1 .TH fio 1 "December 2014" "User Manual"
3 fio \- flexible I/O tester
6 [\fIoptions\fR] [\fIjobfile\fR]...
9 is a tool that will spawn a number of threads or processes doing a
10 particular type of I/O action as specified by the user.
11 The typical use of fio is to write a job file matching the I/O load
12 one wants to simulate.
15 .BI \-\-debug \fR=\fPtype
16 Enable verbose tracing of various fio actions. May be `all' for all types
17 or individual types separated by a comma (eg \-\-debug=io,file). `help' will
18 list all available tracing options.
20 .BI \-\-output \fR=\fPfilename
21 Write output to \fIfilename\fR.
23 .BI \-\-output-format \fR=\fPformat
24 Set the reporting format to \fInormal\fR, \fIterse\fR, or \fIjson\fR.
25 Multiple formats can be selected, separate by a comma.
27 .BI \-\-runtime \fR=\fPruntime
28 Limit run time to \fIruntime\fR seconds.
31 Generate per-job bandwidth logs.
34 Print statistics in a terse, semicolon-delimited format.
37 Display version information and exit.
39 .BI \-\-terse\-version \fR=\fPversion
40 Set terse version output format (Current version 3, or older version 2).
43 Display usage information and exit.
46 Perform test and validation of internal CPU clock
48 .BI \-\-crctest[\fR=\fPtest]
49 Test the speed of the builtin checksumming functions. If no argument is given,
50 all of them are tested. Or a comma separated list can be passed, in which
51 case the given ones are tested.
53 .BI \-\-cmdhelp \fR=\fPcommand
54 Print help information for \fIcommand\fR. May be `all' for all commands.
56 .BI \-\-enghelp \fR=\fPioengine[,command]
57 List all commands defined by \fIioengine\fR, or print help for \fIcommand\fR defined by \fIioengine\fR.
59 .BI \-\-showcmd \fR=\fPjobfile
60 Convert \fIjobfile\fR to a set of command-line options.
62 .BI \-\-eta \fR=\fPwhen
63 Specifies when real-time ETA estimate should be printed. \fIwhen\fR may
64 be one of `always', `never' or `auto'.
66 .BI \-\-eta\-newline \fR=\fPtime
67 Force an ETA newline for every `time` period passed.
69 .BI \-\-status\-interval \fR=\fPtime
70 Report full output status every `time` period passed.
73 Turn on safety read-only checks, preventing any attempted write.
75 .BI \-\-section \fR=\fPsec
76 Only run section \fIsec\fR from job file. This option can be used multiple times to add more sections to run.
78 .BI \-\-alloc\-size \fR=\fPkb
79 Set the internal smalloc pool size to \fIkb\fP kilobytes.
81 .BI \-\-warnings\-fatal
82 All fio parser warnings are fatal, causing fio to exit with an error.
84 .BI \-\-max\-jobs \fR=\fPnr
85 Set the maximum allowed number of jobs (threads/processes) to support.
87 .BI \-\-server \fR=\fPargs
88 Start a backend server, with \fIargs\fP specifying what to listen to. See client/server section.
90 .BI \-\-daemonize \fR=\fPpidfile
91 Background a fio server, writing the pid to the given pid file.
93 .BI \-\-client \fR=\fPhost
94 Instead of running the jobs locally, send and run them on the given host or set of hosts. See client/server section.
96 .BI \-\-idle\-prof \fR=\fPoption
97 Report cpu idleness on a system or percpu basis (\fIoption\fP=system,percpu) or run unit work calibration only (\fIoption\fP=calibrate).
99 Job files are in `ini' format. They consist of one or more
100 job definitions, which begin with a job name in square brackets and
101 extend to the next job name. The job name can be any ASCII string
102 except `global', which has a special meaning. Following the job name is
103 a sequence of zero or more parameters, one per line, that define the
104 behavior of the job. Any line starting with a `;' or `#' character is
105 considered a comment and ignored.
107 If \fIjobfile\fR is specified as `-', the job file will be read from
110 The global section contains default parameters for jobs specified in the
111 job file. A job is only affected by global sections residing above it,
112 and there may be any number of global sections. Specific job definitions
113 may override any parameter set in global sections.
116 Some parameters may take arguments of a specific type.
117 Anywhere a numeric value is required, an arithmetic expression may be used,
118 provided it is surrounded by parentheses. Supported operators are:
126 .B multiplication (*)
132 .B exponentiation (^)
136 For time values in expressions, units are microseconds by default. This is
137 different than for time values not in expressions (not enclosed in
138 parentheses). The types used are:
141 String: a sequence of alphanumeric characters.
144 SI integer: a whole number, possibly containing a suffix denoting the base unit
145 of the value. Accepted suffixes are `k', 'M', 'G', 'T', and 'P', denoting
146 kilo (1024), mega (1024^2), giga (1024^3), tera (1024^4), and peta (1024^5)
147 respectively. If prefixed with '0x', the value is assumed to be base 16
148 (hexadecimal). A suffix may include a trailing 'b', for instance 'kb' is
149 identical to 'k'. You can specify a base 10 value by using 'KiB', 'MiB','GiB',
150 etc. This is useful for disk drives where values are often given in base 10
151 values. Specifying '30GiB' will get you 30*1000^3 bytes.
152 When specifying times the default suffix meaning changes, still denoting the
153 base unit of the value, but accepted suffixes are 'D' (days), 'H' (hours), 'M'
154 (minutes), 'S' Seconds, 'ms' (or msec) milli seconds, 'us' (or 'usec') micro
155 seconds. Time values without a unit specify seconds.
156 The suffixes are not case sensitive.
159 Boolean: a true or false value. `0' denotes false, `1' denotes true.
162 Integer range: a range of integers specified in the format
163 \fIlower\fR:\fIupper\fR or \fIlower\fR\-\fIupper\fR. \fIlower\fR and
164 \fIupper\fR may contain a suffix as described above. If an option allows two
165 sets of ranges, they are separated with a `,' or `/' character. For example:
169 List of floating numbers: A list of floating numbers, separated by
174 May be used to override the job name. On the command line, this parameter
175 has the special purpose of signalling the start of a new job.
177 .BI description \fR=\fPstr
178 Human-readable description of the job. It is printed when the job is run, but
179 otherwise has no special purpose.
181 .BI directory \fR=\fPstr
182 Prefix filenames with this directory. Used to place files in a location other
184 You can specify a number of directories by separating the names with a ':'
185 character. These directories will be assigned equally distributed to job clones
186 creates with \fInumjobs\fR as long as they are using generated filenames.
187 If specific \fIfilename(s)\fR are set fio will use the first listed directory,
188 and thereby matching the \fIfilename\fR semantic which generates a file each
189 clone if not specified, but let all clones use the same if set. See
190 \fIfilename\fR for considerations regarding escaping certain characters on
193 .BI filename \fR=\fPstr
195 normally makes up a file name based on the job name, thread number, and file
196 number. If you want to share files between threads in a job or several jobs,
197 specify a \fIfilename\fR for each of them to override the default.
198 If the I/O engine is file-based, you can specify
199 a number of files by separating the names with a `:' character. `\-' is a
200 reserved name, meaning stdin or stdout, depending on the read/write direction
201 set. On Windows, disk devices are accessed as \\.\PhysicalDrive0 for the first
202 device, \\.\PhysicalDrive1 for the second etc. Note: Windows and FreeBSD
203 prevent write access to areas of the disk containing in-use data
204 (e.g. filesystems). If the wanted filename does need to include a colon, then
205 escape that with a '\\' character. For instance, if the filename is
206 "/dev/dsk/foo@3,0:c", then you would use filename="/dev/dsk/foo@3,0\\:c".
208 .BI filename_format \fR=\fPstr
209 If sharing multiple files between jobs, it is usually necessary to have
210 fio generate the exact names that you want. By default, fio will name a file
211 based on the default file format specification of
212 \fBjobname.jobnumber.filenumber\fP. With this option, that can be
213 customized. Fio will recognize and replace the following keywords in this
219 The name of the worker thread or process.
222 The incremental number of the worker thread or process.
225 The incremental number of the file for that worker thread or process.
228 To have dependent jobs share a set of files, this option can be set to
229 have fio generate filenames that are shared between the two. For instance,
230 if \fBtestfiles.$filenum\fR is specified, file number 4 for any job will
231 be named \fBtestfiles.4\fR. The default of \fB$jobname.$jobnum.$filenum\fR
232 will be used if no other format specifier is given.
236 .BI lockfile \fR=\fPstr
237 Fio defaults to not locking any files before it does IO to them. If a file or
238 file descriptor is shared, fio can serialize IO to that file to make the end
239 result consistent. This is usual for emulating real workloads that share files.
245 No locking. This is the default.
248 Only one thread or process may do IO at a time, excluding all others.
251 Read-write locking on the file. Many readers may access the file at the same
252 time, but writes get exclusive access.
256 .BI opendir \fR=\fPstr
257 Recursively open any files below directory \fIstr\fR.
259 .BI readwrite \fR=\fPstr "\fR,\fP rw" \fR=\fPstr
260 Type of I/O pattern. Accepted values are:
271 Sequential trim (Linux block devices only).
280 Random trim (Linux block devices only).
283 Mixed sequential reads and writes.
286 Mixed random reads and writes.
289 Trim and write mixed workload. Blocks will be trimmed first, then the same
290 blocks will be written to.
