1 .TH fio 1 "September 2007" "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 \-\-output \fR=\fPfilename
16 Write output to \fIfilename\fR.
18 .BI \-\-timeout \fR=\fPtimeout
19 Limit run time to \fItimeout\fR seconds.
22 Generate per-job latency logs.
25 Generate per-job bandwidth logs.
28 Print statistics in a terse, semicolon-delimited format.
30 .BI \-\-showcmd \fR=\fPjobfile
31 Convert \fIjobfile\fR to a set of command-line options.
34 Enable read-only safety checks.
36 .BI \-\-eta \fR=\fPwhen
37 Specifies when real-time ETA estimate should be printed. \fIwhen\fR may
38 be one of `always', `never' or `auto'.
40 .BI \-\-section \fR=\fPsec
41 Only run section \fIsec\fR from job file.
43 .BI \-\-cmdhelp \fR=\fPcommand
44 Print help information for \fIcommand\fR. May be `all' for all commands.
46 .BI \-\-debug \fR=\fPtype
47 Enable verbose tracing of various fio actions. May be `all' for all types
48 or individual types separated by a comma (eg \-\-debug=io,file). `help' will
49 list all available tracing options.
52 Display usage information and exit.
55 Display version information and exit.
57 .B \-\-terse\-version\fR=\fPtype
58 Terse version output format
60 Job files are in `ini' format. They consist of one or more
61 job definitions, which begin with a job name in square brackets and
62 extend to the next job name. The job name can be any ASCII string
63 except `global', which has a special meaning. Following the job name is
64 a sequence of zero or more parameters, one per line, that define the
65 behavior of the job. Any line starting with a `;' or `#' character is
66 considered a comment and ignored.
68 If \fIjobfile\fR is specified as `-', the job file will be read from
71 The global section contains default parameters for jobs specified in the
72 job file. A job is only affected by global sections residing above it,
73 and there may be any number of global sections. Specific job definitions
74 may override any parameter set in global sections.
77 Some parameters may take arguments of a specific type. The types used are:
80 String: a sequence of alphanumeric characters.
83 SI integer: a whole number, possibly containing a suffix denoting the base unit
84 of the value. Accepted suffixes are `k', 'M', 'G', 'T', and 'P', denoting
85 kilo (1024), mega (1024^2), giga (1024^3), tera (1024^4), and peta (1024^5)
86 respectively. The suffix is not case sensitive. If prefixed with '0x', the
87 value is assumed to be base 16 (hexadecimal). A suffix may include a trailing 'b',
88 for instance 'kb' is identical to 'k'. You can specify a base 10 value
89 by using 'KiB', 'MiB', 'GiB', etc. This is useful for disk drives where
90 values are often given in base 10 values. Specifying '30GiB' will get you
94 Boolean: a true or false value. `0' denotes false, `1' denotes true.
97 Integer range: a range of integers specified in the format
98 \fIlower\fR:\fIupper\fR or \fIlower\fR\-\fIupper\fR. \fIlower\fR and
99 \fIupper\fR may contain a suffix as described above. If an option allows two
100 sets of ranges, they are separated with a `,' or `/' character. For example:
104 List of floating numbers: A list of floating numbers, separated by
109 May be used to override the job name. On the command line, this parameter
110 has the special purpose of signalling the start of a new job.
112 .BI description \fR=\fPstr
113 Human-readable description of the job. It is printed when the job is run, but
114 otherwise has no special purpose.
116 .BI directory \fR=\fPstr
117 Prefix filenames with this directory. Used to place files in a location other
120 .BI filename \fR=\fPstr
122 normally makes up a file name based on the job name, thread number, and file
123 number. If you want to share files between threads in a job or several jobs,
124 specify a \fIfilename\fR for each of them to override the default. If the I/O
125 engine used is `net', \fIfilename\fR is the host and port to connect to in the
126 format \fIhost\fR/\fIport\fR. If the I/O engine is file-based, you can specify
127 a number of files by separating the names with a `:' character. `\-' is a
128 reserved name, meaning stdin or stdout, depending on the read/write direction
131 .BI lockfile \fR=\fPstr
132 Fio defaults to not locking any files before it does IO to them. If a file or
133 file descriptor is shared, fio can serialize IO to that file to make the end
134 result consistent. This is usual for emulating real workloads that share files.
140 No locking. This is the default.
143 Only one thread or process may do IO at the time, excluding all others.
146 Read-write locking on the file. Many readers may access the file at the same
147 time, but writes get exclusive access.
150 The option may be post-fixed with a lock batch number. If set, then each
151 thread/process may do that amount of IOs to the file before giving up the lock.
