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, \fIjson\fR, or
25 \fIjson+\fR. Multiple formats can be selected, separate by a comma. \fIterse\fR
26 is a CSV based format. \fIjson+\fR is like \fIjson\fR, except it adds a full
27 dump of the latency buckets.
29 .BI \-\-runtime \fR=\fPruntime
30 Limit run time to \fIruntime\fR seconds.
33 Generate per-job bandwidth logs.
36 Print statistics in a terse, semicolon-delimited format.
39 Print statistics in selected mode AND terse, semicolon-delimited format.
40 Deprecated, use \-\-output-format instead to select multiple formats.
43 Display version information and exit.
45 .BI \-\-terse\-version \fR=\fPversion
46 Set terse version output format (Current version 3, or older version 2).
49 Display usage information and exit.
52 Perform test and validation of internal CPU clock
54 .BI \-\-crctest[\fR=\fPtest]
55 Test the speed of the builtin checksumming functions. If no argument is given,
56 all of them are tested. Or a comma separated list can be passed, in which
57 case the given ones are tested.
59 .BI \-\-cmdhelp \fR=\fPcommand
60 Print help information for \fIcommand\fR. May be `all' for all commands.
62 .BI \-\-enghelp \fR=\fPioengine[,command]
63 List all commands defined by \fIioengine\fR, or print help for \fIcommand\fR defined by \fIioengine\fR.
65 .BI \-\-showcmd \fR=\fPjobfile
66 Convert \fIjobfile\fR to a set of command-line options.
68 .BI \-\-eta \fR=\fPwhen
69 Specifies when real-time ETA estimate should be printed. \fIwhen\fR may
70 be one of `always', `never' or `auto'.
72 .BI \-\-eta\-newline \fR=\fPtime
73 Force an ETA newline for every `time` period passed.
75 .BI \-\-status\-interval \fR=\fPtime
76 Report full output status every `time` period passed.
79 Turn on safety read-only checks, preventing any attempted write.
81 .BI \-\-section \fR=\fPsec
82 Only run section \fIsec\fR from job file. This option can be used multiple times to add more sections to run.
84 .BI \-\-alloc\-size \fR=\fPkb
85 Set the internal smalloc pool size to \fIkb\fP kilobytes.
87 .BI \-\-warnings\-fatal
88 All fio parser warnings are fatal, causing fio to exit with an error.
90 .BI \-\-max\-jobs \fR=\fPnr
91 Set the maximum allowed number of jobs (threads/processes) to support.
93 .BI \-\-server \fR=\fPargs
94 Start a backend server, with \fIargs\fP specifying what to listen to. See client/server section.
96 .BI \-\-daemonize \fR=\fPpidfile
97 Background a fio server, writing the pid to the given pid file.
99 .BI \-\-client \fR=\fPhost
100 Instead of running the jobs locally, send and run them on the given host or set of hosts. See client/server section.
102 .BI \-\-idle\-prof \fR=\fPoption
103 Report cpu idleness on a system or percpu basis (\fIoption\fP=system,percpu) or run unit work calibration only (\fIoption\fP=calibrate).
104 .SH "JOB FILE FORMAT"
105 Job files are in `ini' format. They consist of one or more
106 job definitions, which begin with a job name in square brackets and
107 extend to the next job name. The job name can be any ASCII string
108 except `global', which has a special meaning. Following the job name is
109 a sequence of zero or more parameters, one per line, that define the
110 behavior of the job. Any line starting with a `;' or `#' character is
111 considered a comment and ignored.
113 If \fIjobfile\fR is specified as `-', the job file will be read from
116 The global section contains default parameters for jobs specified in the
117 job file. A job is only affected by global sections residing above it,
118 and there may be any number of global sections. Specific job definitions
119 may override any parameter set in global sections.
122 Some parameters may take arguments of a specific type.
123 Anywhere a numeric value is required, an arithmetic expression may be used,
124 provided it is surrounded by parentheses. Supported operators are:
132 .B multiplication (*)
138 .B exponentiation (^)
142 For time values in expressions, units are microseconds by default. This is
143 different than for time values not in expressions (not enclosed in
144 parentheses). The types used are:
147 String: a sequence of alphanumeric characters.
150 SI integer: a whole number, possibly containing a suffix denoting the base unit
151 of the value. Accepted suffixes are `k', 'M', 'G', 'T', and 'P', denoting
152 kilo (1024), mega (1024^2), giga (1024^3), tera (1024^4), and peta (1024^5)
153 respectively. If prefixed with '0x', the value is assumed to be base 16
154 (hexadecimal). A suffix may include a trailing 'b', for instance 'kb' is
155 identical to 'k'. You can specify a base 10 value by using 'KiB', 'MiB','GiB',
156 etc. This is useful for disk drives where values are often given in base 10
157 values. Specifying '30GiB' will get you 30*1000^3 bytes.
158 When specifying times the default suffix meaning changes, still denoting the
159 base unit of the value, but accepted suffixes are 'D' (days), 'H' (hours), 'M'
160 (minutes), 'S' Seconds, 'ms' (or msec) milli seconds, 'us' (or 'usec') micro
161 seconds. Time values without a unit specify seconds.
162 The suffixes are not case sensitive.
165 Boolean: a true or false value. `0' denotes false, `1' denotes true.
168 Integer range: a range of integers specified in the format
169 \fIlower\fR:\fIupper\fR or \fIlower\fR\-\fIupper\fR. \fIlower\fR and
170 \fIupper\fR may contain a suffix as described above. If an option allows two
171 sets of ranges, they are separated with a `,' or `/' character. For example:
175 List of floating numbers: A list of floating numbers, separated by
180 May be used to override the job name. On the command line, this parameter
181 has the special purpose of signalling the start of a new job.
183 .BI description \fR=\fPstr
184 Human-readable description of the job. It is printed when the job is run, but
185 otherwise has no special purpose.
187 .BI directory \fR=\fPstr
188 Prefix filenames with this directory. Used to place files in a location other
190 You can specify a number of directories by separating the names with a ':'
191 character. These directories will be assigned equally distributed to job clones
192 creates with \fInumjobs\fR as long as they are using generated filenames.
193 If specific \fIfilename(s)\fR are set fio will use the first listed directory,
194 and thereby matching the \fIfilename\fR semantic which generates a file each
195 clone if not specified, but let all clones use the same if set. See
196 \fIfilename\fR for considerations regarding escaping certain characters on
199 .BI filename \fR=\fPstr
201 normally makes up a file name based on the job name, thread number, and file
202 number. If you want to share files between threads in a job or several jobs,
203 specify a \fIfilename\fR for each of them to override the default.
204 If the I/O engine is file-based, you can specify
205 a number of files by separating the names with a `:' character. `\-' is a
206 reserved name, meaning stdin or stdout, depending on the read/write direction
207 set. On Windows, disk devices are accessed as \\.\PhysicalDrive0 for the first
208 device, \\.\PhysicalDrive1 for the second etc. Note: Windows and FreeBSD
209 prevent write access to areas of the disk containing in-use data
210 (e.g. filesystems). If the wanted filename does need to include a colon, then
211 escape that with a '\\' character. For instance, if the filename is
212 "/dev/dsk/foo@3,0:c", then you would use filename="/dev/dsk/foo@3,0\\:c".
214 .BI filename_format \fR=\fPstr
215 If sharing multiple files between jobs, it is usually necessary to have
216 fio generate the exact names that you want. By default, fio will name a file
217 based on the default file format specification of
218 \fBjobname.jobnumber.filenumber\fP. With this option, that can be
219 customized. Fio will recognize and replace the following keywords in this
225 The name of the worker thread or process.
228 The incremental number of the worker thread or process.
231 The incremental number of the file for that worker thread or process.
234 To have dependent jobs share a set of files, this option can be set to
235 have fio generate filenames that are shared between the two. For instance,
236 if \fBtestfiles.$filenum\fR is specified, file number 4 for any job will
237 be named \fBtestfiles.4\fR. The default of \fB$jobname.$jobnum.$filenum\fR
238 will be used if no other format specifier is given.
242 .BI lockfile \fR=\fPstr
243 Fio defaults to not locking any files before it does IO to them. If a file or
244 file descriptor is shared, fio can serialize IO to that file to make the end
245 result consistent. This is usual for emulating real workloads that share files.
251 No locking. This is the default.
254 Only one thread or process may do IO at a time, excluding all others.
257 Read-write locking on the file. Many readers may access the file at the same
258 time, but writes get exclusive access.
262 .BI opendir \fR=\fPstr
263 Recursively open any files below directory \fIstr\fR.
265 .BI readwrite \fR=\fPstr "\fR,\fP rw" \fR=\fPstr
266 Type of I/O pattern. Accepted values are:
277 Sequential trim (Linux block devices only).
286 Random trim (Linux block devices only).
289 Mixed sequential reads and writes.
292 Mixed random reads and writes.
295 Trim and write mixed workload. Blocks will be trimmed first, then the same
296 blocks will be written to.