293 For mixed I/O, the default split is 50/50. For certain types of io the result
294 may still be skewed a bit, since the speed may be different. It is possible to
295 specify a number of IO's to do before getting a new offset, this is done by
296 appending a `:\fI<nr>\fR to the end of the string given. For a random read, it
297 would look like \fBrw=randread:8\fR for passing in an offset modifier with a
298 value of 8. If the postfix is used with a sequential IO pattern, then the value
299 specified will be added to the generated offset for each IO. For instance,
300 using \fBrw=write:4k\fR will skip 4k for every write. It turns sequential IO
301 into sequential IO with holes. See the \fBrw_sequencer\fR option.
304 .BI rw_sequencer \fR=\fPstr
305 If an offset modifier is given by appending a number to the \fBrw=<str>\fR line,
306 then this option controls how that number modifies the IO offset being
307 generated. Accepted values are:
312 Generate sequential offset
315 Generate the same offset
318 \fBsequential\fR is only useful for random IO, where fio would normally
319 generate a new random offset for every IO. If you append eg 8 to randread, you
320 would get a new random offset for every 8 IO's. The result would be a seek for
321 only every 8 IO's, instead of for every IO. Use \fBrw=randread:8\fR to specify
322 that. As sequential IO is already sequential, setting \fBsequential\fR for that
323 would not result in any differences. \fBidentical\fR behaves in a similar
324 fashion, except it sends the same offset 8 number of times before generating a
329 .BI kb_base \fR=\fPint
330 The base unit for a kilobyte. The defacto base is 2^10, 1024. Storage
331 manufacturers like to use 10^3 or 1000 as a base ten unit instead, for obvious
332 reasons. Allowed values are 1024 or 1000, with 1024 being the default.
334 .BI unified_rw_reporting \fR=\fPbool
335 Fio normally reports statistics on a per data direction basis, meaning that
336 read, write, and trim are accounted and reported separately. If this option is
337 set fio sums the results and reports them as "mixed" instead.
339 .BI randrepeat \fR=\fPbool
340 Seed the random number generator used for random I/O patterns in a predictable
341 way so the pattern is repeatable across runs. Default: true.
343 .BI allrandrepeat \fR=\fPbool
344 Seed all random number generators in a predictable way so results are
345 repeatable across runs. Default: false.
347 .BI randseed \fR=\fPint
348 Seed the random number generators based on this seed value, to be able to
349 control what sequence of output is being generated. If not set, the random
350 sequence depends on the \fBrandrepeat\fR setting.
352 .BI fallocate \fR=\fPstr
353 Whether pre-allocation is performed when laying down files. Accepted values
359 Do not pre-allocate space.
362 Pre-allocate via \fBposix_fallocate\fR\|(3).
365 Pre-allocate via \fBfallocate\fR\|(2) with FALLOC_FL_KEEP_SIZE set.
368 Backward-compatible alias for 'none'.
371 Backward-compatible alias for 'posix'.
374 May not be available on all supported platforms. 'keep' is only
375 available on Linux. If using ZFS on Solaris this must be set to 'none'
376 because ZFS doesn't support it. Default: 'posix'.
379 .BI fadvise_hint \fR=\fPbool
380 Use \fBposix_fadvise\fR\|(2) to advise the kernel what I/O patterns
381 are likely to be issued. Default: true.
383 .BI fadvise_stream \fR=\fPint
384 Use \fBposix_fadvise\fR\|(2) to advise the kernel what stream ID the
385 writes issued belong to. Only supported on Linux. Note, this option
386 may change going forward.
389 Total size of I/O for this job. \fBfio\fR will run until this many bytes have
390 been transferred, unless limited by other options (\fBruntime\fR, for instance,
391 or increased/descreased by \fBio_size\fR). Unless \fBnrfiles\fR and
392 \fBfilesize\fR options are given, this amount will be divided between the
393 available files for the job. If not set, fio will use the full size of the
394 given files or devices. If the files do not exist, size must be given. It is
395 also possible to give size as a percentage between 1 and 100. If size=20% is
396 given, fio will use 20% of the full size of the given files or devices.
398 .BI io_size \fR=\fPint "\fR,\fB io_limit \fR=\fPint
399 Normally fio operates within the region set by \fBsize\fR, which means that
400 the \fBsize\fR option sets both the region and size of IO to be performed.
401 Sometimes that is not what you want. With this option, it is possible to
402 define just the amount of IO that fio should do. For instance, if \fBsize\fR
403 is set to 20G and \fBio_limit\fR is set to 5G, fio will perform IO within
404 the first 20G but exit when 5G have been done. The opposite is also
405 possible - if \fBsize\fR is set to 20G, and \fBio_size\fR is set to 40G, then
406 fio will do 40G of IO within the 0..20G region.
408 .BI fill_device \fR=\fPbool "\fR,\fB fill_fs" \fR=\fPbool
409 Sets size to something really large and waits for ENOSPC (no space left on
410 device) as the terminating condition. Only makes sense with sequential write.
411 For a read workload, the mount point will be filled first then IO started on
412 the result. This option doesn't make sense if operating on a raw device node,
413 since the size of that is already known by the file system. Additionally,
414 writing beyond end-of-device will not return ENOSPC there.
416 .BI filesize \fR=\fPirange
417 Individual file sizes. May be a range, in which case \fBfio\fR will select sizes
418 for files at random within the given range, limited to \fBsize\fR in total (if
419 that is given). If \fBfilesize\fR is not specified, each created file is the
422 .BI file_append \fR=\fPbool
423 Perform IO after the end of the file. Normally fio will operate within the
424 size of a file. If this option is set, then fio will append to the file
425 instead. This has identical behavior to setting \fRoffset\fP to the size
426 of a file. This option is ignored on non-regular files.
428 .BI blocksize \fR=\fPint[,int] "\fR,\fB bs" \fR=\fPint[,int]
429 Block size for I/O units. Default: 4k. Values for reads, writes, and trims
430 can be specified separately in the format \fIread\fR,\fIwrite\fR,\fItrim\fR
431 either of which may be empty to leave that value at its default. If a trailing
432 comma isn't given, the remainder will inherit the last value set.
434 .BI blocksize_range \fR=\fPirange[,irange] "\fR,\fB bsrange" \fR=\fPirange[,irange]
435 Specify a range of I/O block sizes. The issued I/O unit will always be a
436 multiple of the minimum size, unless \fBblocksize_unaligned\fR is set. Applies
437 to both reads and writes if only one range is given, but can be specified
438 separately with a comma separating the values. Example: bsrange=1k-4k,2k-8k.
439 Also (see \fBblocksize\fR).
441 .BI bssplit \fR=\fPstr
442 This option allows even finer grained control of the block sizes issued,
443 not just even splits between them. With this option, you can weight various
444 block sizes for exact control of the issued IO for a job that has mixed
445 block sizes. The format of the option is bssplit=blocksize/percentage,
446 optionally adding as many definitions as needed separated by a colon.
447 Example: bssplit=4k/10:64k/50:32k/40 would issue 50% 64k blocks, 10% 4k
448 blocks and 40% 32k blocks. \fBbssplit\fR also supports giving separate
449 splits to reads and writes. The format is identical to what the
450 \fBbs\fR option accepts, the read and write parts are separated with a
453 .B blocksize_unaligned\fR,\fP bs_unaligned
454 If set, any size in \fBblocksize_range\fR may be used. This typically won't
455 work with direct I/O, as that normally requires sector alignment.
457 .BI blockalign \fR=\fPint[,int] "\fR,\fB ba" \fR=\fPint[,int]
458 At what boundary to align random IO offsets. Defaults to the same as 'blocksize'
459 the minimum blocksize given. Minimum alignment is typically 512b
460 for using direct IO, though it usually depends on the hardware block size.
461 This option is mutually exclusive with using a random map for files, so it
462 will turn off that option.
464 .BI bs_is_seq_rand \fR=\fPbool
465 If this option is set, fio will use the normal read,write blocksize settings as
466 sequential,random instead. Any random read or write will use the WRITE
467 blocksize settings, and any sequential read or write will use the READ
471 Initialize buffers with all zeros. Default: fill buffers with random data.
474 If this option is given, fio will refill the IO buffers on every submit. The
475 default is to only fill it at init time and reuse that data. Only makes sense
476 if zero_buffers isn't specified, naturally. If data verification is enabled,
477 refill_buffers is also automatically enabled.
479 .BI scramble_buffers \fR=\fPbool
480 If \fBrefill_buffers\fR is too costly and the target is using data
481 deduplication, then setting this option will slightly modify the IO buffer
482 contents to defeat normal de-dupe attempts. This is not enough to defeat
483 more clever block compression attempts, but it will stop naive dedupe
484 of blocks. Default: true.
486 .BI buffer_compress_percentage \fR=\fPint
487 If this is set, then fio will attempt to provide IO buffer content (on WRITEs)
488 that compress to the specified level. Fio does this by providing a mix of
489 random data and a fixed pattern. The fixed pattern is either zeroes, or the
490 pattern specified by \fBbuffer_pattern\fR. If the pattern option is used, it
491 might skew the compression ratio slightly. Note that this is per block size
492 unit, for file/disk wide compression level that matches this setting. Note
493 that this is per block size unit, for file/disk wide compression level that
494 matches this setting, you'll also want to set refill_buffers.
