes.
1133 .BI trim_backlog \fR=\fPint
1134 Trim after this number of blocks are written.
1136 .BI trim_backlog_batch \fR=\fPint
1137 Trim this number of IO blocks.
1139 .BI experimental_verify \fR=\fPbool
1140 Enable experimental verification.
1142 .B stonewall "\fR,\fP wait_for_previous"
1143 Wait for preceding jobs in the job file to exit before starting this one.
1144 \fBstonewall\fR implies \fBnew_group\fR.
1147 Start a new reporting group. If not given, all jobs in a file will be part
1148 of the same reporting group, unless separated by a stonewall.
1150 .BI numjobs \fR=\fPint
1151 Number of clones (processes/threads performing the same workload) of this job.
1155 If set, display per-group reports instead of per-job when \fBnumjobs\fR is
1159 Use threads created with \fBpthread_create\fR\|(3) instead of processes created
1160 with \fBfork\fR\|(2).
1162 .BI zonesize \fR=\fPint
1163 Divide file into zones of the specified size in bytes. See \fBzoneskip\fR.
1165 .BI zonerange \fR=\fPint
1166 Give size of an IO zone. See \fBzoneskip\fR.
1168 .BI zoneskip \fR=\fPint
1169 Skip the specified number of bytes when \fBzonesize\fR bytes of data have been
1172 .BI write_iolog \fR=\fPstr
1173 Write the issued I/O patterns to the specified file. Specify a separate file
1174 for each job, otherwise the iologs will be interspersed and the file may be
1177 .BI read_iolog \fR=\fPstr
1178 Replay the I/O patterns contained in the specified file generated by
1179 \fBwrite_iolog\fR, or may be a \fBblktrace\fR binary file.
1181 .BI replay_no_stall \fR=\fPint
1182 While replaying I/O patterns using \fBread_iolog\fR the default behavior
1183 attempts to respect timing information between I/Os. Enabling
1184 \fBreplay_no_stall\fR causes I/Os to be replayed as fast as possible while
1185 still respecting ordering.
1187 .BI replay_redirect \fR=\fPstr
1188 While replaying I/O patterns using \fBread_iolog\fR the default behavior
1189 is to replay the IOPS onto the major/minor device that each IOP was recorded
1190 from. Setting \fBreplay_redirect\fR causes all IOPS to be replayed onto the
1191 single specified device regardless of the device it was recorded from.
1193 .BI write_bw_log \fR=\fPstr
1194 If given, write a bandwidth log of the jobs in this job file. Can be used to
1195 store data of the bandwidth of the jobs in their lifetime. The included
1196 fio_generate_plots script uses gnuplot to turn these text files into nice
1197 graphs. See \fBwrite_lat_log\fR for behaviour of given filename. For this
1198 option, the postfix is _bw.x.log, where x is the index of the job (1..N,
1199 where N is the number of jobs)
1201 .BI write_lat_log \fR=\fPstr
1202 Same as \fBwrite_bw_log\fR, but writes I/O completion latencies. If no
1203 filename is given with this option, the default filename of
1204 "jobname_type.x.log" is used, where x is the index of the job (1..N, where
1205 N is the number of jobs). Even if the filename is given, fio will still
1206 append the type of log.
1208 .BI write_iops_log \fR=\fPstr
1209 Same as \fBwrite_bw_log\fR, but writes IOPS. If no filename is given with this
1210 option, the default filename of "jobname_type.x.log" is used, where x is the
1211 index of the job (1..N, where N is the number of jobs). Even if the filename
1212 is given, fio will still append the type of log.
1214 .BI log_avg_msec \fR=\fPint
1215 By default, fio will log an entry in the iops, latency, or bw log for every
1216 IO that completes. When writing to the disk log, that can quickly grow to a
1217 very large size. Setting this option makes fio average the each log entry
1218 over the specified period of time, reducing the resolution of the log.
1221 .BI log_offset \fR=\fPbool
1222 If this is set, the iolog options will include the byte offset for the IO
1223 entry as well as the other data values.
1225 .BI log_compression \fR=\fPint
1226 If this is set, fio will compress the IO logs as it goes, to keep the memory
1227 footprint lower. When a log reaches the specified size, that chunk is removed
1228 and compressed in the background. Given that IO logs are fairly highly
1229 compressible, this yields a nice memory savings for longer runs. The downside
1230 is that the compression will consume some background CPU cycles, so it may
1231 impact the run. This, however, is also true if the logging ends up consuming
1232 most of the system memory. So pick your poison. The IO logs are saved
1233 normally at the end of a run, by decompressing the chunks and storing them
1234 in the specified log file. This feature depends on the availability of zlib.