299 For mixed I/O, the default split is 50/50. For certain types of io the result
300 may still be skewed a bit, since the speed may be different. It is possible to
301 specify a number of IO's to do before getting a new offset, this is done by
302 appending a `:\fI<nr>\fR to the end of the string given. For a random read, it
303 would look like \fBrw=randread:8\fR for passing in an offset modifier with a
304 value of 8. If the postfix is used with a sequential IO pattern, then the value
305 specified will be added to the generated offset for each IO. For instance,
306 using \fBrw=write:4k\fR will skip 4k for every write. It turns sequential IO
307 into sequential IO with holes. See the \fBrw_sequencer\fR option.
310 .BI rw_sequencer \fR=\fPstr
311 If an offset modifier is given by appending a number to the \fBrw=<str>\fR line,
312 then this option controls how that number modifies the IO offset being
313 generated. Accepted values are:
318 Generate sequential offset
321 Generate the same offset
324 \fBsequential\fR is only useful for random IO, where fio would normally
325 generate a new random offset for every IO. If you append eg 8 to randread, you
326 would get a new random offset for every 8 IO's. The result would be a seek for
327 only every 8 IO's, instead of for every IO. Use \fBrw=randread:8\fR to specify
328 that. As sequential IO is already sequential, setting \fBsequential\fR for that
329 would not result in any differences. \fBidentical\fR behaves in a similar
330 fashion, except it sends the same offset 8 number of times before generating a
335 .BI kb_base \fR=\fPint
336 The base unit for a kilobyte. The defacto base is 2^10, 1024. Storage
337 manufacturers like to use 10^3 or 1000 as a base ten unit instead, for obvious
338 reasons. Allowed values are 1024 or 1000, with 1024 being the default.
340 .BI unified_rw_reporting \fR=\fPbool
341 Fio normally reports statistics on a per data direction basis, meaning that
342 read, write, and trim are accounted and reported separately. If this option is
343 set fio sums the results and reports them as "mixed" instead.
345 .BI randrepeat \fR=\fPbool
346 Seed the random number generator used for random I/O patterns in a predictable
347 way so the pattern is repeatable across runs. Default: true.
349 .BI allrandrepeat \fR=\fPbool
350 Seed all random number generators in a predictable way so results are
351 repeatable across runs. Default: false.
353 .BI randseed \fR=\fPint
354 Seed the random number generators based on this seed value, to be able to
355 control what sequence of output is being generated. If not set, the random
356 sequence depends on the \fBrandrepeat\fR setting.
358 .BI fallocate \fR=\fPstr
359 Whether pre-allocation is performed when laying down files. Accepted values
365 Do not pre-allocate space.
368 Pre-allocate via \fBposix_fallocate\fR\|(3).
371 Pre-allocate via \fBfallocate\fR\|(2) with FALLOC_FL_KEEP_SIZE set.
374 Backward-compatible alias for 'none'.
377 Backward-compatible alias for 'posix'.
380 May not be available on all supported platforms. 'keep' is only
381 available on Linux. If using ZFS on Solaris this must be set to 'none'
382 because ZFS doesn't support it. Default: 'posix'.
385 .BI fadvise_hint \fR=\fPbool
386 Use \fBposix_fadvise\fR\|(2) to advise the kernel what I/O patterns
387 are likely to be issued. Default: true.
389 .BI fadvise_stream \fR=\fPint
390 Use \fBposix_fadvise\fR\|(2) to advise the kernel what stream ID the
391 writes issued belong to. Only supported on Linux. Note, this option
392 may change going forward.
395 Total size of I/O for this job. \fBfio\fR will run until this many bytes have
396 been transferred, unless limited by other options (\fBruntime\fR, for instance,
397 or increased/descreased by \fBio_size\fR). Unless \fBnrfiles\fR and
398 \fBfilesize\fR options are given, this amount will be divided between the
399 available files for the job. If not set, fio will use the full size of the
400 given files or devices. If the files do not exist, size must be given. It is
401 also possible to give size as a percentage between 1 and 100. If size=20% is
402 given, fio will use 20% of the full size of the given files or devices.
404 .BI io_size \fR=\fPint "\fR,\fB io_limit \fR=\fPint
405 Normally fio operates within the region set by \fBsize\fR, which means that
406 the \fBsize\fR option sets both the region and size of IO to be performed.
407 Sometimes that is not what you want. With this option, it is possible to
408 define just the amount of IO that fio should do. For instance, if \fBsize\fR
409 is set to 20G and \fBio_limit\fR is set to 5G, fio will perform IO within
410 the first 20G but exit when 5G have been done. The opposite is also
411 possible - if \fBsize\fR is set to 20G, and \fBio_size\fR is set to 40G, then
412 fio will do 40G of IO within the 0..20G region.
414 .BI fill_device \fR=\fPbool "\fR,\fB fill_fs" \fR=\fPbool
415 Sets size to something really large and waits for ENOSPC (no space left on
416 device) as the terminating condition. Only makes sense with sequential write.
417 For a read workload, the mount point will be filled first then IO started on
418 the result. This option doesn't make sense if operating on a raw device node,
419 since the size of that is already known by the file system. Additionally,
420 writing beyond end-of-device will not return ENOSPC there.
422 .BI filesize \fR=\fPirange
423 Individual file sizes. May be a range, in which case \fBfio\fR will select sizes
424 for files at random within the given range, limited to \fBsize\fR in total (if
425 that is given). If \fBfilesize\fR is not specified, each created file is the
428 .BI file_append \fR=\fPbool
429 Perform IO after the end of the file. Normally fio will operate within the
430 size of a file. If this option is set, then fio will append to the file
431 instead. This has identical behavior to setting \fRoffset\fP to the size
432 of a file. This option is ignored on non-regular files.
434 .BI blocksize \fR=\fPint[,int] "\fR,\fB bs" \fR=\fPint[,int]
435 Block size for I/O units. Default: 4k. Values for reads, writes, and trims
436 can be specified separately in the format \fIread\fR,\fIwrite\fR,\fItrim\fR
437 either of which may be empty to leave that value at its default. If a trailing
438 comma isn't given, the remainder will inherit the last value set.
440 .BI blocksize_range \fR=\fPirange[,irange] "\fR,\fB bsrange" \fR=\fPirange[,irange]
441 Specify a range of I/O block sizes. The issued I/O unit will always be a
442 multiple of the minimum size, unless \fBblocksize_unaligned\fR is set. Applies
443 to both reads and writes if only one range is given, but can be specified
444 separately with a comma separating the values. Example: bsrange=1k-4k,2k-8k.
445 Also (see \fBblocksize\fR).
447 .BI bssplit \fR=\fPstr
448 This option allows even finer grained control of the block sizes issued,
449 not just even splits between them. With this option, you can weight various
450 block sizes for exact control of the issued IO for a job that has mixed
451 block sizes. The format of the option is bssplit=blocksize/percentage,
452 optionally adding as many definitions as needed separated by a colon.
453 Example: bssplit=4k/10:64k/50:32k/40 would issue 50% 64k blocks, 10% 4k
454 blocks and 40% 32k blocks. \fBbssplit\fR also supports giving separate
455 splits to reads and writes. The format is identical to what the
456 \fBbs\fR option accepts, the read and write parts are separated with a
459 .B blocksize_unaligned\fR,\fP bs_unaligned
460 If set, any size in \fBblocksize_range\fR may be used. This typically won't
461 work with direct I/O, as that normally requires sector alignment.
463 .BI blockalign \fR=\fPint[,int] "\fR,\fB ba" \fR=\fPint[,int]
464 At what boundary to align random IO offsets. Defaults to the same as 'blocksize'
465 the minimum blocksize given. Minimum alignment is typically 512b
466 for using direct IO, though it usually depends on the hardware block size.
467 This option is mutually exclusive with using a random map for files, so it
468 will turn off that option.
470 .BI bs_is_seq_rand \fR=\fPbool
471 If this option is set, fio will use the normal read,write blocksize settings as
472 sequential,random instead. Any random read or write will use the WRITE
473 blocksize settings, and any sequential read or write will use the READ
477 Initialize buffers with all zeros. Default: fill buffers with random data.
480 If this option is given, fio will refill the IO buffers on every submit. The
481 default is to only fill it at init time and reuse that data. Only makes sense
482 if zero_buffers isn't specified, naturally. If data verification is enabled,
483 refill_buffers is also automatically enabled.
485 .BI scramble_buffers \fR=\fPbool
486 If \fBrefill_buffers\fR is too costly and the target is using data
487 deduplication, then setting this option will slightly modify the IO buffer
488 contents to defeat normal de-dupe attempts. This is not enough to defeat
489 more clever block compression attempts, but it will stop naive dedupe
490 of blocks. Default: true.
492 .BI buffer_compress_percentage \fR=\fPint
493 If this is set, then fio will attempt to provide IO buffer content (on WRITEs)
494 that compress to the specified level. Fio does this by providing a mix of
495 random data and a fixed pattern. The fixed pattern is either zeroes, or the
496 pattern specified by \fBbuffer_pattern\fR. If the pattern option is used, it
497 might skew the compression ratio slightly. Note that this is per block size
498 unit, for file/disk wide compression level that matches this setting. Note
499 that this is per block size unit, for file/disk wide compression level that
500 matches this setting, you'll also want to set refill_buffers.