496 .BI buffer_compress_chunk \fR=\fPint
497 See \fBbuffer_compress_percentage\fR. This setting allows fio to manage how
498 big the ranges of random data and zeroed data is. Without this set, fio will
499 provide \fBbuffer_compress_percentage\fR of blocksize random data, followed by
500 the remaining zeroed. With this set to some chunk size smaller than the block
501 size, fio can alternate random and zeroed data throughout the IO buffer.
503 .BI buffer_pattern \fR=\fPstr
504 If set, fio will fill the IO buffers with this pattern. If not set, the contents
505 of IO buffers is defined by the other options related to buffer contents. The
506 setting can be any pattern of bytes, and can be prefixed with 0x for hex
507 values. It may also be a string, where the string must then be wrapped with
511 \fBbuffer_pattern\fR="abcd"
515 \fBbuffer_pattern\fR=-12
519 \fBbuffer_pattern\fR=0xdeadface
522 Also you can combine everything together in any order:
525 \fBbuffer_pattern\fR=0xdeadface"abcd"-12
529 .BI dedupe_percentage \fR=\fPint
530 If set, fio will generate this percentage of identical buffers when writing.
531 These buffers will be naturally dedupable. The contents of the buffers depend
532 on what other buffer compression settings have been set. It's possible to have
533 the individual buffers either fully compressible, or not at all. This option
534 only controls the distribution of unique buffers.
536 .BI nrfiles \fR=\fPint
537 Number of files to use for this job. Default: 1.
539 .BI openfiles \fR=\fPint
540 Number of files to keep open at the same time. Default: \fBnrfiles\fR.
542 .BI file_service_type \fR=\fPstr
543 Defines how files to service are selected. The following types are defined:
548 Choose a file at random.
551 Round robin over opened files (default).
554 Do each file in the set sequentially.
557 The number of I/Os to issue before switching to a new file can be specified by
558 appending `:\fIint\fR' to the service type.
561 .BI ioengine \fR=\fPstr
562 Defines how the job issues I/O. The following types are defined:
567 Basic \fBread\fR\|(2) or \fBwrite\fR\|(2) I/O. \fBfseek\fR\|(2) is used to
568 position the I/O location.
571 Basic \fBpread\fR\|(2) or \fBpwrite\fR\|(2) I/O.
574 Basic \fBreadv\fR\|(2) or \fBwritev\fR\|(2) I/O. Will emulate queuing by
575 coalescing adjacent IOs into a single submission.
578 Basic \fBpreadv\fR\|(2) or \fBpwritev\fR\|(2) I/O.
581 Linux native asynchronous I/O. This ioengine defines engine specific options.
584 POSIX asynchronous I/O using \fBaio_read\fR\|(3) and \fBaio_write\fR\|(3).
587 Solaris native asynchronous I/O.
590 Windows native asynchronous I/O.
593 File is memory mapped with \fBmmap\fR\|(2) and data copied using
597 \fBsplice\fR\|(2) is used to transfer the data and \fBvmsplice\fR\|(2) to
598 transfer data from user-space to the kernel.
601 Use the syslet system calls to make regular read/write asynchronous.
604 SCSI generic sg v3 I/O. May be either synchronous using the SG_IO ioctl, or if
605 the target is an sg character device, we use \fBread\fR\|(2) and
606 \fBwrite\fR\|(2) for asynchronous I/O.
609 Doesn't transfer any data, just pretends to. Mainly used to exercise \fBfio\fR
610 itself and for debugging and testing purposes.
613 Transfer over the network. The protocol to be used can be defined with the
614 \fBprotocol\fR parameter. Depending on the protocol, \fBfilename\fR,
615 \fBhostname\fR, \fBport\fR, or \fBlisten\fR must be specified.
616 This ioengine defines engine specific options.
619 Like \fBnet\fR, but uses \fBsplice\fR\|(2) and \fBvmsplice\fR\|(2) to map data
620 and send/receive. This ioengine defines engine specific options.
623 Doesn't transfer any data, but burns CPU cycles according to \fBcpuload\fR and
624 \fBcpucycles\fR parameters.
627 The GUASI I/O engine is the Generic Userspace Asynchronous Syscall Interface
628 approach to asynchronous I/O.
630 See <http://www.xmailserver.org/guasi\-lib.html>.
633 The RDMA I/O engine supports both RDMA memory semantics (RDMA_WRITE/RDMA_READ)
634 and channel semantics (Send/Recv) for the InfiniBand, RoCE and iWARP protocols.
637 Loads an external I/O engine object file. Append the engine filename as
641 IO engine that does regular linux native fallocate call to simulate data
642 transfer as fio ioengine
644 DDIR_READ does fallocate(,mode = FALLOC_FL_KEEP_SIZE,)
646 DIR_WRITE does fallocate(,mode = 0)
648 DDIR_TRIM does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE)
651 IO engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate defragment activity
652 request to DDIR_WRITE event
655 IO engine supporting direct access to Ceph Rados Block Devices (RBD) via librbd
656 without the need to use the kernel rbd driver. This ioengine defines engine specific
660 Using Glusterfs libgfapi sync interface to direct access to Glusterfs volumes without
661 having to go through FUSE. This ioengine defines engine specific
665 Using Glusterfs libgfapi async interface to direct access to Glusterfs volumes without
666 having to go through FUSE. This ioengine defines engine specific
670 Read and write through Hadoop (HDFS). The \fBfilename\fR option is used to
671 specify host,port of the hdfs name-node to connect. This engine interprets
672 offsets a little differently. In HDFS, files once created cannot be modified.
673 So random writes are not possible. To imitate this, libhdfs engine expects
674 bunch of small files to be created over HDFS, and engine will randomly pick a
675 file out of those files based on the offset generated by fio backend. (see the
676 example job file to create such files, use rw=write option). Please note, you
677 might want to set necessary environment variables to work with hdfs/libhdfs
681 Read, write and erase an MTD character device (e.g., /dev/mtd0). Discards are
682 treated as erases. Depending on the underlying device type, the I/O may have
683 to go in a certain pattern, e.g., on NAND, writing sequentially to erase blocks
684 and discarding before overwriting. The writetrim mode works well for this
690 .BI iodepth \fR=\fPint
691 Number of I/O units to keep in flight against the file. Note that increasing
692 iodepth beyond 1 will not affect synchronous ioengines (except for small
693 degress when verify_async is in use). Even async engines may impose OS
694 restrictions causing the desired depth not to be achieved. This may happen on
695 Linux when using libaio and not setting \fBdirect\fR=1, since buffered IO is
696 not async on that OS. Keep an eye on the IO depth distribution in the
697 fio output to verify that the achieved depth is as expected. Default: 1.
699 .BI iodepth_batch \fR=\fPint "\fR,\fP iodepth_batch_submit" \fR=\fPint
700 This defines how many pieces of IO to submit at once. It defaults to 1
701 which means that we submit each IO as soon as it is available, but can
702 be raised to submit bigger batches of IO at the time. If it is set to 0
703 the \fBiodepth\fR value will be used.
705 .BI iodepth_batch_complete_min \fR=\fPint "\fR,\fP iodepth_batch_complete" \fR=\fPint
706 This defines how many pieces of IO to retrieve at once. It defaults to 1 which
707 means that we'll ask for a minimum of 1 IO in the retrieval process from the
708 kernel. The IO retrieval will go on until we hit the limit set by
709 \fBiodepth_low\fR. If this variable is set to 0, then fio will always check for
710 completed events before queuing more IO. This helps reduce IO latency, at the
711 cost of more retrieval system calls.
713 .BI iodepth_batch_complete_max \fR=\fPint
714 This defines maximum pieces of IO to
715 retrieve at once. This variable should be used along with
716 \fBiodepth_batch_complete_min\fR=int variable, specifying the range
717 of min and max amount of IO which should be retrieved. By default
718 it is equal to \fBiodepth_batch_complete_min\fR value.
723 \fBiodepth_batch_complete_min\fR=1
725 \fBiodepth_batch_complete_max\fR=<iodepth>
728 which means that we will retrieve at leat 1 IO and up to the
729 whole submitted queue depth. If none of IO has been completed
734 \fBiodepth_batch_complete_min\fR=0
736 \fBiodepth_batch_complete_max\fR=<iodepth>
739 which means that we can retrieve up to the whole submitted
740 queue depth, but if none of IO has been completed yet, we will
741 NOT wait and immediately exit the system call. In this example
742 we simply do polling.
745 .BI iodepth_low \fR=\fPint
746 Low watermark indicating when to start filling the queue again. Default:
749 .BI io_submit_mode \fR=\fPstr
750 This option controls how fio submits the IO to the IO engine. The default is
751 \fBinline\fR, which means that the fio job threads submit and reap IO directly.