152 Since lock acquisition is expensive, batching the lock/unlocks will speed up IO.
155 .BI opendir \fR=\fPstr
156 Recursively open any files below directory \fIstr\fR.
158 .BI readwrite \fR=\fPstr "\fR,\fP rw" \fR=\fPstr
159 Type of I/O pattern. Accepted values are:
176 Mixed sequential reads and writes.
179 Mixed random reads and writes.
182 For mixed I/O, the default split is 50/50. For certain types of io the result
183 may still be skewed a bit, since the speed may be different. It is possible to
184 specify a number of IO's to do before getting a new offset, this is one by
185 appending a `:\fI<nr>\fR to the end of the string given. For a random read, it
186 would look like \fBrw=randread:8\fR for passing in an offset modifier with a
187 value of 8. If the postfix is used with a sequential IO pattern, then the value
188 specified will be added to the generated offset for each IO. For instance,
189 using \fBrw=write:4k\fR will skip 4k for every write. It turns sequential IO
190 into sequential IO with holes. See the \fBrw_sequencer\fR option.
193 .BI rw_sequencer \fR=\fPstr
194 If an offset modifier is given by appending a number to the \fBrw=<str>\fR line,
195 then this option controls how that number modifies the IO offset being
196 generated. Accepted values are:
201 Generate sequential offset
204 Generate the same offset
207 \fBsequential\fR is only useful for random IO, where fio would normally
208 generate a new random offset for every IO. If you append eg 8 to randread, you
209 would get a new random offset for every 8 IO's. The result would be a seek for
210 only every 8 IO's, instead of for every IO. Use \fBrw=randread:8\fR to specify
211 that. As sequential IO is already sequential, setting \fBsequential\fR for that
212 would not result in any differences. \fBidentical\fR behaves in a similar
213 fashion, except it sends the same offset 8 number of times before generating a
218 .BI kb_base \fR=\fPint
219 The base unit for a kilobyte. The defacto base is 2^10, 1024. Storage
220 manufacturers like to use 10^3 or 1000 as a base ten unit instead, for obvious
221 reasons. Allow values are 1024 or 1000, with 1024 being the default.
223 .BI randrepeat \fR=\fPbool
224 Seed the random number generator in a predictable way so results are repeatable
225 across runs. Default: true.
227 .BI use_os_rand \fR=\fPbool
228 Fio can either use the random generator supplied by the OS to generator random
229 offsets, or it can use it's own internal generator (based on Tausworthe).
230 Default is to use the internal generator, which is often of better quality and
231 faster. Default: false.
233 .BI fallocate \fR=\fPstr
234 Whether pre-allocation is performed when laying down files. Accepted values
240 Do not pre-allocate space.
243 Pre-allocate via posix_fallocate().
246 Pre-allocate via fallocate() with FALLOC_FL_KEEP_SIZE set.
249 Backward-compatible alias for 'none'.
252 Backward-compatible alias for 'posix'.
255 May not be available on all supported platforms. 'keep' is only
256 available on Linux. If using ZFS on Solaris this must be set to 'none'
257 because ZFS doesn't support it. Default: 'posix'.
260 .BI fadvise_hint \fR=\fPbool
261 Disable use of \fIposix_fadvise\fR\|(2) to advise the kernel what I/O patterns
262 are likely to be issued. Default: true.
265 Total size of I/O for this job. \fBfio\fR will run until this many bytes have
266 been transfered, unless limited by other options (\fBruntime\fR, for instance).
267 Unless \fBnrfiles\fR and \fBfilesize\fR options are given, this amount will be
268 divided between the available files for the job. If not set, fio will use the
269 full size of the given files or devices. If the the files do not exist, size
270 must be given. It is also possible to give size as a percentage between 1 and
271 100. If size=20% is given, fio will use 20% of the full size of the given files
274 .BI fill_device \fR=\fPbool "\fR,\fB fill_fs" \fR=\fPbool
275 Sets size to something really large and waits for ENOSPC (no space left on
276 device) as the terminating condition. Only makes sense with sequential write.
277 For a read workload, the mount point will be filled first then IO started on
278 the result. This option doesn't make sense if operating on a raw device node,
279 since the size of that is already known by the file system. Additionally,
280 writing beyond end-of-device will not return ENOSPC there.
282 .BI filesize \fR=\fPirange
283 Individual file sizes. May be a range, in which case \fBfio\fR will select sizes
284 for files at random within the given range, limited to \fBsize\fR in total (if
285 that is given). If \fBfilesize\fR is not specified, each created file is the
288 .BI blocksize \fR=\fPint[,int] "\fR,\fB bs" \fR=\fPint[,int]
289 Block size for I/O units. Default: 4k. Values for reads and writes can be
290 specified separately in the format \fIread\fR,\fIwrite\fR, either of
291 which may be empty to leave that value at its default.
293 .BI blocksize_range \fR=\fPirange[,irange] "\fR,\fB bsrange" \fR=\fPirange[,irange]
294 Specify a range of I/O block sizes. The issued I/O unit will always be a
295 multiple of the minimum size, unless \fBblocksize_unaligned\fR is set. Applies
296 to both reads and writes if only one range is given, but can be specified
297 separately with a comma seperating the values. Example: bsrange=1k-4k,2k-8k.