502 .BI buffer_compress_chunk \fR=\fPint
503 See \fBbuffer_compress_percentage\fR. This setting allows fio to manage how
504 big the ranges of random data and zeroed data is. Without this set, fio will
505 provide \fBbuffer_compress_percentage\fR of blocksize random data, followed by
506 the remaining zeroed. With this set to some chunk size smaller than the block
507 size, fio can alternate random and zeroed data throughout the IO buffer.
509 .BI buffer_pattern \fR=\fPstr
510 If set, fio will fill the IO buffers with this pattern. If not set, the contents
511 of IO buffers is defined by the other options related to buffer contents. The
512 setting can be any pattern of bytes, and can be prefixed with 0x for hex
513 values. It may also be a string, where the string must then be wrapped with
517 \fBbuffer_pattern\fR="abcd"
521 \fBbuffer_pattern\fR=-12
525 \fBbuffer_pattern\fR=0xdeadface
528 Also you can combine everything together in any order:
531 \fBbuffer_pattern\fR=0xdeadface"abcd"-12
535 .BI dedupe_percentage \fR=\fPint
536 If set, fio will generate this percentage of identical buffers when writing.
537 These buffers will be naturally dedupable. The contents of the buffers depend
538 on what other buffer compression settings have been set. It's possible to have
539 the individual buffers either fully compressible, or not at all. This option
540 only controls the distribution of unique buffers.
542 .BI nrfiles \fR=\fPint
543 Number of files to use for this job. Default: 1.
545 .BI openfiles \fR=\fPint
546 Number of files to keep open at the same time. Default: \fBnrfiles\fR.
548 .BI file_service_type \fR=\fPstr
549 Defines how files to service are selected. The following types are defined:
554 Choose a file at random.
557 Round robin over opened files (default).
560 Do each file in the set sequentially.
563 The number of I/Os to issue before switching to a new file can be specified by
564 appending `:\fIint\fR' to the service type.
567 .BI ioengine \fR=\fPstr
568 Defines how the job issues I/O. The following types are defined:
573 Basic \fBread\fR\|(2) or \fBwrite\fR\|(2) I/O. \fBfseek\fR\|(2) is used to
574 position the I/O location.
577 Basic \fBpread\fR\|(2) or \fBpwrite\fR\|(2) I/O.
580 Basic \fBreadv\fR\|(2) or \fBwritev\fR\|(2) I/O. Will emulate queuing by
581 coalescing adjacent IOs into a single submission.
584 Basic \fBpreadv\fR\|(2) or \fBpwritev\fR\|(2) I/O.
587 Linux native asynchronous I/O. This ioengine defines engine specific options.
590 POSIX asynchronous I/O using \fBaio_read\fR\|(3) and \fBaio_write\fR\|(3).
593 Solaris native asynchronous I/O.
596 Windows native asynchronous I/O.
599 File is memory mapped with \fBmmap\fR\|(2) and data copied using
603 \fBsplice\fR\|(2) is used to transfer the data and \fBvmsplice\fR\|(2) to
604 transfer data from user-space to the kernel.
607 Use the syslet system calls to make regular read/write asynchronous.
610 SCSI generic sg v3 I/O. May be either synchronous using the SG_IO ioctl, or if
611 the target is an sg character device, we use \fBread\fR\|(2) and
612 \fBwrite\fR\|(2) for asynchronous I/O.
615 Doesn't transfer any data, just pretends to. Mainly used to exercise \fBfio\fR
616 itself and for debugging and testing purposes.
619 Transfer over the network. The protocol to be used can be defined with the
620 \fBprotocol\fR parameter. Depending on the protocol, \fBfilename\fR,
621 \fBhostname\fR, \fBport\fR, or \fBlisten\fR must be specified.
622 This ioengine defines engine specific options.
625 Like \fBnet\fR, but uses \fBsplice\fR\|(2) and \fBvmsplice\fR\|(2) to map data
626 and send/receive. This ioengine defines engine specific options.
629 Doesn't transfer any data, but burns CPU cycles according to \fBcpuload\fR and
630 \fBcpucycles\fR parameters.
633 The GUASI I/O engine is the Generic Userspace Asynchronous Syscall Interface
634 approach to asynchronous I/O.
636 See <http://www.xmailserver.org/guasi\-lib.html>.
639 The RDMA I/O engine supports both RDMA memory semantics (RDMA_WRITE/RDMA_READ)
640 and channel semantics (Send/Recv) for the InfiniBand, RoCE and iWARP protocols.
643 Loads an external I/O engine object file. Append the engine filename as
647 IO engine that does regular linux native fallocate call to simulate data
648 transfer as fio ioengine
650 DDIR_READ does fallocate(,mode = FALLOC_FL_KEEP_SIZE,)
652 DIR_WRITE does fallocate(,mode = 0)
654 DDIR_TRIM does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE)
657 IO engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate defragment activity
658 request to DDIR_WRITE event
661 IO engine supporting direct access to Ceph Rados Block Devices (RBD) via librbd
662 without the need to use the kernel rbd driver. This ioengine defines engine specific
666 Using Glusterfs libgfapi sync interface to direct access to Glusterfs volumes without
667 having to go through FUSE. This ioengine defines engine specific
671 Using Glusterfs libgfapi async interface to direct access to Glusterfs volumes without
672 having to go through FUSE. This ioengine defines engine specific
676 Read and write through Hadoop (HDFS). The \fBfilename\fR option is used to
677 specify host,port of the hdfs name-node to connect. This engine interprets
678 offsets a little differently. In HDFS, files once created cannot be modified.
679 So random writes are not possible. To imitate this, libhdfs engine expects
680 bunch of small files to be created over HDFS, and engine will randomly pick a
681 file out of those files based on the offset generated by fio backend. (see the
682 example job file to create such files, use rw=write option). Please note, you
683 might want to set necessary environment variables to work with hdfs/libhdfs
687 Read, write and erase an MTD character device (e.g., /dev/mtd0). Discards are
688 treated as erases. Depending on the underlying device type, the I/O may have
689 to go in a certain pattern, e.g., on NAND, writing sequentially to erase blocks
690 and discarding before overwriting. The writetrim mode works well for this
696 .BI iodepth \fR=\fPint
697 Number of I/O units to keep in flight against the file. Note that increasing
698 iodepth beyond 1 will not affect synchronous ioengines (except for small
699 degress when verify_async is in use). Even async engines may impose OS
700 restrictions causing the desired depth not to be achieved. This may happen on
701 Linux when using libaio and not setting \fBdirect\fR=1, since buffered IO is
702 not async on that OS. Keep an eye on the IO depth distribution in the
703 fio output to verify that the achieved depth is as expected. Default: 1.
705 .BI iodepth_batch \fR=\fPint "\fR,\fP iodepth_batch_submit" \fR=\fPint
706 This defines how many pieces of IO to submit at once. It defaults to 1
707 which means that we submit each IO as soon as it is available, but can
708 be raised to submit bigger batches of IO at the time. If it is set to 0
709 the \fBiodepth\fR value will be used.
711 .BI iodepth_batch_complete_min \fR=\fPint "\fR,\fP iodepth_batch_complete" \fR=\fPint
712 This defines how many pieces of IO to retrieve at once. It defaults to 1 which
713 means that we'll ask for a minimum of 1 IO in the retrieval process from the
714 kernel. The IO retrieval will go on until we hit the limit set by
715 \fBiodepth_low\fR. If this variable is set to 0, then fio will always check for
716 completed events before queuing more IO. This helps reduce IO latency, at the
717 cost of more retrieval system calls.
719 .BI iodepth_batch_complete_max \fR=\fPint
720 This defines maximum pieces of IO to
721 retrieve at once. This variable should be used along with
722 \fBiodepth_batch_complete_min\fR=int variable, specifying the range
723 of min and max amount of IO which should be retrieved. By default
724 it is equal to \fBiodepth_batch_complete_min\fR value.
729 \fBiodepth_batch_complete_min\fR=1
731 \fBiodepth_batch_complete_max\fR=<iodepth>
734 which means that we will retrieve at leat 1 IO and up to the
735 whole submitted queue depth. If none of IO has been completed
740 \fBiodepth_batch_complete_min\fR=0
742 \fBiodepth_batch_complete_max\fR=<iodepth>
745 which means that we can retrieve up to the whole submitted
746 queue depth, but if none of IO has been completed yet, we will
747 NOT wait and immediately exit the system call. In this example
748 we simply do polling.
751 .BI iodepth_low \fR=\fPint
752 Low watermark indicating when to start filling the queue again. Default:
755 .BI io_submit_mode \fR=\fPstr
756 This option controls how fio submits the IO to the IO engine. The default is
757 \fBinline\fR, which means that the fio job threads submit and reap IO directly.
758 If set to \fBoffload\fR, the job threads will offload IO submission to a
759 dedicated pool of IO threads. This requires some coordination and thus has a
760 bit of extra overhead, especially for lower queue depth IO where it can
761 increase latencies. The benefit is that fio can manage submission rates
762 independently of the device completion rates. This avoids skewed latency
763 reporting if IO gets back up on the device side (the coordinated omission
766 .BI direct \fR=\fPbool
767 If true, use non-buffered I/O (usually O_DIRECT). Default: false.
769 .BI atomic \fR=\fPbool
770 If value is true, attempt to use atomic direct IO. Atomic writes are guaranteed
771 to be stable once acknowledged by the operating system. Only Linux supports
774 .BI buffered \fR=\fPbool
775 If true, use buffered I/O. This is the opposite of the \fBdirect\fR parameter.
778 .BI offset \fR=\fPint
779 Offset in the file to start I/O. Data before the offset will not be touched.
781 .BI offset_increment \fR=\fPint
782 If this is provided, then the real offset becomes the
783 offset + offset_increment * thread_number, where the thread number is a
784 counter that starts at 0 and is incremented for each sub-job (i.e. when
785 numjobs option is specified). This option is useful if there are several jobs
786 which are intended to operate on a file in parallel disjoint segments, with
787 even spacing between the starting points.