752 If set to \fBoffload\fR, the job threads will offload IO submission to a
753 dedicated pool of IO threads. This requires some coordination and thus has a
754 bit of extra overhead, especially for lower queue depth IO where it can
755 increase latencies. The benefit is that fio can manage submission rates
756 independently of the device completion rates. This avoids skewed latency
757 reporting if IO gets back up on the device side (the coordinated omission
760 .BI direct \fR=\fPbool
761 If true, use non-buffered I/O (usually O_DIRECT). Default: false.
763 .BI atomic \fR=\fPbool
764 If value is true, attempt to use atomic direct IO. Atomic writes are guaranteed
765 to be stable once acknowledged by the operating system. Only Linux supports
768 .BI buffered \fR=\fPbool
769 If true, use buffered I/O. This is the opposite of the \fBdirect\fR parameter.
772 .BI offset \fR=\fPint
773 Offset in the file to start I/O. Data before the offset will not be touched.
775 .BI offset_increment \fR=\fPint
776 If this is provided, then the real offset becomes the
777 offset + offset_increment * thread_number, where the thread number is a
778 counter that starts at 0 and is incremented for each sub-job (i.e. when
779 numjobs option is specified). This option is useful if there are several jobs
780 which are intended to operate on a file in parallel disjoint segments, with
781 even spacing between the starting points.
783 .BI number_ios \fR=\fPint
784 Fio will normally perform IOs until it has exhausted the size of the region
785 set by \fBsize\fR, or if it exhaust the allocated time (or hits an error
786 condition). With this setting, the range/size can be set independently of
787 the number of IOs to perform. When fio reaches this number, it will exit
788 normally and report status. Note that this does not extend the amount
789 of IO that will be done, it will only stop fio if this condition is met
790 before other end-of-job criteria.
793 How many I/Os to perform before issuing an \fBfsync\fR\|(2) of dirty data. If
794 0, don't sync. Default: 0.
796 .BI fdatasync \fR=\fPint
797 Like \fBfsync\fR, but uses \fBfdatasync\fR\|(2) instead to only sync the
798 data parts of the file. Default: 0.
800 .BI write_barrier \fR=\fPint
801 Make every Nth write a barrier write.
803 .BI sync_file_range \fR=\fPstr:int
804 Use \fBsync_file_range\fR\|(2) for every \fRval\fP number of write operations. Fio will
805 track range of writes that have happened since the last \fBsync_file_range\fR\|(2) call.
806 \fRstr\fP can currently be one or more of:
810 SYNC_FILE_RANGE_WAIT_BEFORE
813 SYNC_FILE_RANGE_WRITE
816 SYNC_FILE_RANGE_WRITE
820 So if you do sync_file_range=wait_before,write:8, fio would use
821 \fBSYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE\fP for every 8 writes.
822 Also see the \fBsync_file_range\fR\|(2) man page. This option is Linux specific.
824 .BI overwrite \fR=\fPbool
825 If writing, setup the file first and do overwrites. Default: false.
827 .BI end_fsync \fR=\fPbool
828 Sync file contents when a write stage has completed. Default: false.
830 .BI fsync_on_close \fR=\fPbool
831 If true, sync file contents on close. This differs from \fBend_fsync\fR in that
832 it will happen on every close, not just at the end of the job. Default: false.
834 .BI rwmixread \fR=\fPint
835 Percentage of a mixed workload that should be reads. Default: 50.
837 .BI rwmixwrite \fR=\fPint
838 Percentage of a mixed workload that should be writes. If \fBrwmixread\fR and
839 \fBrwmixwrite\fR are given and do not sum to 100%, the latter of the two
840 overrides the first. This may interfere with a given rate setting, if fio is
841 asked to limit reads or writes to a certain rate. If that is the case, then
842 the distribution may be skewed. Default: 50.
844 .BI random_distribution \fR=\fPstr:float
845 By default, fio will use a completely uniform random distribution when asked
846 to perform random IO. Sometimes it is useful to skew the distribution in
847 specific ways, ensuring that some parts of the data is more hot than others.
848 Fio includes the following distribution models:
852 Uniform random distribution
862 When using a zipf or pareto distribution, an input value is also needed to
863 define the access pattern. For zipf, this is the zipf theta. For pareto,
864 it's the pareto power. Fio includes a test program, genzipf, that can be
865 used visualize what the given input values will yield in terms of hit rates.
866 If you wanted to use zipf with a theta of 1.2, you would use
867 random_distribution=zipf:1.2 as the option. If a non-uniform model is used,
868 fio will disable use of the random map.
870 .BI percentage_random \fR=\fPint
871 For a random workload, set how big a percentage should be random. This defaults
872 to 100%, in which case the workload is fully random. It can be set from
873 anywhere from 0 to 100. Setting it to 0 would make the workload fully
874 sequential. It is possible to set different values for reads, writes, and
875 trim. To do so, simply use a comma separated list. See \fBblocksize\fR.
878 Normally \fBfio\fR will cover every block of the file when doing random I/O. If
879 this parameter is given, a new offset will be chosen without looking at past
880 I/O history. This parameter is mutually exclusive with \fBverify\fR.
882 .BI softrandommap \fR=\fPbool
883 See \fBnorandommap\fR. If fio runs with the random block map enabled and it
884 fails to allocate the map, if this option is set it will continue without a
885 random block map. As coverage will not be as complete as with random maps, this
886 option is disabled by default.
888 .BI random_generator \fR=\fPstr
889 Fio supports the following engines for generating IO offsets for random IO:
893 Strong 2^88 cycle random number generator
896 Linear feedback shift register generator
899 Strong 64-bit 2^258 cycle random number generator
903 Tausworthe is a strong random number generator, but it requires tracking on the
904 side if we want to ensure that blocks are only read or written once. LFSR
905 guarantees that we never generate the same offset twice, and it's also less
906 computationally expensive. It's not a true random generator, however, though
907 for IO purposes it's typically good enough. LFSR only works with single block
908 sizes, not with workloads that use multiple block sizes. If used with such a
909 workload, fio may read or write some blocks multiple times.
912 Run job with given nice value. See \fBnice\fR\|(2).
915 Set I/O priority value of this job between 0 (highest) and 7 (lowest). See
918 .BI prioclass \fR=\fPint
919 Set I/O priority class. See \fBionice\fR\|(1).
921 .BI thinktime \fR=\fPint
922 Stall job for given number of microseconds between issuing I/Os.
924 .BI thinktime_spin \fR=\fPint
925 Pretend to spend CPU time for given number of microseconds, sleeping the rest
926 of the time specified by \fBthinktime\fR. Only valid if \fBthinktime\fR is set.
928 .BI thinktime_blocks \fR=\fPint
929 Only valid if thinktime is set - control how many blocks to issue, before
930 waiting \fBthinktime\fR microseconds. If not set, defaults to 1 which will
931 make fio wait \fBthinktime\fR microseconds after every block. This
932 effectively makes any queue depth setting redundant, since no more than 1 IO
933 will be queued before we have to complete it and do our thinktime. In other
934 words, this setting effectively caps the queue depth if the latter is larger.
938 Cap bandwidth used by this job. The number is in bytes/sec, the normal postfix
939 rules apply. You can use \fBrate\fR=500k to limit reads and writes to 500k each,
940 or you can specify read and writes separately. Using \fBrate\fR=1m,500k would
941 limit reads to 1MB/sec and writes to 500KB/sec. Capping only reads or writes
942 can be done with \fBrate\fR=,500k or \fBrate\fR=500k,. The former will only
943 limit writes (to 500KB/sec), the latter will only limit reads.
945 .BI ratemin \fR=\fPint
946 Tell \fBfio\fR to do whatever it can to maintain at least the given bandwidth.
947 Failing to meet this requirement will cause the job to exit. The same format
948 as \fBrate\fR is used for read vs write separation.
950 .BI rate_iops \fR=\fPint
951 Cap the bandwidth to this number of IOPS. Basically the same as rate, just
952 specified independently of bandwidth. The same format as \fBrate\fR is used for
953 read vs write separation. If \fBblocksize\fR is a range, the smallest block
954 size is used as the metric.
956 .BI rate_iops_min \fR=\fPint
957 If this rate of I/O is not met, the job will exit. The same format as \fBrate\fR
958 is used for read vs write separation.
960 .BI ratecycle \fR=\fPint
961 Average bandwidth for \fBrate\fR and \fBratemin\fR over this number of
962 milliseconds. Default: 1000ms.
964 .BI latency_target \fR=\fPint
965 If set, fio will attempt to find the max performance point that the given
966 workload will run at while maintaining a latency below this target. The
967 values is given in microseconds. See \fBlatency_window\fR and
968 \fBlatency_percentile\fR.
970 .BI latency_window \fR=\fPint
971 Used with \fBlatency_target\fR to specify the sample window that the job
972 is run at varying queue depths to test the performance. The value is given
975 .BI latency_percentile \fR=\fPfloat
976 The percentage of IOs that must fall within the criteria specified by
977 \fBlatency_target\fR and \fBlatency_window\fR. If not set, this defaults
978 to 100.0, meaning that all IOs must be equal or below to the value set
979 by \fBlatency_target\fR.
981 .BI max_latency \fR=\fPint
982 If set, fio will exit the job if it exceeds this maximum latency. It will exit
985 .BI cpumask \fR=\fPint
986 Set CPU affinity for this job. \fIint\fR is a bitmask of allowed CPUs the job
987 may run on. See \fBsched_setaffinity\fR\|(2).