1236 .BI log_store_compressed \fR=\fPbool
1237 If set, and \fBlog\fR_compression is also set, fio will store the log files in
1238 a compressed format. They can be decompressed with fio, using the
1239 \fB\-\-inflate-log\fR command line parameter. The files will be stored with a
1242 .BI disable_lat \fR=\fPbool
1243 Disable measurements of total latency numbers. Useful only for cutting
1244 back the number of calls to \fBgettimeofday\fR\|(2), as that does impact performance at
1245 really high IOPS rates. Note that to really get rid of a large amount of these
1246 calls, this option must be used with disable_slat and disable_bw as well.
1248 .BI disable_clat \fR=\fPbool
1249 Disable measurements of completion latency numbers. See \fBdisable_lat\fR.
1251 .BI disable_slat \fR=\fPbool
1252 Disable measurements of submission latency numbers. See \fBdisable_lat\fR.
1254 .BI disable_bw_measurement \fR=\fPbool
1255 Disable measurements of throughput/bandwidth numbers. See \fBdisable_lat\fR.
1257 .BI lockmem \fR=\fPint
1258 Pin the specified amount of memory with \fBmlock\fR\|(2). Can be used to
1259 simulate a smaller amount of memory. The amount specified is per worker.
1261 .BI exec_prerun \fR=\fPstr
1262 Before running the job, execute the specified command with \fBsystem\fR\|(3).
1264 Output is redirected in a file called \fBjobname.prerun.txt\fR
1267 .BI exec_postrun \fR=\fPstr
1268 Same as \fBexec_prerun\fR, but the command is executed after the job completes.
1270 Output is redirected in a file called \fBjobname.postrun.txt\fR
1273 .BI ioscheduler \fR=\fPstr
1274 Attempt to switch the device hosting the file to the specified I/O scheduler.
1276 .BI disk_util \fR=\fPbool
1277 Generate disk utilization statistics if the platform supports it. Default: true.
1279 .BI clocksource \fR=\fPstr
1280 Use the given clocksource as the base of timing. The supported options are:
1284 \fBgettimeofday\fR\|(2)
1287 \fBclock_gettime\fR\|(2)
1290 Internal CPU clock source
1294 \fBcpu\fR is the preferred clocksource if it is reliable, as it is very fast
1295 (and fio is heavy on time calls). Fio will automatically use this clocksource
1296 if it's supported and considered reliable on the system it is running on,
1297 unless another clocksource is specifically set. For x86/x86-64 CPUs, this
1298 means supporting TSC Invariant.
1300 .BI gtod_reduce \fR=\fPbool
1301 Enable all of the \fBgettimeofday\fR\|(2) reducing options (disable_clat, disable_slat,
1302 disable_bw) plus reduce precision of the timeout somewhat to really shrink the
1303 \fBgettimeofday\fR\|(2) call count. With this option enabled, we only do about 0.4% of
1304 the gtod() calls we would have done if all time keeping was enabled.
1306 .BI gtod_cpu \fR=\fPint
1307 Sometimes it's cheaper to dedicate a single thread of execution to just getting
1308 the current time. Fio (and databases, for instance) are very intensive on
1309 \fBgettimeofday\fR\|(2) calls. With this option, you can set one CPU aside for doing
1310 nothing but logging current time to a shared memory location. Then the other
1311 threads/processes that run IO workloads need only copy that segment, instead of
1312 entering the kernel with a \fBgettimeofday\fR\|(2) call. The CPU set aside for doing
1313 these time calls will be excluded from other uses. Fio will manually clear it
1314 from the CPU mask of other jobs.
1316 .BI ignore_error \fR=\fPstr
1317 Sometimes you want to ignore some errors during test in that case you can specify
1318 error list for each error type.
1320 ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST
1322 errors for given error type is separated with ':'.
1323 Error may be symbol ('ENOSPC', 'ENOMEM') or an integer.
1325 Example: ignore_error=EAGAIN,ENOSPC:122 .
1327 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from WRITE.