789 .BI number_ios \fR=\fPint
790 Fio will normally perform IOs until it has exhausted the size of the region
791 set by \fBsize\fR, or if it exhaust the allocated time (or hits an error
792 condition). With this setting, the range/size can be set independently of
793 the number of IOs to perform. When fio reaches this number, it will exit
794 normally and report status. Note that this does not extend the amount
795 of IO that will be done, it will only stop fio if this condition is met
796 before other end-of-job criteria.
799 How many I/Os to perform before issuing an \fBfsync\fR\|(2) of dirty data. If
800 0, don't sync. Default: 0.
802 .BI fdatasync \fR=\fPint
803 Like \fBfsync\fR, but uses \fBfdatasync\fR\|(2) instead to only sync the
804 data parts of the file. Default: 0.
806 .BI write_barrier \fR=\fPint
807 Make every Nth write a barrier write.
809 .BI sync_file_range \fR=\fPstr:int
810 Use \fBsync_file_range\fR\|(2) for every \fRval\fP number of write operations. Fio will
811 track range of writes that have happened since the last \fBsync_file_range\fR\|(2) call.
812 \fRstr\fP can currently be one or more of:
816 SYNC_FILE_RANGE_WAIT_BEFORE
819 SYNC_FILE_RANGE_WRITE
822 SYNC_FILE_RANGE_WRITE
826 So if you do sync_file_range=wait_before,write:8, fio would use
827 \fBSYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE\fP for every 8 writes.
828 Also see the \fBsync_file_range\fR\|(2) man page. This option is Linux specific.
830 .BI overwrite \fR=\fPbool
831 If writing, setup the file first and do overwrites. Default: false.
833 .BI end_fsync \fR=\fPbool
834 Sync file contents when a write stage has completed. Default: false.
836 .BI fsync_on_close \fR=\fPbool
837 If true, sync file contents on close. This differs from \fBend_fsync\fR in that
838 it will happen on every close, not just at the end of the job. Default: false.
840 .BI rwmixread \fR=\fPint
841 Percentage of a mixed workload that should be reads. Default: 50.
843 .BI rwmixwrite \fR=\fPint
844 Percentage of a mixed workload that should be writes. If \fBrwmixread\fR and
845 \fBrwmixwrite\fR are given and do not sum to 100%, the latter of the two
846 overrides the first. This may interfere with a given rate setting, if fio is
847 asked to limit reads or writes to a certain rate. If that is the case, then
848 the distribution may be skewed. Default: 50.
850 .BI random_distribution \fR=\fPstr:float
851 By default, fio will use a completely uniform random distribution when asked
852 to perform random IO. Sometimes it is useful to skew the distribution in
853 specific ways, ensuring that some parts of the data is more hot than others.
854 Fio includes the following distribution models:
858 Uniform random distribution
868 When using a zipf or pareto distribution, an input value is also needed to
869 define the access pattern. For zipf, this is the zipf theta. For pareto,
870 it's the pareto power. Fio includes a test program, genzipf, that can be
871 used visualize what the given input values will yield in terms of hit rates.
872 If you wanted to use zipf with a theta of 1.2, you would use
873 random_distribution=zipf:1.2 as the option. If a non-uniform model is used,
874 fio will disable use of the random map.
876 .BI percentage_random \fR=\fPint
877 For a random workload, set how big a percentage should be random. This defaults
878 to 100%, in which case the workload is fully random. It can be set from
879 anywhere from 0 to 100. Setting it to 0 would make the workload fully
880 sequential. It is possible to set different values for reads, writes, and
881 trim. To do so, simply use a comma separated list. See \fBblocksize\fR.
884 Normally \fBfio\fR will cover every block of the file when doing random I/O. If
885 this parameter is given, a new offset will be chosen without looking at past
886 I/O history. This parameter is mutually exclusive with \fBverify\fR.
888 .BI softrandommap \fR=\fPbool
889 See \fBnorandommap\fR. If fio runs with the random block map enabled and it
890 fails to allocate the map, if this option is set it will continue without a
891 random block map. As coverage will not be as complete as with random maps, this
892 option is disabled by default.
894 .BI random_generator \fR=\fPstr
895 Fio supports the following engines for generating IO offsets for random IO:
899 Strong 2^88 cycle random number generator
902 Linear feedback shift register generator
905 Strong 64-bit 2^258 cycle random number generator
909 Tausworthe is a strong random number generator, but it requires tracking on the
910 side if we want to ensure that blocks are only read or written once. LFSR
911 guarantees that we never generate the same offset twice, and it's also less
912 computationally expensive. It's not a true random generator, however, though
913 for IO purposes it's typically good enough. LFSR only works with single block
914 sizes, not with workloads that use multiple block sizes. If used with such a
915 workload, fio may read or write some blocks multiple times.
918 Run job with given nice value. See \fBnice\fR\|(2).
921 Set I/O priority value of this job between 0 (highest) and 7 (lowest). See
924 .BI prioclass \fR=\fPint
925 Set I/O priority class. See \fBionice\fR\|(1).
927 .BI thinktime \fR=\fPint
928 Stall job for given number of microseconds between issuing I/Os.
930 .BI thinktime_spin \fR=\fPint
931 Pretend to spend CPU time for given number of microseconds, sleeping the rest
932 of the time specified by \fBthinktime\fR. Only valid if \fBthinktime\fR is set.
934 .BI thinktime_blocks \fR=\fPint
935 Only valid if thinktime is set - control how many blocks to issue, before
936 waiting \fBthinktime\fR microseconds. If not set, defaults to 1 which will
937 make fio wait \fBthinktime\fR microseconds after every block. This
938 effectively makes any queue depth setting redundant, since no more than 1 IO
939 will be queued before we have to complete it and do our thinktime. In other
940 words, this setting effectively caps the queue depth if the latter is larger.
944 Cap bandwidth used by this job. The number is in bytes/sec, the normal postfix
945 rules apply. You can use \fBrate\fR=500k to limit reads and writes to 500k each,
946 or you can specify read and writes separately. Using \fBrate\fR=1m,500k would
947 limit reads to 1MB/sec and writes to 500KB/sec. Capping only reads or writes
948 can be done with \fBrate\fR=,500k or \fBrate\fR=500k,. The former will only
949 limit writes (to 500KB/sec), the latter will only limit reads.
951 .BI rate_min \fR=\fPint
952 Tell \fBfio\fR to do whatever it can to maintain at least the given bandwidth.
953 Failing to meet this requirement will cause the job to exit. The same format
954 as \fBrate\fR is used for read vs write separation.
956 .BI rate_iops \fR=\fPint
957 Cap the bandwidth to this number of IOPS. Basically the same as rate, just
958 specified independently of bandwidth. The same format as \fBrate\fR is used for
959 read vs write separation. If \fBblocksize\fR is a range, the smallest block
960 size is used as the metric.
962 .BI rate_iops_min \fR=\fPint
963 If this rate of I/O is not met, the job will exit. The same format as \fBrate\fR
964 is used for read vs write separation.
966 .BI rate_process \fR=\fPstr
967 This option controls how fio manages rated IO submissions. The default is
968 \fBlinear\fR, which submits IO in a linear fashion with fixed delays between
969 IOs that gets adjusted based on IO completion rates. If this is set to
970 \fBpoisson\fR, fio will submit IO based on a more real world random request
971 flow, known as the Poisson process
972 (https://en.wikipedia.org/wiki/Poisson_process). The lambda will be
973 10^6 / IOPS for the given workload.
975 .BI rate_cycle \fR=\fPint
976 Average bandwidth for \fBrate\fR and \fBrate_min\fR over this number of
977 milliseconds. Default: 1000ms.
979 .BI latency_target \fR=\fPint
980 If set, fio will attempt to find the max performance point that the given
981 workload will run at while maintaining a latency below this target. The
982 values is given in microseconds. See \fBlatency_window\fR and
983 \fBlatency_percentile\fR.
985 .BI latency_window \fR=\fPint
986 Used with \fBlatency_target\fR to specify the sample window that the job
987 is run at varying queue depths to test the performance. The value is given
990 .BI latency_percentile \fR=\fPfloat
991 The percentage of IOs that must fall within the criteria specified by
992 \fBlatency_target\fR and \fBlatency_window\fR. If not set, this defaults
993 to 100.0, meaning that all IOs must be equal or below to the value set
994 by \fBlatency_target\fR.
996 .BI max_latency \fR=\fPint
997 If set, fio will exit the job if it exceeds this maximum latency. It will exit
1000 .BI cpumask \fR=\fPint
1001 Set CPU affinity for this job. \fIint\fR is a bitmask of allowed CPUs the job
1002 may run on. See \fBsched_setaffinity\fR\|(2).