989 .BI cpus_allowed \fR=\fPstr
990 Same as \fBcpumask\fR, but allows a comma-delimited list of CPU numbers.
992 .BI cpus_allowed_policy \fR=\fPstr
993 Set the policy of how fio distributes the CPUs specified by \fBcpus_allowed\fR
994 or \fBcpumask\fR. Two policies are supported:
999 All jobs will share the CPU set specified.
1002 Each job will get a unique CPU from the CPU set.
1005 \fBshared\fR is the default behaviour, if the option isn't specified. If
1006 \fBsplit\fR is specified, then fio will assign one cpu per job. If not enough
1007 CPUs are given for the jobs listed, then fio will roundrobin the CPUs in
1012 .BI numa_cpu_nodes \fR=\fPstr
1013 Set this job running on specified NUMA nodes' CPUs. The arguments allow
1014 comma delimited list of cpu numbers, A-B ranges, or 'all'.
1016 .BI numa_mem_policy \fR=\fPstr
1017 Set this job's memory policy and corresponding NUMA nodes. Format of
1021 .B <mode>[:<nodelist>]
1024 is one of the following memory policy:
1026 .B default, prefer, bind, interleave, local
1029 For \fBdefault\fR and \fBlocal\fR memory policy, no \fBnodelist\fR is
1030 needed to be specified. For \fBprefer\fR, only one node is
1031 allowed. For \fBbind\fR and \fBinterleave\fR, \fBnodelist\fR allows
1032 comma delimited list of numbers, A-B ranges, or 'all'.
1034 .BI startdelay \fR=\fPirange
1035 Delay start of job for the specified number of seconds. Supports all time
1036 suffixes to allow specification of hours, minutes, seconds and
1037 milliseconds - seconds are the default if a unit is omitted.
1038 Can be given as a range which causes each thread to choose randomly out of the
1041 .BI runtime \fR=\fPint
1042 Terminate processing after the specified number of seconds.
1045 If given, run for the specified \fBruntime\fR duration even if the files are
1046 completely read or written. The same workload will be repeated as many times
1047 as \fBruntime\fR allows.
1049 .BI ramp_time \fR=\fPint
1050 If set, fio will run the specified workload for this amount of time before
1051 logging any performance numbers. Useful for letting performance settle before
1052 logging results, thus minimizing the runtime required for stable results. Note
1053 that the \fBramp_time\fR is considered lead in time for a job, thus it will
1054 increase the total runtime if a special timeout or runtime is specified.
1056 .BI invalidate \fR=\fPbool
1057 Invalidate buffer-cache for the file prior to starting I/O. Default: true.
1059 .BI sync \fR=\fPbool
1060 Use synchronous I/O for buffered writes. For the majority of I/O engines,
1061 this means using O_SYNC. Default: false.
1063 .BI iomem \fR=\fPstr "\fR,\fP mem" \fR=\fPstr
1064 Allocation method for I/O unit buffer. Allowed values are:
1069 Allocate memory with \fBmalloc\fR\|(3).
1072 Use shared memory buffers allocated through \fBshmget\fR\|(2).
1075 Same as \fBshm\fR, but use huge pages as backing.
1078 Use \fBmmap\fR\|(2) for allocation. Uses anonymous memory unless a filename
1079 is given after the option in the format `:\fIfile\fR'.
1082 Same as \fBmmap\fR, but use huge files as backing.
1085 The amount of memory allocated is the maximum allowed \fBblocksize\fR for the
1086 job multiplied by \fBiodepth\fR. For \fBshmhuge\fR or \fBmmaphuge\fR to work,
1087 the system must have free huge pages allocated. \fBmmaphuge\fR also needs to
1088 have hugetlbfs mounted, and \fIfile\fR must point there. At least on Linux,
1089 huge pages must be manually allocated. See \fB/proc/sys/vm/nr_hugehages\fR
1090 and the documentation for that. Normally you just need to echo an appropriate
1091 number, eg echoing 8 will ensure that the OS has 8 huge pages ready for
1095 .BI iomem_align \fR=\fPint "\fR,\fP mem_align" \fR=\fPint
1096 This indicates the memory alignment of the IO memory buffers. Note that the
1097 given alignment is applied to the first IO unit buffer, if using \fBiodepth\fR
1098 the alignment of the following buffers are given by the \fBbs\fR used. In
1099 other words, if using a \fBbs\fR that is a multiple of the page sized in the
1100 system, all buffers will be aligned to this value. If using a \fBbs\fR that
1101 is not page aligned, the alignment of subsequent IO memory buffers is the
1102 sum of the \fBiomem_align\fR and \fBbs\fR used.
1104 .BI hugepage\-size \fR=\fPint
1105 Defines the size of a huge page. Must be at least equal to the system setting.
1106 Should be a multiple of 1MB. Default: 4MB.
1109 Terminate all jobs when one finishes. Default: wait for each job to finish.
1111 .BI bwavgtime \fR=\fPint
1112 Average bandwidth calculations over the given time in milliseconds. Default:
1115 .BI iopsavgtime \fR=\fPint
1116 Average IOPS calculations over the given time in milliseconds. Default:
1119 .BI create_serialize \fR=\fPbool
1120 If true, serialize file creation for the jobs. Default: true.
1122 .BI create_fsync \fR=\fPbool
1123 \fBfsync\fR\|(2) data file after creation. Default: true.
1125 .BI create_on_open \fR=\fPbool
1126 If true, the files are not created until they are opened for IO by the job.
1128 .BI create_only \fR=\fPbool
1129 If true, fio will only run the setup phase of the job. If files need to be
1130 laid out or updated on disk, only that will be done. The actual job contents
1133 .BI allow_file_create \fR=\fPbool
1134 If true, fio is permitted to create files as part of its workload. This is
1135 the default behavior. If this option is false, then fio will error out if the
1136 files it needs to use don't already exist. Default: true.
1138 .BI allow_mounted_write \fR=\fPbool
1139 If this isn't set, fio will abort jobs that are destructive (eg that write)
1140 to what appears to be a mounted device or partition. This should help catch
1141 creating inadvertently destructive tests, not realizing that the test will
1142 destroy data on the mounted file system. Default: false.
1144 .BI pre_read \fR=\fPbool
1145 If this is given, files will be pre-read into memory before starting the given
1146 IO operation. This will also clear the \fR \fBinvalidate\fR flag, since it is
1147 pointless to pre-read and then drop the cache. This will only work for IO
1148 engines that are seekable, since they allow you to read the same data
1149 multiple times. Thus it will not work on eg network or splice IO.
1151 .BI unlink \fR=\fPbool
1152 Unlink job files when done. Default: false.
1154 .BI loops \fR=\fPint
1155 Specifies the number of iterations (runs of the same workload) of this job.
1158 .BI verify_only \fR=\fPbool
1159 Do not perform the specified workload, only verify data still matches previous
1160 invocation of this workload. This option allows one to check data multiple
1161 times at a later date without overwriting it. This option makes sense only for
1162 workloads that write data, and does not support workloads with the
1163 \fBtime_based\fR option set.
1165 .BI do_verify \fR=\fPbool
1166 Run the verify phase after a write phase. Only valid if \fBverify\fR is set.
1169 .BI verify \fR=\fPstr
1170 Method of verifying file contents after each iteration of the job. Each
1171 verification method also implies verification of special header, which is
1172 written to the beginning of each block. This header also includes meta
1173 information, like offset of the block, block number, timestamp when block
1174 was written, etc. \fBverify\fR=str can be combined with \fBverify_pattern\fR=str
1175 option. The allowed values are:
1179 .B md5 crc16 crc32 crc32c crc32c-intel crc64 crc7 sha256 sha512 sha1 xxhash
1180 Store appropriate checksum in the header of each block. crc32c-intel is
1181 hardware accelerated SSE4.2 driven, falls back to regular crc32c if
1182 not supported by the system.
1185 This option is deprecated, since now meta information is included in generic
1186 verification header and meta verification happens by default. For detailed
1187 information see the description of the \fBverify\fR=str setting. This option
1188 is kept because of compatibility's sake with old configurations. Do not use it.
1191 Verify a strict pattern. Normally fio includes a header with some basic
1192 information and checksumming, but if this option is set, only the
1193 specific pattern set with \fBverify_pattern\fR is verified.
1196 Pretend to verify. Used for testing internals.
1199 This option can be used for repeated burn-in tests of a system to make sure
1200 that the written data is also correctly read back. If the data direction given
1201 is a read or random read, fio will assume that it should verify a previously
1202 written file. If the data direction includes any form of write, the verify will
1203 be of the newly written data.
1206 .BI verifysort \fR=\fPbool
1207 If true, written verify blocks are sorted if \fBfio\fR deems it to be faster to
1208 read them back in a sorted manner. Default: true.
1210 .BI verifysort_nr \fR=\fPint
1211 Pre-load and sort verify blocks for a read workload.
1213 .BI verify_offset \fR=\fPint
1214 Swap the verification header with data somewhere else in the block before
1215 writing. It is swapped back before verifying.
1217 .BI verify_interval \fR=\fPint
1218 Write the verification header for this number of bytes, which should divide
1219 \fBblocksize\fR. Default: \fBblocksize\fR.