298 Also (see \fBblocksize\fR).
300 .BI bssplit \fR=\fPstr
301 This option allows even finer grained control of the block sizes issued,
302 not just even splits between them. With this option, you can weight various
303 block sizes for exact control of the issued IO for a job that has mixed
304 block sizes. The format of the option is bssplit=blocksize/percentage,
305 optionally adding as many definitions as needed separated by a colon.
306 Example: bssplit=4k/10:64k/50:32k/40 would issue 50% 64k blocks, 10% 4k
307 blocks and 40% 32k blocks. \fBbssplit\fR also supports giving separate
308 splits to reads and writes. The format is identical to what the
309 \fBbs\fR option accepts, the read and write parts are separated with a
312 .B blocksize_unaligned\fR,\fP bs_unaligned
313 If set, any size in \fBblocksize_range\fR may be used. This typically won't
314 work with direct I/O, as that normally requires sector alignment.
316 .BI blockalign \fR=\fPint[,int] "\fR,\fB ba" \fR=\fPint[,int]
317 At what boundary to align random IO offsets. Defaults to the same as 'blocksize'
318 the minimum blocksize given. Minimum alignment is typically 512b
319 for using direct IO, though it usually depends on the hardware block size.
320 This option is mutually exclusive with using a random map for files, so it
321 will turn off that option.
324 Initialise buffers with all zeros. Default: fill buffers with random data.
327 If this option is given, fio will refill the IO buffers on every submit. The
328 default is to only fill it at init time and reuse that data. Only makes sense
329 if zero_buffers isn't specified, naturally. If data verification is enabled,
330 refill_buffers is also automatically enabled.
332 .BI scramble_buffers \fR=\fPbool
333 If \fBrefill_buffers\fR is too costly and the target is using data
334 deduplication, then setting this option will slightly modify the IO buffer
335 contents to defeat normal de-dupe attempts. This is not enough to defeat
336 more clever block compression attempts, but it will stop naive dedupe
337 of blocks. Default: true.
339 .BI nrfiles \fR=\fPint
340 Number of files to use for this job. Default: 1.
342 .BI openfiles \fR=\fPint
343 Number of files to keep open at the same time. Default: \fBnrfiles\fR.
345 .BI file_service_type \fR=\fPstr
346 Defines how files to service are selected. The following types are defined:
351 Choose a file at random
354 Round robin over open files (default).
356 Do each file in the set sequentially.
359 The number of I/Os to issue before switching a new file can be specified by
360 appending `:\fIint\fR' to the service type.
363 .BI ioengine \fR=\fPstr
364 Defines how the job issues I/O. The following types are defined:
369 Basic \fIread\fR\|(2) or \fIwrite\fR\|(2) I/O. \fIfseek\fR\|(2) is used to
370 position the I/O location.
373 Basic \fIpread\fR\|(2) or \fIpwrite\fR\|(2) I/O.
376 Basic \fIreadv\fR\|(2) or \fIwritev\fR\|(2) I/O. Will emulate queuing by
377 coalescing adjacents IOs into a single submission.
380 Linux native asynchronous I/O. This engine also has a sub-option,
381 \fBuserspace_reap\fR. To set it, use \fBioengine=libaio:userspace_reap\fR.
382 Normally, with the libaio engine in use, fio will use the
383 \fIio_getevents\fR\|(3) system call to reap newly returned events. With this
384 flag turned on, the AIO ring will be read directly from user-space to reap
385 events. The reaping mode is only enabled when polling for a minimum of \fB0\fR
386 events (eg when \fBiodepth_batch_complete=0\fR).
389 POSIX asynchronous I/O using \fIaio_read\fR\|(3) and \fIaio_write\fR\|(3).
392 Solaris native asynchronous I/O.
395 Windows native asynchronous I/O.
398 File is memory mapped with \fImmap\fR\|(2) and data copied using
402 \fIsplice\fR\|(2) is used to transfer the data and \fIvmsplice\fR\|(2) to
403 transfer data from user-space to the kernel.
406 Use the syslet system calls to make regular read/write asynchronous.
409 SCSI generic sg v3 I/O. May be either synchronous using the SG_IO ioctl, or if
410 the target is an sg character device, we use \fIread\fR\|(2) and
411 \fIwrite\fR\|(2) for asynchronous I/O.
414 Doesn't transfer any data, just pretends to. Mainly used to exercise \fBfio\fR
415 itself and for debugging and testing purposes.
418 Transfer over the network. \fBfilename\fR must be set appropriately to
419 `\fIhost\fR/\fIport\fR' regardless of data direction. If receiving, only the
420 \fIport\fR argument is used.
423 Like \fBnet\fR, but uses \fIsplice\fR\|(2) and \fIvmsplice\fR\|(2) to map data
427 Doesn't transfer any data, but burns CPU cycles according to \fBcpuload\fR and
428 \fBcpucycles\fR parameters.
431 The GUASI I/O engine is the Generic Userspace Asynchronous Syscall Interface
432 approach to asycnronous I/O.
434 See <http://www.xmailserver.org/guasi\-lib.html>.
437 The RDMA I/O engine supports both RDMA memory semantics (RDMA_WRITE/RDMA_READ)
438 and channel semantics (Send/Recv) for the InfiniBand, RoCE and iWARP protocols.