1004 .BI cpus_allowed \fR=\fPstr
1005 Same as \fBcpumask\fR, but allows a comma-delimited list of CPU numbers.
1007 .BI cpus_allowed_policy \fR=\fPstr
1008 Set the policy of how fio distributes the CPUs specified by \fBcpus_allowed\fR
1009 or \fBcpumask\fR. Two policies are supported:
1014 All jobs will share the CPU set specified.
1017 Each job will get a unique CPU from the CPU set.
1020 \fBshared\fR is the default behaviour, if the option isn't specified. If
1021 \fBsplit\fR is specified, then fio will assign one cpu per job. If not enough
1022 CPUs are given for the jobs listed, then fio will roundrobin the CPUs in
1027 .BI numa_cpu_nodes \fR=\fPstr
1028 Set this job running on specified NUMA nodes' CPUs. The arguments allow
1029 comma delimited list of cpu numbers, A-B ranges, or 'all'.
1031 .BI numa_mem_policy \fR=\fPstr
1032 Set this job's memory policy and corresponding NUMA nodes. Format of
1036 .B <mode>[:<nodelist>]
1039 is one of the following memory policy:
1041 .B default, prefer, bind, interleave, local
1044 For \fBdefault\fR and \fBlocal\fR memory policy, no \fBnodelist\fR is
1045 needed to be specified. For \fBprefer\fR, only one node is
1046 allowed. For \fBbind\fR and \fBinterleave\fR, \fBnodelist\fR allows
1047 comma delimited list of numbers, A-B ranges, or 'all'.
1049 .BI startdelay \fR=\fPirange
1050 Delay start of job for the specified number of seconds. Supports all time
1051 suffixes to allow specification of hours, minutes, seconds and
1052 milliseconds - seconds are the default if a unit is omitted.
1053 Can be given as a range which causes each thread to choose randomly out of the
1056 .BI runtime \fR=\fPint
1057 Terminate processing after the specified number of seconds.
1060 If given, run for the specified \fBruntime\fR duration even if the files are
1061 completely read or written. The same workload will be repeated as many times
1062 as \fBruntime\fR allows.
1064 .BI ramp_time \fR=\fPint
1065 If set, fio will run the specified workload for this amount of time before
1066 logging any performance numbers. Useful for letting performance settle before
1067 logging results, thus minimizing the runtime required for stable results. Note
1068 that the \fBramp_time\fR is considered lead in time for a job, thus it will
1069 increase the total runtime if a special timeout or runtime is specified.
1071 .BI invalidate \fR=\fPbool
1072 Invalidate buffer-cache for the file prior to starting I/O. Default: true.
1074 .BI sync \fR=\fPbool
1075 Use synchronous I/O for buffered writes. For the majority of I/O engines,
1076 this means using O_SYNC. Default: false.
1078 .BI iomem \fR=\fPstr "\fR,\fP mem" \fR=\fPstr
1079 Allocation method for I/O unit buffer. Allowed values are:
1084 Allocate memory with \fBmalloc\fR\|(3).
1087 Use shared memory buffers allocated through \fBshmget\fR\|(2).
1090 Same as \fBshm\fR, but use huge pages as backing.
1093 Use \fBmmap\fR\|(2) for allocation. Uses anonymous memory unless a filename
1094 is given after the option in the format `:\fIfile\fR'.
1097 Same as \fBmmap\fR, but use huge files as backing.
1100 Same as \fBmmap\fR, but use a MMAP_SHARED mapping.
1103 The amount of memory allocated is the maximum allowed \fBblocksize\fR for the
1104 job multiplied by \fBiodepth\fR. For \fBshmhuge\fR or \fBmmaphuge\fR to work,
1105 the system must have free huge pages allocated. \fBmmaphuge\fR also needs to
1106 have hugetlbfs mounted, and \fIfile\fR must point there. At least on Linux,
1107 huge pages must be manually allocated. See \fB/proc/sys/vm/nr_hugehages\fR
1108 and the documentation for that. Normally you just need to echo an appropriate
1109 number, eg echoing 8 will ensure that the OS has 8 huge pages ready for
1113 .BI iomem_align \fR=\fPint "\fR,\fP mem_align" \fR=\fPint
1114 This indicates the memory alignment of the IO memory buffers. Note that the
1115 given alignment is applied to the first IO unit buffer, if using \fBiodepth\fR
1116 the alignment of the following buffers are given by the \fBbs\fR used. In
1117 other words, if using a \fBbs\fR that is a multiple of the page sized in the
1118 system, all buffers will be aligned to this value. If using a \fBbs\fR that
1119 is not page aligned, the alignment of subsequent IO memory buffers is the
1120 sum of the \fBiomem_align\fR and \fBbs\fR used.
1122 .BI hugepage\-size \fR=\fPint
1123 Defines the size of a huge page. Must be at least equal to the system setting.
1124 Should be a multiple of 1MB. Default: 4MB.
1127 Terminate all jobs when one finishes. Default: wait for each job to finish.
1129 .BI bwavgtime \fR=\fPint
1130 Average bandwidth calculations over the given time in milliseconds. Default:
1133 .BI iopsavgtime \fR=\fPint
1134 Average IOPS calculations over the given time in milliseconds. Default:
1137 .BI create_serialize \fR=\fPbool
1138 If true, serialize file creation for the jobs. Default: true.
1140 .BI create_fsync \fR=\fPbool
1141 \fBfsync\fR\|(2) data file after creation. Default: true.
1143 .BI create_on_open \fR=\fPbool
1144 If true, the files are not created until they are opened for IO by the job.
1146 .BI create_only \fR=\fPbool
1147 If true, fio will only run the setup phase of the job. If files need to be
1148 laid out or updated on disk, only that will be done. The actual job contents
1151 .BI allow_file_create \fR=\fPbool
1152 If true, fio is permitted to create files as part of its workload. This is
1153 the default behavior. If this option is false, then fio will error out if the
1154 files it needs to use don't already exist. Default: true.
1156 .BI allow_mounted_write \fR=\fPbool
1157 If this isn't set, fio will abort jobs that are destructive (eg that write)
1158 to what appears to be a mounted device or partition. This should help catch
1159 creating inadvertently destructive tests, not realizing that the test will
1160 destroy data on the mounted file system. Default: false.
1162 .BI pre_read \fR=\fPbool
1163 If this is given, files will be pre-read into memory before starting the given
1164 IO operation. This will also clear the \fR \fBinvalidate\fR flag, since it is
1165 pointless to pre-read and then drop the cache. This will only work for IO
1166 engines that are seekable, since they allow you to read the same data
1167 multiple times. Thus it will not work on eg network or splice IO.
1169 .BI unlink \fR=\fPbool
1170 Unlink job files when done. Default: false.
1172 .BI loops \fR=\fPint
1173 Specifies the number of iterations (runs of the same workload) of this job.
1176 .BI verify_only \fR=\fPbool
1177 Do not perform the specified workload, only verify data still matches previous
1178 invocation of this workload. This option allows one to check data multiple
1179 times at a later date without overwriting it. This option makes sense only for
1180 workloads that write data, and does not support workloads with the
1181 \fBtime_based\fR option set.
1183 .BI do_verify \fR=\fPbool
1184 Run the verify phase after a write phase. Only valid if \fBverify\fR is set.
1187 .BI verify \fR=\fPstr
1188 Method of verifying file contents after each iteration of the job. Each
1189 verification method also implies verification of special header, which is
1190 written to the beginning of each block. This header also includes meta
1191 information, like offset of the block, block number, timestamp when block
1192 was written, etc. \fBverify\fR=str can be combined with \fBverify_pattern\fR=str
1193 option. The allowed values are:
1197 .B md5 crc16 crc32 crc32c crc32c-intel crc64 crc7 sha256 sha512 sha1 xxhash
1198 Store appropriate checksum in the header of each block. crc32c-intel is
1199 hardware accelerated SSE4.2 driven, falls back to regular crc32c if
1200 not supported by the system.
1203 This option is deprecated, since now meta information is included in generic
1204 verification header and meta verification happens by default. For detailed
1205 information see the description of the \fBverify\fR=str setting. This option
1206 is kept because of compatibility's sake with old configurations. Do not use it.
1209 Verify a strict pattern. Normally fio includes a header with some basic
1210 information and checksumming, but if this option is set, only the
1211 specific pattern set with \fBverify_pattern\fR is verified.
1214 Pretend to verify. Used for testing internals.
1217 This option can be used for repeated burn-in tests of a system to make sure
1218 that the written data is also correctly read back. If the data direction given
1219 is a read or random read, fio will assume that it should verify a previously
1220 written file. If the data direction includes any form of write, the verify will
1221 be of the newly written data.
1224 .BI verifysort \fR=\fPbool
1225 If true, written verify blocks are sorted if \fBfio\fR deems it to be faster to
1226 read them back in a sorted manner. Default: true.
1228 .BI verifysort_nr \fR=\fPint
1229 Pre-load and sort verify blocks for a read workload.
1231 .BI verify_offset \fR=\fPint
1232 Swap the verification header with data somewhere else in the block before
1233 writing. It is swapped back before verifying.