1221 .BI verify_pattern \fR=\fPstr
1222 If set, fio will fill the io buffers with this pattern. Fio defaults to filling
1223 with totally random bytes, but sometimes it's interesting to fill with a known
1224 pattern for io verification purposes. Depending on the width of the pattern,
1225 fio will fill 1/2/3/4 bytes of the buffer at the time(it can be either a
1226 decimal or a hex number). The verify_pattern if larger than a 32-bit quantity
1227 has to be a hex number that starts with either "0x" or "0X". Use with
1228 \fBverify\fP=str. Also, verify_pattern supports %o format, which means that for
1229 each block offset will be written and then verifyied back, e.g.:
1232 \fBverify_pattern\fR=%o
1234 Or use combination of everything:
1237 \fBverify_pattern\fR=0xff%o"abcd"-21
1241 .BI verify_fatal \fR=\fPbool
1242 If true, exit the job on the first observed verification failure. Default:
1245 .BI verify_dump \fR=\fPbool
1246 If set, dump the contents of both the original data block and the data block we
1247 read off disk to files. This allows later analysis to inspect just what kind of
1248 data corruption occurred. Off by default.
1250 .BI verify_async \fR=\fPint
1251 Fio will normally verify IO inline from the submitting thread. This option
1252 takes an integer describing how many async offload threads to create for IO
1253 verification instead, causing fio to offload the duty of verifying IO contents
1254 to one or more separate threads. If using this offload option, even sync IO
1255 engines can benefit from using an \fBiodepth\fR setting higher than 1, as it
1256 allows them to have IO in flight while verifies are running.
1258 .BI verify_async_cpus \fR=\fPstr
1259 Tell fio to set the given CPU affinity on the async IO verification threads.
1260 See \fBcpus_allowed\fP for the format used.
1262 .BI verify_backlog \fR=\fPint
1263 Fio will normally verify the written contents of a job that utilizes verify
1264 once that job has completed. In other words, everything is written then
1265 everything is read back and verified. You may want to verify continually
1266 instead for a variety of reasons. Fio stores the meta data associated with an
1267 IO block in memory, so for large verify workloads, quite a bit of memory would
1268 be used up holding this meta data. If this option is enabled, fio will write
1269 only N blocks before verifying these blocks.
1271 .BI verify_backlog_batch \fR=\fPint
1272 Control how many blocks fio will verify if verify_backlog is set. If not set,
1273 will default to the value of \fBverify_backlog\fR (meaning the entire queue is
1274 read back and verified). If \fBverify_backlog_batch\fR is less than
1275 \fBverify_backlog\fR then not all blocks will be verified, if
1276 \fBverify_backlog_batch\fR is larger than \fBverify_backlog\fR, some blocks
1277 will be verified more than once.
1279 .BI trim_percentage \fR=\fPint
1280 Number of verify blocks to discard/trim.
1282 .BI trim_verify_zero \fR=\fPbool
1283 Verify that trim/discarded blocks are returned as zeroes.
1285 .BI trim_backlog \fR=\fPint
1286 Trim after this number of blocks are written.
1288 .BI trim_backlog_batch \fR=\fPint
1289 Trim this number of IO blocks.
1291 .BI experimental_verify \fR=\fPbool
1292 Enable experimental verification.
1294 .BI verify_state_save \fR=\fPbool
1295 When a job exits during the write phase of a verify workload, save its
1296 current state. This allows fio to replay up until that point, if the
1297 verify state is loaded for the verify read phase.
1299 .BI verify_state_load \fR=\fPbool
1300 If a verify termination trigger was used, fio stores the current write
1301 state of each thread. This can be used at verification time so that fio
1302 knows how far it should verify. Without this information, fio will run
1303 a full verification pass, according to the settings in the job file used.
1305 .B stonewall "\fR,\fP wait_for_previous"
1306 Wait for preceding jobs in the job file to exit before starting this one.
1307 \fBstonewall\fR implies \fBnew_group\fR.
1310 Start a new reporting group. If not given, all jobs in a file will be part
1311 of the same reporting group, unless separated by a stonewall.
1313 .BI numjobs \fR=\fPint
1314 Number of clones (processes/threads performing the same workload) of this job.
1318 If set, display per-group reports instead of per-job when \fBnumjobs\fR is
1322 Use threads created with \fBpthread_create\fR\|(3) instead of processes created
1323 with \fBfork\fR\|(2).
1325 .BI zonesize \fR=\fPint
1326 Divide file into zones of the specified size in bytes. See \fBzoneskip\fR.
1328 .BI zonerange \fR=\fPint
1329 Give size of an IO zone. See \fBzoneskip\fR.
1331 .BI zoneskip \fR=\fPint
1332 Skip the specified number of bytes when \fBzonesize\fR bytes of data have been
1335 .BI write_iolog \fR=\fPstr
1336 Write the issued I/O patterns to the specified file. Specify a separate file
1337 for each job, otherwise the iologs will be interspersed and the file may be
1340 .BI read_iolog \fR=\fPstr
1341 Replay the I/O patterns contained in the specified file generated by
1342 \fBwrite_iolog\fR, or may be a \fBblktrace\fR binary file.
1344 .BI replay_no_stall \fR=\fPint
1345 While replaying I/O patterns using \fBread_iolog\fR the default behavior
1346 attempts to respect timing information between I/Os. Enabling
1347 \fBreplay_no_stall\fR causes I/Os to be replayed as fast as possible while
1348 still respecting ordering.
1350 .BI replay_redirect \fR=\fPstr
1351 While replaying I/O patterns using \fBread_iolog\fR the default behavior
1352 is to replay the IOPS onto the major/minor device that each IOP was recorded
1353 from. Setting \fBreplay_redirect\fR causes all IOPS to be replayed onto the
1354 single specified device regardless of the device it was recorded from.
1356 .BI replay_align \fR=\fPint
1357 Force alignment of IO offsets and lengths in a trace to this power of 2 value.
1359 .BI replay_scale \fR=\fPint
1360 Scale sector offsets down by this factor when replaying traces.
1362 .BI per_job_logs \fR=\fPbool
1363 If set, this generates bw/clat/iops log with per file private filenames. If
1364 not set, jobs with identical names will share the log filename. Default: true.
1366 .BI write_bw_log \fR=\fPstr
1367 If given, write a bandwidth log of the jobs in this job file. Can be used to
1368 store data of the bandwidth of the jobs in their lifetime. The included
1369 fio_generate_plots script uses gnuplot to turn these text files into nice
1370 graphs. See \fBwrite_lat_log\fR for behaviour of given filename. For this
1371 option, the postfix is _bw.x.log, where x is the index of the job (1..N,
1372 where N is the number of jobs). If \fBper_job_logs\fR is false, then the
1373 filename will not include the job index.
1375 .BI write_lat_log \fR=\fPstr
1376 Same as \fBwrite_bw_log\fR, but writes I/O completion latencies. If no
1377 filename is given with this option, the default filename of
1378 "jobname_type.x.log" is used, where x is the index of the job (1..N, where
1379 N is the number of jobs). Even if the filename is given, fio will still
1380 append the type of log. If \fBper_job_logs\fR is false, then the filename will
1381 not include the job index.
1383 .BI write_iops_log \fR=\fPstr
1384 Same as \fBwrite_bw_log\fR, but writes IOPS. If no filename is given with this
1385 option, the default filename of "jobname_type.x.log" is used, where x is the
1386 index of the job (1..N, where N is the number of jobs). Even if the filename
1387 is given, fio will still append the type of log. If \fBper_job_logs\fR is false,
1388 then the filename will not include the job index.
1390 .BI log_avg_msec \fR=\fPint
1391 By default, fio will log an entry in the iops, latency, or bw log for every
1392 IO that completes. When writing to the disk log, that can quickly grow to a
1393 very large size. Setting this option makes fio average the each log entry
1394 over the specified period of time, reducing the resolution of the log.
1397 .BI log_offset \fR=\fPbool
1398 If this is set, the iolog options will include the byte offset for the IO
1399 entry as well as the other data values.
1401 .BI log_compression \fR=\fPint
1402 If this is set, fio will compress the IO logs as it goes, to keep the memory
1403 footprint lower. When a log reaches the specified size, that chunk is removed
1404 and compressed in the background. Given that IO logs are fairly highly
1405 compressible, this yields a nice memory savings for longer runs. The downside
1406 is that the compression will consume some background CPU cycles, so it may
1407 impact the run. This, however, is also true if the logging ends up consuming
1408 most of the system memory. So pick your poison. The IO logs are saved
1409 normally at the end of a run, by decompressing the chunks and storing them
1410 in the specified log file. This feature depends on the availability of zlib.
1412 .BI log_store_compressed \fR=\fPbool
1413 If set, and \fBlog\fR_compression is also set, fio will store the log files in
1414 a compressed format. They can be decompressed with fio, using the
1415 \fB\-\-inflate-log\fR command line parameter. The files will be stored with a
1418 .BI block_error_percentiles \fR=\fPbool
1419 If set, record errors in trim block-sized units from writes and trims and output
1420 a histogram of how many trims it took to get to errors, and what kind of error
1423 .BI disable_lat \fR=\fPbool
1424 Disable measurements of total latency numbers. Useful only for cutting
1425 back the number of calls to \fBgettimeofday\fR\|(2), as that does impact performance at
1426 really high IOPS rates. Note that to really get rid of a large amount of these
1427 calls, this option must be used with disable_slat and disable_bw as well.