441 Loads an external I/O engine object file. Append the engine filename as
446 .BI iodepth \fR=\fPint
447 Number of I/O units to keep in flight against the file. Note that increasing
448 iodepth beyond 1 will not affect synchronous ioengines (except for small
449 degress when verify_async is in use). Even async engines my impose OS
450 restrictions causing the desired depth not to be achieved. This may happen on
451 Linux when using libaio and not setting \fBdirect\fR=1, since buffered IO is
452 not async on that OS. Keep an eye on the IO depth distribution in the
453 fio output to verify that the achieved depth is as expected. Default: 1.
455 .BI iodepth_batch \fR=\fPint
456 Number of I/Os to submit at once. Default: \fBiodepth\fR.
458 .BI iodepth_batch_complete \fR=\fPint
459 This defines how many pieces of IO to retrieve at once. It defaults to 1 which
460 means that we'll ask for a minimum of 1 IO in the retrieval process from the
461 kernel. The IO retrieval will go on until we hit the limit set by
462 \fBiodepth_low\fR. If this variable is set to 0, then fio will always check for
463 completed events before queuing more IO. This helps reduce IO latency, at the
464 cost of more retrieval system calls.
466 .BI iodepth_low \fR=\fPint
467 Low watermark indicating when to start filling the queue again. Default:
470 .BI direct \fR=\fPbool
471 If true, use non-buffered I/O (usually O_DIRECT). Default: false.
473 .BI buffered \fR=\fPbool
474 If true, use buffered I/O. This is the opposite of the \fBdirect\fR parameter.
477 .BI offset \fR=\fPint
478 Offset in the file to start I/O. Data before the offset will not be touched.
481 How many I/Os to perform before issuing an \fBfsync\fR\|(2) of dirty data. If
482 0, don't sync. Default: 0.
484 .BI fdatasync \fR=\fPint
485 Like \fBfsync\fR, but uses \fBfdatasync\fR\|(2) instead to only sync the
486 data parts of the file. Default: 0.
488 .BI sync_file_range \fR=\fPstr:int
489 Use sync_file_range() for every \fRval\fP number of write operations. Fio will
490 track range of writes that have happened since the last sync_file_range() call.
491 \fRstr\fP can currently be one or more of:
495 SYNC_FILE_RANGE_WAIT_BEFORE
498 SYNC_FILE_RANGE_WRITE
501 SYNC_FILE_RANGE_WRITE
505 So if you do sync_file_range=wait_before,write:8, fio would use
506 \fBSYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE\fP for every 8 writes.
507 Also see the sync_file_range(2) man page. This option is Linux specific.
509 .BI overwrite \fR=\fPbool
510 If writing, setup the file first and do overwrites. Default: false.
512 .BI end_fsync \fR=\fPbool
513 Sync file contents when job exits. Default: false.
515 .BI fsync_on_close \fR=\fPbool
516 If true, sync file contents on close. This differs from \fBend_fsync\fR in that
517 it will happen on every close, not just at the end of the job. Default: false.
519 .BI rwmixcycle \fR=\fPint
520 How many milliseconds before switching between reads and writes for a mixed
521 workload. Default: 500ms.
523 .BI rwmixread \fR=\fPint
524 Percentage of a mixed workload that should be reads. Default: 50.
526 .BI rwmixwrite \fR=\fPint
527 Percentage of a mixed workload that should be writes. If \fBrwmixread\fR and
528 \fBrwmixwrite\fR are given and do not sum to 100%, the latter of the two
529 overrides the first. This may interfere with a given rate setting, if fio is
530 asked to limit reads or writes to a certain rate. If that is the case, then
531 the distribution may be skewed. Default: 50.
534 Normally \fBfio\fR will cover every block of the file when doing random I/O. If
535 this parameter is given, a new offset will be chosen without looking at past
536 I/O history. This parameter is mutually exclusive with \fBverify\fR.
538 .BI softrandommap \fR=\fPbool
539 See \fBnorandommap\fR. If fio runs with the random block map enabled and it
540 fails to allocate the map, if this option is set it will continue without a
541 random block map. As coverage will not be as complete as with random maps, this
542 option is disabled by default.
545 Run job with given nice value. See \fInice\fR\|(2).
548 Set I/O priority value of this job between 0 (highest) and 7 (lowest). See
551 .BI prioclass \fR=\fPint
552 Set I/O priority class. See \fIionice\fR\|(1).
554 .BI thinktime \fR=\fPint
555 Stall job for given number of microseconds between issuing I/Os.
557 .BI thinktime_spin \fR=\fPint
558 Pretend to spend CPU time for given number of microseconds, sleeping the rest
559 of the time specified by \fBthinktime\fR. Only valid if \fBthinktime\fR is set.