1235 .BI verify_interval \fR=\fPint
1236 Write the verification header for this number of bytes, which should divide
1237 \fBblocksize\fR. Default: \fBblocksize\fR.
1239 .BI verify_pattern \fR=\fPstr
1240 If set, fio will fill the io buffers with this pattern. Fio defaults to filling
1241 with totally random bytes, but sometimes it's interesting to fill with a known
1242 pattern for io verification purposes. Depending on the width of the pattern,
1243 fio will fill 1/2/3/4 bytes of the buffer at the time(it can be either a
1244 decimal or a hex number). The verify_pattern if larger than a 32-bit quantity
1245 has to be a hex number that starts with either "0x" or "0X". Use with
1246 \fBverify\fP=str. Also, verify_pattern supports %o format, which means that for
1247 each block offset will be written and then verifyied back, e.g.:
1250 \fBverify_pattern\fR=%o
1252 Or use combination of everything:
1255 \fBverify_pattern\fR=0xff%o"abcd"-21
1259 .BI verify_fatal \fR=\fPbool
1260 If true, exit the job on the first observed verification failure. Default:
1263 .BI verify_dump \fR=\fPbool
1264 If set, dump the contents of both the original data block and the data block we
1265 read off disk to files. This allows later analysis to inspect just what kind of
1266 data corruption occurred. Off by default.
1268 .BI verify_async \fR=\fPint
1269 Fio will normally verify IO inline from the submitting thread. This option
1270 takes an integer describing how many async offload threads to create for IO
1271 verification instead, causing fio to offload the duty of verifying IO contents
1272 to one or more separate threads. If using this offload option, even sync IO
1273 engines can benefit from using an \fBiodepth\fR setting higher than 1, as it
1274 allows them to have IO in flight while verifies are running.
1276 .BI verify_async_cpus \fR=\fPstr
1277 Tell fio to set the given CPU affinity on the async IO verification threads.
1278 See \fBcpus_allowed\fP for the format used.
1280 .BI verify_backlog \fR=\fPint
1281 Fio will normally verify the written contents of a job that utilizes verify
1282 once that job has completed. In other words, everything is written then
1283 everything is read back and verified. You may want to verify continually
1284 instead for a variety of reasons. Fio stores the meta data associated with an
1285 IO block in memory, so for large verify workloads, quite a bit of memory would
1286 be used up holding this meta data. If this option is enabled, fio will write
1287 only N blocks before verifying these blocks.
1289 .BI verify_backlog_batch \fR=\fPint
1290 Control how many blocks fio will verify if verify_backlog is set. If not set,
1291 will default to the value of \fBverify_backlog\fR (meaning the entire queue is
1292 read back and verified). If \fBverify_backlog_batch\fR is less than
1293 \fBverify_backlog\fR then not all blocks will be verified, if
1294 \fBverify_backlog_batch\fR is larger than \fBverify_backlog\fR, some blocks
1295 will be verified more than once.
1297 .BI trim_percentage \fR=\fPint
1298 Number of verify blocks to discard/trim.
1300 .BI trim_verify_zero \fR=\fPbool
1301 Verify that trim/discarded blocks are returned as zeroes.
1303 .BI trim_backlog \fR=\fPint
1304 Trim after this number of blocks are written.
1306 .BI trim_backlog_batch \fR=\fPint
1307 Trim this number of IO blocks.
1309 .BI experimental_verify \fR=\fPbool
1310 Enable experimental verification.
1312 .BI verify_state_save \fR=\fPbool
1313 When a job exits during the write phase of a verify workload, save its
1314 current state. This allows fio to replay up until that point, if the
1315 verify state is loaded for the verify read phase.
1317 .BI verify_state_load \fR=\fPbool
1318 If a verify termination trigger was used, fio stores the current write
1319 state of each thread. This can be used at verification time so that fio
1320 knows how far it should verify. Without this information, fio will run
1321 a full verification pass, according to the settings in the job file used.
1323 .B stonewall "\fR,\fP wait_for_previous"
1324 Wait for preceding jobs in the job file to exit before starting this one.
1325 \fBstonewall\fR implies \fBnew_group\fR.
1328 Start a new reporting group. If not given, all jobs in a file will be part
1329 of the same reporting group, unless separated by a stonewall.
1331 .BI numjobs \fR=\fPint
1332 Number of clones (processes/threads performing the same workload) of this job.
1336 If set, display per-group reports instead of per-job when \fBnumjobs\fR is
1340 Use threads created with \fBpthread_create\fR\|(3) instead of processes created
1341 with \fBfork\fR\|(2).
1343 .BI zonesize \fR=\fPint
1344 Divide file into zones of the specified size in bytes. See \fBzoneskip\fR.
1346 .BI zonerange \fR=\fPint
1347 Give size of an IO zone. See \fBzoneskip\fR.
1349 .BI zoneskip \fR=\fPint
1350 Skip the specified number of bytes when \fBzonesize\fR bytes of data have been
1353 .BI write_iolog \fR=\fPstr
1354 Write the issued I/O patterns to the specified file. Specify a separate file
1355 for each job, otherwise the iologs will be interspersed and the file may be
1358 .BI read_iolog \fR=\fPstr
1359 Replay the I/O patterns contained in the specified file generated by
1360 \fBwrite_iolog\fR, or may be a \fBblktrace\fR binary file.
1362 .BI replay_no_stall \fR=\fPint
1363 While replaying I/O patterns using \fBread_iolog\fR the default behavior
1364 attempts to respect timing information between I/Os. Enabling
1365 \fBreplay_no_stall\fR causes I/Os to be replayed as fast as possible while
1366 still respecting ordering.
1368 .BI replay_redirect \fR=\fPstr
1369 While replaying I/O patterns using \fBread_iolog\fR the default behavior
1370 is to replay the IOPS onto the major/minor device that each IOP was recorded
1371 from. Setting \fBreplay_redirect\fR causes all IOPS to be replayed onto the
1372 single specified device regardless of the device it was recorded from.
1374 .BI replay_align \fR=\fPint
1375 Force alignment of IO offsets and lengths in a trace to this power of 2 value.
1377 .BI replay_scale \fR=\fPint
1378 Scale sector offsets down by this factor when replaying traces.
1380 .BI per_job_logs \fR=\fPbool
1381 If set, this generates bw/clat/iops log with per file private filenames. If
1382 not set, jobs with identical names will share the log filename. Default: true.
1384 .BI write_bw_log \fR=\fPstr
1385 If given, write a bandwidth log of the jobs in this job file. Can be used to
1386 store data of the bandwidth of the jobs in their lifetime. The included
1387 fio_generate_plots script uses gnuplot to turn these text files into nice
1388 graphs. See \fBwrite_lat_log\fR for behaviour of given filename. For this
1389 option, the postfix is _bw.x.log, where x is the index of the job (1..N,
1390 where N is the number of jobs). If \fBper_job_logs\fR is false, then the
1391 filename will not include the job index.
1393 .BI write_lat_log \fR=\fPstr
1394 Same as \fBwrite_bw_log\fR, but writes I/O completion latencies. If no
1395 filename is given with this option, the default filename of
1396 "jobname_type.x.log" is used, where x is the index of the job (1..N, where
1397 N is the number of jobs). Even if the filename is given, fio will still
1398 append the type of log. If \fBper_job_logs\fR is false, then the filename will
1399 not include the job index.
1401 .BI write_iops_log \fR=\fPstr
1402 Same as \fBwrite_bw_log\fR, but writes IOPS. If no filename is given with this
1403 option, the default filename of "jobname_type.x.log" is used, where x is the
1404 index of the job (1..N, where N is the number of jobs). Even if the filename
1405 is given, fio will still append the type of log. If \fBper_job_logs\fR is false,
1406 then the filename will not include the job index.
1408 .BI log_avg_msec \fR=\fPint
1409 By default, fio will log an entry in the iops, latency, or bw log for every
1410 IO that completes. When writing to the disk log, that can quickly grow to a
1411 very large size. Setting this option makes fio average the each log entry
1412 over the specified period of time, reducing the resolution of the log.
1415 .BI log_offset \fR=\fPbool
1416 If this is set, the iolog options will include the byte offset for the IO
1417 entry as well as the other data values.
1419 .BI log_compression \fR=\fPint
1420 If this is set, fio will compress the IO logs as it goes, to keep the memory
1421 footprint lower. When a log reaches the specified size, that chunk is removed
1422 and compressed in the background. Given that IO logs are fairly highly
1423 compressible, this yields a nice memory savings for longer runs. The downside
1424 is that the compression will consume some background CPU cycles, so it may
1425 impact the run. This, however, is also true if the logging ends up consuming
1426 most of the system memory. So pick your poison. The IO logs are saved
1427 normally at the end of a run, by decompressing the chunks and storing them
1428 in the specified log file. This feature depends on the availability of zlib.
1430 .BI log_store_compressed \fR=\fPbool
1431 If set, and \fBlog\fR_compression is also set, fio will store the log files in
1432 a compressed format. They can be decompressed with fio, using the
1433 \fB\-\-inflate-log\fR command line parameter. The files will be stored with a
1436 .BI block_error_percentiles \fR=\fPbool
1437 If set, record errors in trim block-sized units from writes and trims and output
1438 a histogram of how many trims it took to get to errors, and what kind of error
1441 .BI disable_lat \fR=\fPbool
1442 Disable measurements of total latency numbers. Useful only for cutting
1443 back the number of calls to \fBgettimeofday\fR\|(2), as that does impact performance at
1444 really high IOPS rates. Note that to really get rid of a large amount of these
1445 calls, this option must be used with disable_slat and disable_bw as well.