1429 .BI disable_clat \fR=\fPbool
1430 Disable measurements of completion latency numbers. See \fBdisable_lat\fR.
1432 .BI disable_slat \fR=\fPbool
1433 Disable measurements of submission latency numbers. See \fBdisable_lat\fR.
1435 .BI disable_bw_measurement \fR=\fPbool
1436 Disable measurements of throughput/bandwidth numbers. See \fBdisable_lat\fR.
1438 .BI lockmem \fR=\fPint
1439 Pin the specified amount of memory with \fBmlock\fR\|(2). Can be used to
1440 simulate a smaller amount of memory. The amount specified is per worker.
1442 .BI exec_prerun \fR=\fPstr
1443 Before running the job, execute the specified command with \fBsystem\fR\|(3).
1445 Output is redirected in a file called \fBjobname.prerun.txt\fR
1448 .BI exec_postrun \fR=\fPstr
1449 Same as \fBexec_prerun\fR, but the command is executed after the job completes.
1451 Output is redirected in a file called \fBjobname.postrun.txt\fR
1454 .BI ioscheduler \fR=\fPstr
1455 Attempt to switch the device hosting the file to the specified I/O scheduler.
1457 .BI disk_util \fR=\fPbool
1458 Generate disk utilization statistics if the platform supports it. Default: true.
1460 .BI clocksource \fR=\fPstr
1461 Use the given clocksource as the base of timing. The supported options are:
1465 \fBgettimeofday\fR\|(2)
1468 \fBclock_gettime\fR\|(2)
1471 Internal CPU clock source
1475 \fBcpu\fR is the preferred clocksource if it is reliable, as it is very fast
1476 (and fio is heavy on time calls). Fio will automatically use this clocksource
1477 if it's supported and considered reliable on the system it is running on,
1478 unless another clocksource is specifically set. For x86/x86-64 CPUs, this
1479 means supporting TSC Invariant.
1481 .BI gtod_reduce \fR=\fPbool
1482 Enable all of the \fBgettimeofday\fR\|(2) reducing options (disable_clat, disable_slat,
1483 disable_bw) plus reduce precision of the timeout somewhat to really shrink the
1484 \fBgettimeofday\fR\|(2) call count. With this option enabled, we only do about 0.4% of
1485 the gtod() calls we would have done if all time keeping was enabled.
1487 .BI gtod_cpu \fR=\fPint
1488 Sometimes it's cheaper to dedicate a single thread of execution to just getting
1489 the current time. Fio (and databases, for instance) are very intensive on
1490 \fBgettimeofday\fR\|(2) calls. With this option, you can set one CPU aside for doing
1491 nothing but logging current time to a shared memory location. Then the other
1492 threads/processes that run IO workloads need only copy that segment, instead of
1493 entering the kernel with a \fBgettimeofday\fR\|(2) call. The CPU set aside for doing
1494 these time calls will be excluded from other uses. Fio will manually clear it
1495 from the CPU mask of other jobs.
1497 .BI ignore_error \fR=\fPstr
1498 Sometimes you want to ignore some errors during test in that case you can specify
1499 error list for each error type.
1501 ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST
1503 errors for given error type is separated with ':'.
1504 Error may be symbol ('ENOSPC', 'ENOMEM') or an integer.
1506 Example: ignore_error=EAGAIN,ENOSPC:122 .
1508 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from WRITE.
1510 .BI error_dump \fR=\fPbool
1511 If set dump every error even if it is non fatal, true by default. If disabled
1512 only fatal error will be dumped
1514 .BI profile \fR=\fPstr
1515 Select a specific builtin performance test.
1517 .BI cgroup \fR=\fPstr
1518 Add job to this control group. If it doesn't exist, it will be created.
1519 The system must have a mounted cgroup blkio mount point for this to work. If
1520 your system doesn't have it mounted, you can do so with:
1522 # mount \-t cgroup \-o blkio none /cgroup
1524 .BI cgroup_weight \fR=\fPint
1525 Set the weight of the cgroup to this value. See the documentation that comes
1526 with the kernel, allowed values are in the range of 100..1000.
1528 .BI cgroup_nodelete \fR=\fPbool
1529 Normally fio will delete the cgroups it has created after the job completion.
1530 To override this behavior and to leave cgroups around after the job completion,
1531 set cgroup_nodelete=1. This can be useful if one wants to inspect various
1532 cgroup files after job completion. Default: false
1535 Instead of running as the invoking user, set the user ID to this value before
1536 the thread/process does any work.
1539 Set group ID, see \fBuid\fR.
1541 .BI unit_base \fR=\fPint
1542 Base unit for reporting. Allowed values are:
1546 Use auto-detection (default).
1556 .BI flow_id \fR=\fPint
1557 The ID of the flow. If not specified, it defaults to being a global flow. See
1561 Weight in token-based flow control. If this value is used, then there is a
1562 \fBflow counter\fR which is used to regulate the proportion of activity between
1563 two or more jobs. fio attempts to keep this flow counter near zero. The
1564 \fBflow\fR parameter stands for how much should be added or subtracted to the
1565 flow counter on each iteration of the main I/O loop. That is, if one job has
1566 \fBflow=8\fR and another job has \fBflow=-1\fR, then there will be a roughly
1567 1:8 ratio in how much one runs vs the other.
1569 .BI flow_watermark \fR=\fPint
1570 The maximum value that the absolute value of the flow counter is allowed to
1571 reach before the job must wait for a lower value of the counter.
1573 .BI flow_sleep \fR=\fPint
1574 The period of time, in microseconds, to wait after the flow watermark has been
1575 exceeded before retrying operations
1577 .BI clat_percentiles \fR=\fPbool
1578 Enable the reporting of percentiles of completion latencies.
1580 .BI percentile_list \fR=\fPfloat_list
1581 Overwrite the default list of percentiles for completion latencies and the
1582 block error histogram. Each number is a floating number in the range (0,100],
1583 and the maximum length of the list is 20. Use ':' to separate the
1584 numbers. For example, \-\-percentile_list=99.5:99.9 will cause fio to
1585 report the values of completion latency below which 99.5% and 99.9% of
1586 the observed latencies fell, respectively.
1587 .SS "Ioengine Parameters List"
1588 Some parameters are only valid when a specific ioengine is in use. These are
1589 used identically to normal parameters, with the caveat that when used on the
1590 command line, they must come after the ioengine.
1592 .BI (cpu)cpuload \fR=\fPint
1593 Attempt to use the specified percentage of CPU cycles.
1595 .BI (cpu)cpuchunks \fR=\fPint
1596 Split the load into cycles of the given time. In microseconds.
1598 .BI (cpu)exit_on_io_done \fR=\fPbool
1599 Detect when IO threads are done, then exit.
1601 .BI (libaio)userspace_reap
1602 Normally, with the libaio engine in use, fio will use
1603 the io_getevents system call to reap newly returned events.
1604 With this flag turned on, the AIO ring will be read directly
1605 from user-space to reap events. The reaping mode is only
1606 enabled when polling for a minimum of 0 events (eg when
1607 iodepth_batch_complete=0).
1609 .BI (net,netsplice)hostname \fR=\fPstr
1610 The host name or IP address to use for TCP or UDP based IO.
1611 If the job is a TCP listener or UDP reader, the hostname is not
1612 used and must be omitted unless it is a valid UDP multicast address.
1614 .BI (net,netsplice)port \fR=\fPint
1615 The TCP or UDP port to bind to or connect to. If this is used with
1616 \fBnumjobs\fR to spawn multiple instances of the same job type, then
1617 this will be the starting port number since fio will use a range of ports.
1619 .BI (net,netsplice)interface \fR=\fPstr
1620 The IP address of the network interface used to send or receive UDP multicast
1623 .BI (net,netsplice)ttl \fR=\fPint
1624 Time-to-live value for outgoing UDP multicast packets. Default: 1
1626 .BI (net,netsplice)nodelay \fR=\fPbool
1627 Set TCP_NODELAY on TCP connections.
1629 .BI (net,netsplice)protocol \fR=\fPstr "\fR,\fP proto" \fR=\fPstr
1630 The network protocol to use. Accepted values are:
1635 Transmission control protocol
1638 Transmission control protocol V6
1641 User datagram protocol
1644 User datagram protocol V6
1650 When the protocol is TCP or UDP, the port must also be given,
1651 as well as the hostname if the job is a TCP listener or UDP
1652 reader. For unix sockets, the normal filename option should be
1653 used and the port is invalid.
1656 .BI (net,netsplice)listen
1657 For TCP network connections, tell fio to listen for incoming
1658 connections rather than initiating an outgoing connection. The
1659 hostname must be omitted if this option is used.
1661 .BI (net, pingpong) \fR=\fPbool
1662 Normally a network writer will just continue writing data, and a network reader
1663 will just consume packets. If pingpong=1 is set, a writer will send its normal
1664 payload to the reader, then wait for the reader to send the same payload back.