561 .BI thinktime_blocks \fR=\fPint
562 Number of blocks to issue before waiting \fBthinktime\fR microseconds.
566 Cap bandwidth used by this job. The number is in bytes/sec, the normal postfix
567 rules apply. You can use \fBrate\fR=500k to limit reads and writes to 500k each,
568 or you can specify read and writes separately. Using \fBrate\fR=1m,500k would
569 limit reads to 1MB/sec and writes to 500KB/sec. Capping only reads or writes
570 can be done with \fBrate\fR=,500k or \fBrate\fR=500k,. The former will only
571 limit writes (to 500KB/sec), the latter will only limit reads.
573 .BI ratemin \fR=\fPint
574 Tell \fBfio\fR to do whatever it can to maintain at least the given bandwidth.
575 Failing to meet this requirement will cause the job to exit. The same format
576 as \fBrate\fR is used for read vs write separation.
578 .BI rate_iops \fR=\fPint
579 Cap the bandwidth to this number of IOPS. Basically the same as rate, just
580 specified independently of bandwidth. The same format as \fBrate\fR is used for
581 read vs write seperation. If \fBblocksize\fR is a range, the smallest block
582 size is used as the metric.
584 .BI rate_iops_min \fR=\fPint
585 If this rate of I/O is not met, the job will exit. The same format as \fBrate\fR
586 is used for read vs write seperation.
588 .BI ratecycle \fR=\fPint
589 Average bandwidth for \fBrate\fR and \fBratemin\fR over this number of
590 milliseconds. Default: 1000ms.
592 .BI cpumask \fR=\fPint
593 Set CPU affinity for this job. \fIint\fR is a bitmask of allowed CPUs the job
594 may run on. See \fBsched_setaffinity\fR\|(2).
596 .BI cpus_allowed \fR=\fPstr
597 Same as \fBcpumask\fR, but allows a comma-delimited list of CPU numbers.
599 .BI startdelay \fR=\fPint
600 Delay start of job for the specified number of seconds.
602 .BI runtime \fR=\fPint
603 Terminate processing after the specified number of seconds.
606 If given, run for the specified \fBruntime\fR duration even if the files are
607 completely read or written. The same workload will be repeated as many times
608 as \fBruntime\fR allows.
610 .BI ramp_time \fR=\fPint
611 If set, fio will run the specified workload for this amount of time before
612 logging any performance numbers. Useful for letting performance settle before
613 logging results, thus minimizing the runtime required for stable results. Note
614 that the \fBramp_time\fR is considered lead in time for a job, thus it will
615 increase the total runtime if a special timeout or runtime is specified.
617 .BI invalidate \fR=\fPbool
618 Invalidate buffer-cache for the file prior to starting I/O. Default: true.
621 Use synchronous I/O for buffered writes. For the majority of I/O engines,
622 this means using O_SYNC. Default: false.
624 .BI iomem \fR=\fPstr "\fR,\fP mem" \fR=\fPstr
625 Allocation method for I/O unit buffer. Allowed values are:
630 Allocate memory with \fImalloc\fR\|(3).
633 Use shared memory buffers allocated through \fIshmget\fR\|(2).
636 Same as \fBshm\fR, but use huge pages as backing.
639 Use \fImmap\fR\|(2) for allocation. Uses anonymous memory unless a filename
640 is given after the option in the format `:\fIfile\fR'.
643 Same as \fBmmap\fR, but use huge files as backing.
646 The amount of memory allocated is the maximum allowed \fBblocksize\fR for the
647 job multiplied by \fBiodepth\fR. For \fBshmhuge\fR or \fBmmaphuge\fR to work,
648 the system must have free huge pages allocated. \fBmmaphuge\fR also needs to
649 have hugetlbfs mounted, and \fIfile\fR must point there. At least on Linux,
650 huge pages must be manually allocated. See \fB/proc/sys/vm/nr_hugehages\fR
651 and the documentation for that. Normally you just need to echo an appropriate
652 number, eg echoing 8 will ensure that the OS has 8 huge pages ready for
656 .BI iomem_align \fR=\fPint "\fR,\fP mem_align" \fR=\fPint
657 This indiciates the memory alignment of the IO memory buffers. Note that the
658 given alignment is applied to the first IO unit buffer, if using \fBiodepth\fR
659 the alignment of the following buffers are given by the \fBbs\fR used. In
660 other words, if using a \fBbs\fR that is a multiple of the page sized in the
661 system, all buffers will be aligned to this value. If using a \fBbs\fR that
662 is not page aligned, the alignment of subsequent IO memory buffers is the
663 sum of the \fBiomem_align\fR and \fBbs\fR used.
665 .BI hugepage\-size \fR=\fPint
666 Defines the size of a huge page. Must be at least equal to the system setting.
667 Should be a multiple of 1MB. Default: 4MB.
670 Terminate all jobs when one finishes. Default: wait for each job to finish.
672 .BI bwavgtime \fR=\fPint
673 Average bandwidth calculations over the given time in milliseconds. Default:
676 .BI create_serialize \fR=\fPbool
677 If true, serialize file creation for the jobs. Default: true.
679 .BI create_fsync \fR=\fPbool
680 \fIfsync\fR\|(2) data file after creation. Default: true.