1329 .BI error_dump \fR=\fPbool
1330 If set dump every error even if it is non fatal, true by default. If disabled
1331 only fatal error will be dumped
1333 .BI profile \fR=\fPstr
1334 Select a specific builtin performance test.
1336 .BI cgroup \fR=\fPstr
1337 Add job to this control group. If it doesn't exist, it will be created.
1338 The system must have a mounted cgroup blkio mount point for this to work. If
1339 your system doesn't have it mounted, you can do so with:
1341 # mount \-t cgroup \-o blkio none /cgroup
1343 .BI cgroup_weight \fR=\fPint
1344 Set the weight of the cgroup to this value. See the documentation that comes
1345 with the kernel, allowed values are in the range of 100..1000.
1347 .BI cgroup_nodelete \fR=\fPbool
1348 Normally fio will delete the cgroups it has created after the job completion.
1349 To override this behavior and to leave cgroups around after the job completion,
1350 set cgroup_nodelete=1. This can be useful if one wants to inspect various
1351 cgroup files after job completion. Default: false
1354 Instead of running as the invoking user, set the user ID to this value before
1355 the thread/process does any work.
1358 Set group ID, see \fBuid\fR.
1360 .BI unit_base \fR=\fPint
1361 Base unit for reporting. Allowed values are:
1365 Use auto-detection (default).
1375 .BI flow_id \fR=\fPint
1376 The ID of the flow. If not specified, it defaults to being a global flow. See
1380 Weight in token-based flow control. If this value is used, then there is a
1381 \fBflow counter\fR which is used to regulate the proportion of activity between
1382 two or more jobs. fio attempts to keep this flow counter near zero. The
1383 \fBflow\fR parameter stands for how much should be added or subtracted to the
1384 flow counter on each iteration of the main I/O loop. That is, if one job has
1385 \fBflow=8\fR and another job has \fBflow=-1\fR, then there will be a roughly
1386 1:8 ratio in how much one runs vs the other.
1388 .BI flow_watermark \fR=\fPint
1389 The maximum value that the absolute value of the flow counter is allowed to
1390 reach before the job must wait for a lower value of the counter.
1392 .BI flow_sleep \fR=\fPint
1393 The period of time, in microseconds, to wait after the flow watermark has been
1394 exceeded before retrying operations
1396 .BI clat_percentiles \fR=\fPbool
1397 Enable the reporting of percentiles of completion latencies.
1399 .BI percentile_list \fR=\fPfloat_list
1400 Overwrite the default list of percentiles for completion
1401 latencies. Each number is a floating number in the range (0,100], and
1402 the maximum length of the list is 20. Use ':' to separate the
1403 numbers. For example, \-\-percentile_list=99.5:99.9 will cause fio to
1404 report the values of completion latency below which 99.5% and 99.9% of
1405 the observed latencies fell, respectively.
1406 .SS "Ioengine Parameters List"
1407 Some parameters are only valid when a specific ioengine is in use. These are
1408 used identically to normal parameters, with the caveat that when used on the
1409 command line, they must come after the ioengine.
1411 .BI (cpu)cpuload \fR=\fPint
1412 Attempt to use the specified percentage of CPU cycles.
1414 .BI (cpu)cpuchunks \fR=\fPint
1415 Split the load into cycles of the given time. In microseconds.
1417 .BI (cpu)exit_on_io_done \fR=\fPbool
1418 Detect when IO threads are done, then exit.
1420 .BI (libaio)userspace_reap
1421 Normally, with the libaio engine in use, fio will use
1422 the io_getevents system call to reap newly returned events.
1423 With this flag turned on, the AIO ring will be read directly
1424 from user-space to reap events. The reaping mode is only
1425 enabled when polling for a minimum of 0 events (eg when
1426 iodepth_batch_complete=0).
1428 .BI (net,netsplice)hostname \fR=\fPstr
1429 The host name or IP address to use for TCP or UDP based IO.
1430 If the job is a TCP listener or UDP reader, the hostname is not
1431 used and must be omitted unless it is a valid UDP multicast address.