1665 This allows fio to measure network latencies. The submission and completion
1666 latencies then measure local time spent sending or receiving, and the
1667 completion latency measures how long it took for the other end to receive and
1668 send back. For UDP multicast traffic pingpong=1 should only be set for a single
1669 reader when multiple readers are listening to the same address.
1671 .BI (net, window_size) \fR=\fPint
1672 Set the desired socket buffer size for the connection.
1674 .BI (net, mss) \fR=\fPint
1675 Set the TCP maximum segment size (TCP_MAXSEG).
1677 .BI (e4defrag,donorname) \fR=\fPstr
1678 File will be used as a block donor (swap extents between files)
1680 .BI (e4defrag,inplace) \fR=\fPint
1681 Configure donor file block allocation strategy
1684 Preallocate donor's file on init
1687 allocate space immediately inside defragment event, and free right after event
1690 .BI (rbd)rbdname \fR=\fPstr
1691 Specifies the name of the RBD.
1693 .BI (rbd)pool \fR=\fPstr
1694 Specifies the name of the Ceph pool containing the RBD.
1696 .BI (rbd)clientname \fR=\fPstr
1697 Specifies the username (without the 'client.' prefix) used to access the Ceph cluster.
1699 .BI (mtd)skipbad \fR=\fPbool
1700 Skip operations against known bad blocks.
1702 While running, \fBfio\fR will display the status of the created jobs. For
1706 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
1709 The characters in the first set of brackets denote the current status of each
1710 threads. The possible values are:
1716 Setup but not started.
1722 Initialized, waiting.
1725 Running, doing sequential reads.
1728 Running, doing random reads.
1731 Running, doing sequential writes.
1734 Running, doing random writes.
1737 Running, doing mixed sequential reads/writes.
1740 Running, doing mixed random reads/writes.
1743 Running, currently waiting for \fBfsync\fR\|(2).
1746 Running, verifying written data.
1749 Exited, not reaped by main thread.
1752 Exited, thread reaped.
1756 The second set of brackets shows the estimated completion percentage of
1757 the current group. The third set shows the read and write I/O rate,
1758 respectively. Finally, the estimated run time of the job is displayed.
1760 When \fBfio\fR completes (or is interrupted by Ctrl-C), it will show data
1761 for each thread, each group of threads, and each disk, in that order.
1763 Per-thread statistics first show the threads client number, group-id, and
1764 error code. The remaining figures are as follows:
1768 Number of megabytes of I/O performed.
1771 Average data rate (bandwidth).
1777 Submission latency minimum, maximum, average and standard deviation. This is
1778 the time it took to submit the I/O.
1781 Completion latency minimum, maximum, average and standard deviation. This
1782 is the time between submission and completion.
1785 Bandwidth minimum, maximum, percentage of aggregate bandwidth received, average
1786 and standard deviation.
1789 CPU usage statistics. Includes user and system time, number of context switches
1790 this thread went through and number of major and minor page faults.
1793 Distribution of I/O depths. Each depth includes everything less than (or equal)
1794 to it, but greater than the previous depth.
1797 Number of read/write requests issued, and number of short read/write requests.
1800 Distribution of I/O completion latencies. The numbers follow the same pattern
1804 The group statistics show:
1809 Number of megabytes I/O performed.
1812 Aggregate bandwidth of threads in the group.
1815 Minimum average bandwidth a thread saw.
1818 Maximum average bandwidth a thread saw.
1821 Shortest runtime of threads in the group.
1824 Longest runtime of threads in the group.
1828 Finally, disk statistics are printed with reads first:
1833 Number of I/Os performed by all groups.
1836 Number of merges in the I/O scheduler.
1839 Number of ticks we kept the disk busy.
1842 Total time spent in the disk queue.
1849 It is also possible to get fio to dump the current output while it is
1850 running, without terminating the job. To do that, send fio the \fBUSR1\fR
1853 If the \fB\-\-minimal\fR / \fB\-\-append-terse\fR options are given, the
1854 results will be printed/appended in a semicolon-delimited format suitable for
1856 A job description (if provided) follows on a new line. Note that the first
1857 number in the line is the version number. If the output has to be changed
1858 for some reason, this number will be incremented by 1 to signify that
1859 change. The fields are:
1862 .B terse version, fio version, jobname, groupid, error
1866 .B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, IOPS, runtime \fR(ms)\fP
1870 .B min, max, mean, standard deviation
1874 .B min, max, mean, standard deviation
1876 Completion latency percentiles (20 fields):
1878 .B Xth percentile=usec
1882 .B min, max, mean, standard deviation
1886 .B min, max, aggregate percentage of total, mean, standard deviation
1892 .B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, IOPS, runtime \fR(ms)\fP
1896 .B min, max, mean, standard deviation
1900 .B min, max, mean, standard deviation
1902 Completion latency percentiles (20 fields):
1904 .B Xth percentile=usec
1908 .B min, max, mean, standard deviation
1912 .B min, max, aggregate percentage of total, mean, standard deviation
1918 .B user, system, context switches, major page faults, minor page faults
1921 IO depth distribution:
1923 .B <=1, 2, 4, 8, 16, 32, >=64
1926 IO latency distribution:
1930 .B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
1934 .B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
1938 Disk utilization (1 for each disk used):
1940 .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
1943 Error Info (dependent on continue_on_error, default off):
1945 .B total # errors, first error code
1948 .B text description (if provided in config - appears on newline)
1951 Normally you would run fio as a stand-alone application on the machine
1952 where the IO workload should be generated. However, it is also possible to
1953 run the frontend and backend of fio separately. This makes it possible to
1954 have a fio server running on the machine(s) where the IO workload should
1955 be running, while controlling it from another machine.
1957 To start the server, you would do:
1959 \fBfio \-\-server=args\fR
1961 on that machine, where args defines what fio listens to. The arguments
1962 are of the form 'type:hostname or IP:port'. 'type' is either 'ip' (or ip4)
1963 for TCP/IP v4, 'ip6' for TCP/IP v6, or 'sock' for a local unix domain
1964 socket. 'hostname' is either a hostname or IP address, and 'port' is the port to
1965 listen to (only valid for TCP/IP, not a local socket). Some examples:
1969 Start a fio server, listening on all interfaces on the default port (8765).
1971 2) fio \-\-server=ip:hostname,4444
1973 Start a fio server, listening on IP belonging to hostname and on port 4444.
1975 3) fio \-\-server=ip6:::1,4444
1977 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
1979 4) fio \-\-server=,4444
1981 Start a fio server, listening on all interfaces on port 4444.
1983 5) fio \-\-server=1.2.3.4
1985 Start a fio server, listening on IP 1.2.3.4 on the default port.
1987 6) fio \-\-server=sock:/tmp/fio.sock
1989 Start a fio server, listening on the local socket /tmp/fio.sock.
1991 When a server is running, you can connect to it from a client. The client
1994 fio \-\-local-args \-\-client=server \-\-remote-args <job file(s)>
1996 where \-\-local-args are arguments that are local to the client where it is
1997 running, 'server' is the connect string, and \-\-remote-args and <job file(s)>
1998 are sent to the server. The 'server' string follows the same format as it
1999 does on the server side, to allow IP/hostname/socket and port strings.
2000 You can connect to multiple clients as well, to do that you could run:
2002 fio \-\-client=server2 \-\-client=server2 <job file(s)>
2004 If the job file is located on the fio server, then you can tell the server
2005 to load a local file as well. This is done by using \-\-remote-config:
2007 fio \-\-client=server \-\-remote-config /path/to/file.fio
2009 Then fio will open this local (to the server) job file instead
2010 of being passed one from the client.
2012 If you have many servers (example: 100 VMs/containers), you can input a pathname
2013 of a file containing host IPs/names as the parameter value for the \-\-client option.
2014 For example, here is an example "host.list" file containing 2 hostnames:
2016 host1.your.dns.domain
2018 host2.your.dns.domain
2020 The fio command would then be:
2022 fio \-\-client=host.list <job file>
2024 In this mode, you cannot input server-specific parameters or job files, and all
2025 servers receive the same job file.
2027 In order to enable fio \-\-client runs utilizing a shared filesystem from multiple hosts,
2028 fio \-\-client now prepends the IP address of the server to the filename. For example,
2029 if fio is using directory /mnt/nfs/fio and is writing filename fileio.tmp,
2030 with a \-\-client hostfile
2031 containing two hostnames h1 and h2 with IP addresses 192.168.10.120 and 192.168.10.121, then
2032 fio will create two files:
2034 /mnt/nfs/fio/192.168.10.120.fileio.tmp
2036 /mnt/nfs/fio/192.168.10.121.fileio.tmp
2041 was written by Jens Axboe <jens.axboe@oracle.com>,
2042 now Jens Axboe <axboe@fb.com>.
2044 This man page was written by Aaron Carroll <aaronc@cse.unsw.edu.au> based
2045 on documentation by Jens Axboe.
2046 .SH "REPORTING BUGS"
2047 Report bugs to the \fBfio\fR mailing list <fio@vger.kernel.org>.
2050 For further documentation see \fBHOWTO\fR and \fBREADME\fR.
2052 Sample jobfiles are available in the \fBexamples\fR directory.
projects / fio.git / blob