682 .BI create_on_open \fR=\fPbool
683 If true, the files are not created until they are opened for IO by the job.
685 .BI pre_read \fR=\fPbool
686 If this is given, files will be pre-read into memory before starting the given
687 IO operation. This will also clear the \fR \fBinvalidate\fR flag, since it is
688 pointless to pre-read and then drop the cache. This will only work for IO
689 engines that are seekable, since they allow you to read the same data
690 multiple times. Thus it will not work on eg network or splice IO.
692 .BI unlink \fR=\fPbool
693 Unlink job files when done. Default: false.
696 Specifies the number of iterations (runs of the same workload) of this job.
699 .BI do_verify \fR=\fPbool
700 Run the verify phase after a write phase. Only valid if \fBverify\fR is set.
703 .BI verify \fR=\fPstr
704 Method of verifying file contents after each iteration of the job. Allowed
709 .B md5 crc16 crc32 crc32c crc32c-intel crc64 crc7 sha256 sha512 sha1
710 Store appropriate checksum in the header of each block. crc32c-intel is
711 hardware accelerated SSE4.2 driven, falls back to regular crc32c if
712 not supported by the system.
715 Write extra information about each I/O (timestamp, block number, etc.). The
716 block number is verified. See \fBverify_pattern\fR as well.
719 Pretend to verify. Used for testing internals.
722 This option can be used for repeated burn-in tests of a system to make sure
723 that the written data is also correctly read back. If the data direction given
724 is a read or random read, fio will assume that it should verify a previously
725 written file. If the data direction includes any form of write, the verify will
726 be of the newly written data.
729 .BI verify_sort \fR=\fPbool
730 If true, written verify blocks are sorted if \fBfio\fR deems it to be faster to
731 read them back in a sorted manner. Default: true.
733 .BI verify_offset \fR=\fPint
734 Swap the verification header with data somewhere else in the block before
735 writing. It is swapped back before verifying.
737 .BI verify_interval \fR=\fPint
738 Write the verification header for this number of bytes, which should divide
739 \fBblocksize\fR. Default: \fBblocksize\fR.
741 .BI verify_pattern \fR=\fPstr
742 If set, fio will fill the io buffers with this pattern. Fio defaults to filling
743 with totally random bytes, but sometimes it's interesting to fill with a known
744 pattern for io verification purposes. Depending on the width of the pattern,
745 fio will fill 1/2/3/4 bytes of the buffer at the time(it can be either a
746 decimal or a hex number). The verify_pattern if larger than a 32-bit quantity
747 has to be a hex number that starts with either "0x" or "0X". Use with
750 .BI verify_fatal \fR=\fPbool
751 If true, exit the job on the first observed verification failure. Default:
754 .BI verify_dump \fR=\fPbool
755 If set, dump the contents of both the original data block and the data block we
756 read off disk to files. This allows later analysis to inspect just what kind of
757 data corruption occurred. On by default.
759 .BI verify_async \fR=\fPint
760 Fio will normally verify IO inline from the submitting thread. This option
761 takes an integer describing how many async offload threads to create for IO
762 verification instead, causing fio to offload the duty of verifying IO contents
763 to one or more separate threads. If using this offload option, even sync IO
764 engines can benefit from using an \fBiodepth\fR setting higher than 1, as it
765 allows them to have IO in flight while verifies are running.
767 .BI verify_async_cpus \fR=\fPstr
768 Tell fio to set the given CPU affinity on the async IO verification threads.
769 See \fBcpus_allowed\fP for the format used.
771 .BI verify_backlog \fR=\fPint
772 Fio will normally verify the written contents of a job that utilizes verify
773 once that job has completed. In other words, everything is written then
774 everything is read back and verified. You may want to verify continually
775 instead for a variety of reasons. Fio stores the meta data associated with an
776 IO block in memory, so for large verify workloads, quite a bit of memory would
777 be used up holding this meta data. If this option is enabled, fio will write
778 only N blocks before verifying these blocks.
780 .BI verify_backlog_batch \fR=\fPint
781 Control how many blocks fio will verify if verify_backlog is set. If not set,
782 will default to the value of \fBverify_backlog\fR (meaning the entire queue is
783 read back and verified). If \fBverify_backlog_batch\fR is less than
784 \fBverify_backlog\fR then not all blocks will be verified, if
785 \fBverify_backlog_batch\fR is larger than \fBverify_backlog\fR, some blocks
786 will be verified more than once.
788 .B stonewall "\fR,\fP wait_for_previous"
789 Wait for preceding jobs in the job file to exit before starting this one.
790 \fBstonewall\fR implies \fBnew_group\fR.
793 Start a new reporting group. If not given, all jobs in a file will be part
794 of the same reporting group, unless separated by a stonewall.
796 .BI numjobs \fR=\fPint
797 Number of clones (processes/threads performing the same workload) of this job.
801 If set, display per-group reports instead of per-job when \fBnumjobs\fR is
805 Use threads created with \fBpthread_create\fR\|(3) instead of processes created
806 with \fBfork\fR\|(2).