1447 .BI disable_clat \fR=\fPbool
1448 Disable measurements of completion latency numbers. See \fBdisable_lat\fR.
1450 .BI disable_slat \fR=\fPbool
1451 Disable measurements of submission latency numbers. See \fBdisable_lat\fR.
1453 .BI disable_bw_measurement \fR=\fPbool
1454 Disable measurements of throughput/bandwidth numbers. See \fBdisable_lat\fR.
1456 .BI lockmem \fR=\fPint
1457 Pin the specified amount of memory with \fBmlock\fR\|(2). Can be used to
1458 simulate a smaller amount of memory. The amount specified is per worker.
1460 .BI exec_prerun \fR=\fPstr
1461 Before running the job, execute the specified command with \fBsystem\fR\|(3).
1463 Output is redirected in a file called \fBjobname.prerun.txt\fR
1466 .BI exec_postrun \fR=\fPstr
1467 Same as \fBexec_prerun\fR, but the command is executed after the job completes.
1469 Output is redirected in a file called \fBjobname.postrun.txt\fR
1472 .BI ioscheduler \fR=\fPstr
1473 Attempt to switch the device hosting the file to the specified I/O scheduler.
1475 .BI disk_util \fR=\fPbool
1476 Generate disk utilization statistics if the platform supports it. Default: true.
1478 .BI clocksource \fR=\fPstr
1479 Use the given clocksource as the base of timing. The supported options are:
1483 \fBgettimeofday\fR\|(2)
1486 \fBclock_gettime\fR\|(2)
1489 Internal CPU clock source
1493 \fBcpu\fR is the preferred clocksource if it is reliable, as it is very fast
1494 (and fio is heavy on time calls). Fio will automatically use this clocksource
1495 if it's supported and considered reliable on the system it is running on,
1496 unless another clocksource is specifically set. For x86/x86-64 CPUs, this
1497 means supporting TSC Invariant.
1499 .BI gtod_reduce \fR=\fPbool
1500 Enable all of the \fBgettimeofday\fR\|(2) reducing options (disable_clat, disable_slat,
1501 disable_bw) plus reduce precision of the timeout somewhat to really shrink the
1502 \fBgettimeofday\fR\|(2) call count. With this option enabled, we only do about 0.4% of
1503 the gtod() calls we would have done if all time keeping was enabled.
1505 .BI gtod_cpu \fR=\fPint
1506 Sometimes it's cheaper to dedicate a single thread of execution to just getting
1507 the current time. Fio (and databases, for instance) are very intensive on
1508 \fBgettimeofday\fR\|(2) calls. With this option, you can set one CPU aside for doing
1509 nothing but logging current time to a shared memory location. Then the other
1510 threads/processes that run IO workloads need only copy that segment, instead of
1511 entering the kernel with a \fBgettimeofday\fR\|(2) call. The CPU set aside for doing
1512 these time calls will be excluded from other uses. Fio will manually clear it
1513 from the CPU mask of other jobs.
1515 .BI ignore_error \fR=\fPstr
1516 Sometimes you want to ignore some errors during test in that case you can specify
1517 error list for each error type.
1519 ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST
1521 errors for given error type is separated with ':'.
1522 Error may be symbol ('ENOSPC', 'ENOMEM') or an integer.
1524 Example: ignore_error=EAGAIN,ENOSPC:122 .
1526 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from WRITE.
1528 .BI error_dump \fR=\fPbool
1529 If set dump every error even if it is non fatal, true by default. If disabled
1530 only fatal error will be dumped
1532 .BI profile \fR=\fPstr
1533 Select a specific builtin performance test.
1535 .BI cgroup \fR=\fPstr
1536 Add job to this control group. If it doesn't exist, it will be created.
1537 The system must have a mounted cgroup blkio mount point for this to work. If
1538 your system doesn't have it mounted, you can do so with:
1540 # mount \-t cgroup \-o blkio none /cgroup
1542 .BI cgroup_weight \fR=\fPint
1543 Set the weight of the cgroup to this value. See the documentation that comes
1544 with the kernel, allowed values are in the range of 100..1000.
1546 .BI cgroup_nodelete \fR=\fPbool
1547 Normally fio will delete the cgroups it has created after the job completion.
1548 To override this behavior and to leave cgroups around after the job completion,
1549 set cgroup_nodelete=1. This can be useful if one wants to inspect various
1550 cgroup files after job completion. Default: false
1553 Instead of running as the invoking user, set the user ID to this value before
1554 the thread/process does any work.
1557 Set group ID, see \fBuid\fR.
1559 .BI unit_base \fR=\fPint
1560 Base unit for reporting. Allowed values are:
1564 Use auto-detection (default).
1574 .BI flow_id \fR=\fPint
1575 The ID of the flow. If not specified, it defaults to being a global flow. See
1579 Weight in token-based flow control. If this value is used, then there is a
1580 \fBflow counter\fR which is used to regulate the proportion of activity between
1581 two or more jobs. fio attempts to keep this flow counter near zero. The
1582 \fBflow\fR parameter stands for how much should be added or subtracted to the
1583 flow counter on each iteration of the main I/O loop. That is, if one job has
1584 \fBflow=8\fR and another job has \fBflow=-1\fR, then there will be a roughly
1585 1:8 ratio in how much one runs vs the other.
1587 .BI flow_watermark \fR=\fPint
1588 The maximum value that the absolute value of the flow counter is allowed to
1589 reach before the job must wait for a lower value of the counter.
1591 .BI flow_sleep \fR=\fPint
1592 The period of time, in microseconds, to wait after the flow watermark has been
1593 exceeded before retrying operations
1595 .BI clat_percentiles \fR=\fPbool
1596 Enable the reporting of percentiles of completion latencies.
1598 .BI percentile_list \fR=\fPfloat_list
1599 Overwrite the default list of percentiles for completion latencies and the
1600 block error histogram. Each number is a floating number in the range (0,100],
1601 and the maximum length of the list is 20. Use ':' to separate the
1602 numbers. For example, \-\-percentile_list=99.5:99.9 will cause fio to
1603 report the values of completion latency below which 99.5% and 99.9% of
1604 the observed latencies fell, respectively.
1605 .SS "Ioengine Parameters List"
1606 Some parameters are only valid when a specific ioengine is in use. These are
1607 used identically to normal parameters, with the caveat that when used on the
1608 command line, they must come after the ioengine.
1610 .BI (cpu)cpuload \fR=\fPint
1611 Attempt to use the specified percentage of CPU cycles.
1613 .BI (cpu)cpuchunks \fR=\fPint
1614 Split the load into cycles of the given time. In microseconds.
1616 .BI (cpu)exit_on_io_done \fR=\fPbool
1617 Detect when IO threads are done, then exit.
1619 .BI (libaio)userspace_reap
1620 Normally, with the libaio engine in use, fio will use
1621 the io_getevents system call to reap newly returned events.
1622 With this flag turned on, the AIO ring will be read directly
1623 from user-space to reap events. The reaping mode is only
1624 enabled when polling for a minimum of 0 events (eg when
1625 iodepth_batch_complete=0).
1627 .BI (net,netsplice)hostname \fR=\fPstr
1628 The host name or IP address to use for TCP or UDP based IO.
1629 If the job is a TCP listener or UDP reader, the hostname is not
1630 used and must be omitted unless it is a valid UDP multicast address.
1632 .BI (net,netsplice)port \fR=\fPint
1633 The TCP or UDP port to bind to or connect to. If this is used with
1634 \fBnumjobs\fR to spawn multiple instances of the same job type, then
1635 this will be the starting port number since fio will use a range of ports.
1637 .BI (net,netsplice)interface \fR=\fPstr
1638 The IP address of the network interface used to send or receive UDP multicast
1641 .BI (net,netsplice)ttl \fR=\fPint
1642 Time-to-live value for outgoing UDP multicast packets. Default: 1
1644 .BI (net,netsplice)nodelay \fR=\fPbool
1645 Set TCP_NODELAY on TCP connections.
1647 .BI (net,netsplice)protocol \fR=\fPstr "\fR,\fP proto" \fR=\fPstr
1648 The network protocol to use. Accepted values are:
1653 Transmission control protocol
1656 Transmission control protocol V6
1659 User datagram protocol
1662 User datagram protocol V6
1668 When the protocol is TCP or UDP, the port must also be given,
1669 as well as the hostname if the job is a TCP listener or UDP
1670 reader. For unix sockets, the normal filename option should be
1671 used and the port is invalid.
1674 .BI (net,netsplice)listen
1675 For TCP network connections, tell fio to listen for incoming
1676 connections rather than initiating an outgoing connection. The
1677 hostname must be omitted if this option is used.
1679 .BI (net, pingpong) \fR=\fPbool
1680 Normally a network writer will just continue writing data, and a network reader
1681 will just consume packets. If pingpong=1 is set, a writer will send its normal
1682 payload to the reader, then wait for the reader to send the same payload back.