1433 .BI (net,netsplice)port \fR=\fPint
1434 The TCP or UDP port to bind to or connect to.
1436 .BI (net,netsplice)interface \fR=\fPstr
1437 The IP address of the network interface used to send or receive UDP multicast
1440 .BI (net,netsplice)ttl \fR=\fPint
1441 Time-to-live value for outgoing UDP multicast packets. Default: 1
1443 .BI (net,netsplice)nodelay \fR=\fPbool
1444 Set TCP_NODELAY on TCP connections.
1446 .BI (net,netsplice)protocol \fR=\fPstr "\fR,\fP proto" \fR=\fPstr
1447 The network protocol to use. Accepted values are:
1452 Transmission control protocol
1455 Transmission control protocol V6
1458 User datagram protocol
1461 User datagram protocol V6
1467 When the protocol is TCP or UDP, the port must also be given,
1468 as well as the hostname if the job is a TCP listener or UDP
1469 reader. For unix sockets, the normal filename option should be
1470 used and the port is invalid.
1473 .BI (net,netsplice)listen
1474 For TCP network connections, tell fio to listen for incoming
1475 connections rather than initiating an outgoing connection. The
1476 hostname must be omitted if this option is used.
1478 .BI (net, pingpong) \fR=\fPbool
1479 Normally a network writer will just continue writing data, and a network reader
1480 will just consume packets. If pingpong=1 is set, a writer will send its normal
1481 payload to the reader, then wait for the reader to send the same payload back.
1482 This allows fio to measure network latencies. The submission and completion
1483 latencies then measure local time spent sending or receiving, and the
1484 completion latency measures how long it took for the other end to receive and
1485 send back. For UDP multicast traffic pingpong=1 should only be set for a single
1486 reader when multiple readers are listening to the same address.
1488 .BI (e4defrag,donorname) \fR=\fPstr
1489 File will be used as a block donor (swap extents between files)
1491 .BI (e4defrag,inplace) \fR=\fPint
1492 Configure donor file block allocation strategy
1495 Preallocate donor's file on init
1498 allocate space immediately inside defragment event, and free right after event
1501 .BI (rbd)rbdname \fR=\fPstr
1502 Specifies the name of the RBD.
1504 .BI (rbd)pool \fR=\fPstr
1505 Specifies the name of the Ceph pool containing the RBD.
1507 .BI (rbd)clientname \fR=\fPstr
1508 Specifies the username (without the 'client.' prefix) used to access the Ceph cluster.
1510 While running, \fBfio\fR will display the status of the created jobs. For
1514 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
1517 The characters in the first set of brackets denote the current status of each
1518 threads. The possible values are:
1524 Setup but not started.
1530 Initialized, waiting.
1533 Running, doing sequential reads.
1536 Running, doing random reads.
1539 Running, doing sequential writes.
1542 Running, doing random writes.
1545 Running, doing mixed sequential reads/writes.
1548 Running, doing mixed random reads/writes.
1551 Running, currently waiting for \fBfsync\fR\|(2).
1554 Running, verifying written data.
1557 Exited, not reaped by main thread.
1560 Exited, thread reaped.
1564 The second set of brackets shows the estimated completion percentage of
1565 the current group. The third set shows the read and write I/O rate,
1566 respectively. Finally, the estimated run time of the job is displayed.
1568 When \fBfio\fR completes (or is interrupted by Ctrl-C), it will show data
1569 for each thread, each group of threads, and each disk, in that order.
1571 Per-thread statistics first show the threads client number, group-id, and
1572 error code. The remaining figures are as follows:
1576 Number of megabytes of I/O performed.
1579 Average data rate (bandwidth).
1585 Submission latency minimum, maximum, average and standard deviation. This is
1586 the time it took to submit the I/O.
1589 Completion latency minimum, maximum, average and standard deviation. This
1590 is the time between submission and completion.
1593 Bandwidth minimum, maximum, percentage of aggregate bandwidth received, average
1594 and standard deviation.
1597 CPU usage statistics. Includes user and system time, number of context switches
1598 this thread went through and number of major and minor page faults.
1601 Distribution of I/O depths. Each depth includes everything less than (or equal)
1602 to it, but greater than the previous depth.