808 .BI zonesize \fR=\fPint
809 Divide file into zones of the specified size in bytes. See \fBzoneskip\fR.
811 .BI zoneskip \fR=\fPint
812 Skip the specified number of bytes when \fBzonesize\fR bytes of data have been
815 .BI write_iolog \fR=\fPstr
816 Write the issued I/O patterns to the specified file. Specify a separate file
817 for each job, otherwise the iologs will be interspersed and the file may be
820 .BI read_iolog \fR=\fPstr
821 Replay the I/O patterns contained in the specified file generated by
822 \fBwrite_iolog\fR, or may be a \fBblktrace\fR binary file.
824 .BI replay_no_stall \fR=\fPint
825 While replaying I/O patterns using \fBread_iolog\fR the default behavior
826 attempts to respect timing information between I/Os. Enabling
827 \fBreplay_no_stall\fR causes I/Os to be replayed as fast as possible while
828 still respecting ordering.
830 .BI replay_redirect \fR=\fPstr
831 While replaying I/O patterns using \fBread_iolog\fR the default behavior
832 is to replay the IOPS onto the major/minor device that each IOP was recorded
833 from. Setting \fBreplay_redirect\fR causes all IOPS to be replayed onto the
834 single specified device regardless of the device it was recorded from.
836 .BI write_bw_log \fR=\fPstr
837 If given, write a bandwidth log of the jobs in this job file. Can be used to
838 store data of the bandwidth of the jobs in their lifetime. The included
839 fio_generate_plots script uses gnuplot to turn these text files into nice
840 graphs. See \fBwrite_log_log\fR for behaviour of given filename. For this
841 option, the postfix is _bw.log.
843 .BI write_lat_log \fR=\fPstr
844 Same as \fBwrite_bw_log\fR, but writes I/O completion latencies. If no
845 filename is given with this option, the default filename of "jobname_type.log"
846 is used. Even if the filename is given, fio will still append the type of log.
848 .BI disable_lat \fR=\fPbool
849 Disable measurements of total latency numbers. Useful only for cutting
850 back the number of calls to gettimeofday, as that does impact performance at
851 really high IOPS rates. Note that to really get rid of a large amount of these
852 calls, this option must be used with disable_slat and disable_bw as well.
854 .BI disable_clat \fR=\fPbool
855 Disable measurements of completion latency numbers. See \fBdisable_lat\fR.
857 .BI disable_slat \fR=\fPbool
858 Disable measurements of submission latency numbers. See \fBdisable_lat\fR.
860 .BI disable_bw_measurement \fR=\fPbool
861 Disable measurements of throughput/bandwidth numbers. See \fBdisable_lat\fR.
863 .BI lockmem \fR=\fPint
864 Pin the specified amount of memory with \fBmlock\fR\|(2). Can be used to
865 simulate a smaller amount of memory.
867 .BI exec_prerun \fR=\fPstr
868 Before running the job, execute the specified command with \fBsystem\fR\|(3).
870 .BI exec_postrun \fR=\fPstr
871 Same as \fBexec_prerun\fR, but the command is executed after the job completes.
873 .BI ioscheduler \fR=\fPstr
874 Attempt to switch the device hosting the file to the specified I/O scheduler.
876 .BI cpuload \fR=\fPint
877 If the job is a CPU cycle-eater, attempt to use the specified percentage of
880 .BI cpuchunks \fR=\fPint
881 If the job is a CPU cycle-eater, split the load into cycles of the
882 given time in milliseconds.
884 .BI disk_util \fR=\fPbool
885 Generate disk utilization statistics if the platform supports it. Default: true.
887 .BI gtod_reduce \fR=\fPbool
888 Enable all of the gettimeofday() reducing options (disable_clat, disable_slat,
889 disable_bw) plus reduce precision of the timeout somewhat to really shrink the
890 gettimeofday() call count. With this option enabled, we only do about 0.4% of
891 the gtod() calls we would have done if all time keeping was enabled.
893 .BI gtod_cpu \fR=\fPint
894 Sometimes it's cheaper to dedicate a single thread of execution to just getting
895 the current time. Fio (and databases, for instance) are very intensive on
896 gettimeofday() calls. With this option, you can set one CPU aside for doing
897 nothing but logging current time to a shared memory location. Then the other
898 threads/processes that run IO workloads need only copy that segment, instead of
899 entering the kernel with a gettimeofday() call. The CPU set aside for doing
900 these time calls will be excluded from other uses. Fio will manually clear it
901 from the CPU mask of other jobs.
903 .BI cgroup \fR=\fPstr
904 Add job to this control group. If it doesn't exist, it will be created.
905 The system must have a mounted cgroup blkio mount point for this to work. If
906 your system doesn't have it mounted, you can do so with:
908 # mount \-t cgroup \-o blkio none /cgroup
910 .BI cgroup_weight \fR=\fPint
911 Set the weight of the cgroup to this value. See the documentation that comes
912 with the kernel, allowed values are in the range of 100..1000.