1683 This allows fio to measure network latencies. The submission and completion
1684 latencies then measure local time spent sending or receiving, and the
1685 completion latency measures how long it took for the other end to receive and
1686 send back. For UDP multicast traffic pingpong=1 should only be set for a single
1687 reader when multiple readers are listening to the same address.
1689 .BI (net, window_size) \fR=\fPint
1690 Set the desired socket buffer size for the connection.
1692 .BI (net, mss) \fR=\fPint
1693 Set the TCP maximum segment size (TCP_MAXSEG).
1695 .BI (e4defrag,donorname) \fR=\fPstr
1696 File will be used as a block donor (swap extents between files)
1698 .BI (e4defrag,inplace) \fR=\fPint
1699 Configure donor file block allocation strategy
1702 Preallocate donor's file on init
1705 allocate space immediately inside defragment event, and free right after event
1708 .BI (rbd)rbdname \fR=\fPstr
1709 Specifies the name of the RBD.
1711 .BI (rbd)pool \fR=\fPstr
1712 Specifies the name of the Ceph pool containing the RBD.
1714 .BI (rbd)clientname \fR=\fPstr
1715 Specifies the username (without the 'client.' prefix) used to access the Ceph cluster.
1717 .BI (mtd)skipbad \fR=\fPbool
1718 Skip operations against known bad blocks.
1720 While running, \fBfio\fR will display the status of the created jobs. For
1724 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
1727 The characters in the first set of brackets denote the current status of each
1728 threads. The possible values are:
1734 Setup but not started.
1740 Initialized, waiting.
1743 Running, doing sequential reads.
1746 Running, doing random reads.
1749 Running, doing sequential writes.
1752 Running, doing random writes.
1755 Running, doing mixed sequential reads/writes.
1758 Running, doing mixed random reads/writes.
1761 Running, currently waiting for \fBfsync\fR\|(2).
1764 Running, verifying written data.
1767 Exited, not reaped by main thread.
1770 Exited, thread reaped.
1774 The second set of brackets shows the estimated completion percentage of
1775 the current group. The third set shows the read and write I/O rate,
1776 respectively. Finally, the estimated run time of the job is displayed.
1778 When \fBfio\fR completes (or is interrupted by Ctrl-C), it will show data
1779 for each thread, each group of threads, and each disk, in that order.
1781 Per-thread statistics first show the threads client number, group-id, and
1782 error code. The remaining figures are as follows:
1786 Number of megabytes of I/O performed.
1789 Average data rate (bandwidth).
1795 Submission latency minimum, maximum, average and standard deviation. This is
1796 the time it took to submit the I/O.
1799 Completion latency minimum, maximum, average and standard deviation. This
1800 is the time between submission and completion.
1803 Bandwidth minimum, maximum, percentage of aggregate bandwidth received, average
1804 and standard deviation.
1807 CPU usage statistics. Includes user and system time, number of context switches
1808 this thread went through and number of major and minor page faults.
1811 Distribution of I/O depths. Each depth includes everything less than (or equal)
1812 to it, but greater than the previous depth.
1815 Number of read/write requests issued, and number of short read/write requests.
1818 Distribution of I/O completion latencies. The numbers follow the same pattern
1822 The group statistics show:
1827 Number of megabytes I/O performed.
1830 Aggregate bandwidth of threads in the group.
1833 Minimum average bandwidth a thread saw.
1836 Maximum average bandwidth a thread saw.
1839 Shortest runtime of threads in the group.
1842 Longest runtime of threads in the group.
1846 Finally, disk statistics are printed with reads first:
1851 Number of I/Os performed by all groups.
1854 Number of merges in the I/O scheduler.
1857 Number of ticks we kept the disk busy.
1860 Total time spent in the disk queue.
1867 It is also possible to get fio to dump the current output while it is
1868 running, without terminating the job. To do that, send fio the \fBUSR1\fR
1871 If the \fB\-\-minimal\fR / \fB\-\-append-terse\fR options are given, the
1872 results will be printed/appended in a semicolon-delimited format suitable for
1874 A job description (if provided) follows on a new line. Note that the first
1875 number in the line is the version number. If the output has to be changed
1876 for some reason, this number will be incremented by 1 to signify that
1877 change. The fields are:
1880 .B terse version, fio version, jobname, groupid, error
1884 .B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, IOPS, runtime \fR(ms)\fP
1888 .B min, max, mean, standard deviation
1892 .B min, max, mean, standard deviation
1894 Completion latency percentiles (20 fields):
1896 .B Xth percentile=usec
1900 .B min, max, mean, standard deviation
1904 .B min, max, aggregate percentage of total, mean, standard deviation
1910 .B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, IOPS, runtime \fR(ms)\fP
1914 .B min, max, mean, standard deviation
1918 .B min, max, mean, standard deviation
1920 Completion latency percentiles (20 fields):
1922 .B Xth percentile=usec
1926 .B min, max, mean, standard deviation
1930 .B min, max, aggregate percentage of total, mean, standard deviation
1936 .B user, system, context switches, major page faults, minor page faults
1939 IO depth distribution:
1941 .B <=1, 2, 4, 8, 16, 32, >=64
1944 IO latency distribution:
1948 .B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
1952 .B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
1956 Disk utilization (1 for each disk used):
1958 .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
1961 Error Info (dependent on continue_on_error, default off):
1963 .B total # errors, first error code
1966 .B text description (if provided in config - appears on newline)
1969 Normally you would run fio as a stand-alone application on the machine
1970 where the IO workload should be generated. However, it is also possible to
1971 run the frontend and backend of fio separately. This makes it possible to
1972 have a fio server running on the machine(s) where the IO workload should
1973 be running, while controlling it from another machine.
1975 To start the server, you would do:
1977 \fBfio \-\-server=args\fR
1979 on that machine, where args defines what fio listens to. The arguments
1980 are of the form 'type:hostname or IP:port'. 'type' is either 'ip' (or ip4)
1981 for TCP/IP v4, 'ip6' for TCP/IP v6, or 'sock' for a local unix domain
1982 socket. 'hostname' is either a hostname or IP address, and 'port' is the port to
1983 listen to (only valid for TCP/IP, not a local socket). Some examples:
1987 Start a fio server, listening on all interfaces on the default port (8765).
1989 2) fio \-\-server=ip:hostname,4444
1991 Start a fio server, listening on IP belonging to hostname and on port 4444.
1993 3) fio \-\-server=ip6:::1,4444
1995 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
1997 4) fio \-\-server=,4444
1999 Start a fio server, listening on all interfaces on port 4444.
2001 5) fio \-\-server=1.2.3.4
2003 Start a fio server, listening on IP 1.2.3.4 on the default port.
2005 6) fio \-\-server=sock:/tmp/fio.sock
2007 Start a fio server, listening on the local socket /tmp/fio.sock.
2009 When a server is running, you can connect to it from a client. The client
2012 fio \-\-local-args \-\-client=server \-\-remote-args <job file(s)>
2014 where \-\-local-args are arguments that are local to the client where it is
2015 running, 'server' is the connect string, and \-\-remote-args and <job file(s)>
2016 are sent to the server. The 'server' string follows the same format as it
2017 does on the server side, to allow IP/hostname/socket and port strings.
2018 You can connect to multiple clients as well, to do that you could run:
2020 fio \-\-client=server2 \-\-client=server2 <job file(s)>
2022 If the job file is located on the fio server, then you can tell the server
2023 to load a local file as well. This is done by using \-\-remote-config:
2025 fio \-\-client=server \-\-remote-config /path/to/file.fio
2027 Then fio will open this local (to the server) job file instead
2028 of being passed one from the client.
2030 If you have many servers (example: 100 VMs/containers), you can input a pathname
2031 of a file containing host IPs/names as the parameter value for the \-\-client option.
2032 For example, here is an example "host.list" file containing 2 hostnames:
2034 host1.your.dns.domain
2036 host2.your.dns.domain
2038 The fio command would then be:
2040 fio \-\-client=host.list <job file>
2042 In this mode, you cannot input server-specific parameters or job files, and all
2043 servers receive the same job file.
2045 In order to enable fio \-\-client runs utilizing a shared filesystem from multiple hosts,
2046 fio \-\-client now prepends the IP address of the server to the filename. For example,
2047 if fio is using directory /mnt/nfs/fio and is writing filename fileio.tmp,
2048 with a \-\-client hostfile
2049 containing two hostnames h1 and h2 with IP addresses 192.168.10.120 and 192.168.10.121, then
2050 fio will create two files:
2052 /mnt/nfs/fio/192.168.10.120.fileio.tmp
2054 /mnt/nfs/fio/192.168.10.121.fileio.tmp
2059 was written by Jens Axboe <jens.axboe@oracle.com>,
2060 now Jens Axboe <axboe@fb.com>.
2062 This man page was written by Aaron Carroll <aaronc@cse.unsw.edu.au> based
2063 on documentation by Jens Axboe.
2064 .SH "REPORTING BUGS"
2065 Report bugs to the \fBfio\fR mailing list <fio@vger.kernel.org>.
2068 For further documentation see \fBHOWTO\fR and \fBREADME\fR.
2070 Sample jobfiles are available in the \fBexamples\fR directory.
projects / fio.git / blob