1605 Number of read/write requests issued, and number of short read/write requests.
1608 Distribution of I/O completion latencies. The numbers follow the same pattern
1612 The group statistics show:
1617 Number of megabytes I/O performed.
1620 Aggregate bandwidth of threads in the group.
1623 Minimum average bandwidth a thread saw.
1626 Maximum average bandwidth a thread saw.
1629 Shortest runtime of threads in the group.
1632 Longest runtime of threads in the group.
1636 Finally, disk statistics are printed with reads first:
1641 Number of I/Os performed by all groups.
1644 Number of merges in the I/O scheduler.
1647 Number of ticks we kept the disk busy.
1650 Total time spent in the disk queue.
1657 It is also possible to get fio to dump the current output while it is
1658 running, without terminating the job. To do that, send fio the \fBUSR1\fR
1661 If the \fB\-\-minimal\fR / \fB\-\-append-terse\fR options are given, the
1662 results will be printed/appended in a semicolon-delimited format suitable for
1664 A job description (if provided) follows on a new line. Note that the first
1665 number in the line is the version number. If the output has to be changed
1666 for some reason, this number will be incremented by 1 to signify that
1667 change. The fields are:
1670 .B terse version, fio version, jobname, groupid, error
1674 .B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, IOPS, runtime \fR(ms)\fP
1678 .B min, max, mean, standard deviation
1682 .B min, max, mean, standard deviation
1684 Completion latency percentiles (20 fields):
1686 .B Xth percentile=usec
1690 .B min, max, mean, standard deviation
1694 .B min, max, aggregate percentage of total, mean, standard deviation
1700 .B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, IOPS, runtime \fR(ms)\fP
1704 .B min, max, mean, standard deviation
1708 .B min, max, mean, standard deviation
1710 Completion latency percentiles (20 fields):
1712 .B Xth percentile=usec
1716 .B min, max, mean, standard deviation
1720 .B min, max, aggregate percentage of total, mean, standard deviation
1726 .B user, system, context switches, major page faults, minor page faults
1729 IO depth distribution:
1731 .B <=1, 2, 4, 8, 16, 32, >=64
1734 IO latency distribution:
1738 .B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
1742 .B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
1746 Disk utilization (1 for each disk used):
1748 .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
1751 Error Info (dependent on continue_on_error, default off):
1753 .B total # errors, first error code
1756 .B text description (if provided in config - appears on newline)
1759 Normally you would run fio as a stand-alone application on the machine
1760 where the IO workload should be generated. However, it is also possible to
1761 run the frontend and backend of fio separately. This makes it possible to
1762 have a fio server running on the machine(s) where the IO workload should
1763 be running, while controlling it from another machine.
1765 To start the server, you would do:
1767 \fBfio \-\-server=args\fR
1769 on that machine, where args defines what fio listens to. The arguments
1770 are of the form 'type:hostname or IP:port'. 'type' is either 'ip' (or ip4)
1771 for TCP/IP v4, 'ip6' for TCP/IP v6, or 'sock' for a local unix domain
1772 socket. 'hostname' is either a hostname or IP address, and 'port' is the port to
1773 listen to (only valid for TCP/IP, not a local socket). Some examples:
1777 Start a fio server, listening on all interfaces on the default port (8765).
1779 2) fio \-\-server=ip:hostname,4444
1781 Start a fio server, listening on IP belonging to hostname and on port 4444.
1783 3) fio \-\-server=ip6:::1,4444
1785 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
1787 4) fio \-\-server=,4444
1789 Start a fio server, listening on all interfaces on port 4444.
1791 5) fio \-\-server=1.2.3.4
1793 Start a fio server, listening on IP 1.2.3.4 on the default port.
1795 6) fio \-\-server=sock:/tmp/fio.sock
1797 Start a fio server, listening on the local socket /tmp/fio.sock.
1799 When a server is running, you can connect to it from a client. The client
1802 fio \-\-local-args \-\-client=server \-\-remote-args <job file(s)>
1804 where \-\-local-args are arguments that are local to the client where it is
1805 running, 'server' is the connect string, and \-\-remote-args and <job file(s)>
1806 are sent to the server. The 'server' string follows the same format as it
1807 does on the server side, to allow IP/hostname/socket and port strings.
1808 You can connect to multiple clients as well, to do that you could run:
1810 fio \-\-client=server2 \-\-client=server2 <job file(s)>
1814 was written by Jens Axboe <jens.axboe@oracle.com>,
1815 now Jens Axboe <jaxboe@fusionio.com>.
1817 This man page was written by Aaron Carroll <aaronc@cse.unsw.edu.au> based
1818 on documentation by Jens Axboe.
1819 .SH "REPORTING BUGS"
1820 Report bugs to the \fBfio\fR mailing list <fio@vger.kernel.org>.
1823 For further documentation see \fBHOWTO\fR and \fBREADME\fR.
1825 Sample jobfiles are available in the \fBexamples\fR directory.