914 .BI cgroup_nodelete \fR=\fPbool
915 Normally fio will delete the cgroups it has created after the job completion.
916 To override this behavior and to leave cgroups around after the job completion,
917 set cgroup_nodelete=1. This can be useful if one wants to inspect various
918 cgroup files after job completion. Default: false
921 Instead of running as the invoking user, set the user ID to this value before
922 the thread/process does any work.
925 Set group ID, see \fBuid\fR.
927 .BI clat_percentiles \fR=\fPbool
928 Enable the reporting of percentiles of completion latencies.
930 .BI percentile_list \fR=\fPfloat_list
931 Overwrite the default list of percentiles for completion
932 latencies. Each number is a floating number in the range (0,100], and
933 the maximum length of the list is 20. Use ':' to separate the
934 numbers. For example, \-\-percentile_list=99.5:99.9 will cause fio to
935 report the values of completion latency below which 99.5% and 99.9% of
936 the observed latencies fell, respectively.
938 While running, \fBfio\fR will display the status of the created jobs. For
942 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
945 The characters in the first set of brackets denote the current status of each
946 threads. The possible values are:
952 Setup but not started.
958 Initialized, waiting.
961 Running, doing sequential reads.
964 Running, doing random reads.
967 Running, doing sequential writes.
970 Running, doing random writes.
973 Running, doing mixed sequential reads/writes.
976 Running, doing mixed random reads/writes.
979 Running, currently waiting for \fBfsync\fR\|(2).
982 Running, verifying written data.
985 Exited, not reaped by main thread.
988 Exited, thread reaped.
992 The second set of brackets shows the estimated completion percentage of
993 the current group. The third set shows the read and write I/O rate,
994 respectively. Finally, the estimated run time of the job is displayed.
996 When \fBfio\fR completes (or is interrupted by Ctrl-C), it will show data
997 for each thread, each group of threads, and each disk, in that order.
999 Per-thread statistics first show the threads client number, group-id, and
1000 error code. The remaining figures are as follows:
1004 Number of megabytes of I/O performed.
1007 Average data rate (bandwidth).
1013 Submission latency minimum, maximum, average and standard deviation. This is
1014 the time it took to submit the I/O.
1017 Completion latency minimum, maximum, average and standard deviation. This
1018 is the time between submission and completion.
1021 Bandwidth minimum, maximum, percentage of aggregate bandwidth received, average
1022 and standard deviation.
1025 CPU usage statistics. Includes user and system time, number of context switches
1026 this thread went through and number of major and minor page faults.
1029 Distribution of I/O depths. Each depth includes everything less than (or equal)
1030 to it, but greater than the previous depth.
1033 Number of read/write requests issued, and number of short read/write requests.
1036 Distribution of I/O completion latencies. The numbers follow the same pattern
1040 The group statistics show:
1045 Number of megabytes I/O performed.
1048 Aggregate bandwidth of threads in the group.
1051 Minimum average bandwidth a thread saw.
1054 Maximum average bandwidth a thread saw.
1057 Shortest runtime of threads in the group.
1060 Longest runtime of threads in the group.
1064 Finally, disk statistics are printed with reads first:
1069 Number of I/Os performed by all groups.
1072 Number of merges in the I/O scheduler.
1075 Number of ticks we kept the disk busy.
1078 Total time spent in the disk queue.
1085 If the \fB\-\-minimal\fR option is given, the results will be printed in a
1086 semicolon-delimited format suitable for scripted use - a job description
1087 (if provided) follows on a new line. Note that the first
1088 number in the line is the version number. If the output has to be changed
1089 for some reason, this number will be incremented by 1 to signify that
1090 change. The fields are:
1093 .B version, jobname, groupid, error
1097 .B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, runtime \fR(ms)\fP
1101 .B min, max, mean, standard deviation
1105 .B min, max, mean, standard deviation
1109 .B min, max, mean, standard deviation
1113 .B min, max, aggregate percentage of total, mean, standard deviation
1119 .B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, runtime \fR(ms)\fP
1123 .B min, max, mean, standard deviation
1127 .B min, max, mean, standard deviation
1131 .B min, max, mean, standard deviation
1135 .B min, max, aggregate percentage of total, mean, standard deviation
1141 .B user, system, context switches, major page faults, minor page faults
1144 IO depth distribution:
1146 .B <=1, 2, 4, 8, 16, 32, >=64
1149 IO latency distribution:
1153 .B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
1157 .B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
1161 Error Info (dependent on continue_on_error, default off):
1163 .B total # errors, first error code
1166 .B text description (if provided in config - appears on newline)
1170 was written by Jens Axboe <jens.axboe@oracle.com>,
1171 now Jens Axboe <jaxboe@fusionio.com>.
1173 This man page was written by Aaron Carroll <aaronc@cse.unsw.edu.au> based
1174 on documentation by Jens Axboe.
1175 .SH "REPORTING BUGS"
1176 Report bugs to the \fBfio\fR mailing list <fio@vger.kernel.org>.
1179 For further documentation see \fBHOWTO\fR and \fBREADME\fR.
1181 Sample jobfiles are available in the \fBexamples\fR directory.