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 seperated 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 Job files are in `ini' format. They consist of one or more
58 job definitions, which begin with a job name in square brackets and
59 extend to the next job name. The job name can be any ASCII string
60 except `global', which has a special meaning. Following the job name is
61 a sequence of zero or more parameters, one per line, that define the
62 behavior of the job. Any line starting with a `;' or `#' character is
63 considered a comment and ignored.
65 If \fIjobfile\fR is specified as `-', the job file will be read from
68 The global section contains default parameters for jobs specified in the
69 job file. A job is only affected by global sections residing above it,
70 and there may be any number of global sections. Specific job definitions
71 may override any parameter set in global sections.
74 Some parameters may take arguments of a specific type. The types used are:
77 String: a sequence of alphanumeric characters.
80 SI integer: a whole number, possibly containing a suffix denoting the base unit
81 of the value. Accepted suffixes are `k', 'M', 'G', 'T', and 'P', denoting
82 kilo (1024), mega (1024^2), giga (1024^3), tera (1024^4), and peta (1024^5)
83 respectively. The suffix is not case sensitive. If prefixed with '0x', the
84 value is assumed to be base 16 (hexadecimal). A suffix may include a trailing
85 'b', for instance 'kb' is identical to 'k'. You can specify a base 10 value
86 by using 'KiB', 'MiB', 'GiB', etc. This is useful for disk drives where
87 values are often given in base 10 values. Specifying '30GiB' will get you
91 Boolean: a true or false value. `0' denotes false, `1' denotes true.
94 Integer range: a range of integers specified in the format
95 \fIlower\fR:\fIupper\fR or \fIlower\fR\-\fIupper\fR. \fIlower\fR and
96 \fIupper\fR may contain a suffix as described above. If an option allows two
97 sets of ranges, they are separated with a `,' or `/' character. For example:
102 May be used to override the job name. On the command line, this parameter
103 has the special purpose of signalling the start of a new job.
105 .BI description \fR=\fPstr
106 Human-readable description of the job. It is printed when the job is run, but
107 otherwise has no special purpose.
109 .BI directory \fR=\fPstr
110 Prefix filenames with this directory. Used to place files in a location other
113 .BI filename \fR=\fPstr
115 normally makes up a file name based on the job name, thread number, and file
116 number. If you want to share files between threads in a job or several jobs,
117 specify a \fIfilename\fR for each of them to override the default. If the I/O
118 engine used is `net', \fIfilename\fR is the host and port to connect to in the
119 format \fIhost\fR/\fIport\fR. If the I/O engine is file-based, you can specify
120 a number of files by separating the names with a `:' character. `\-' is a
121 reserved name, meaning stdin or stdout, depending on the read/write direction
124 .BI lockfile \fR=\fPstr
125 Fio defaults to not locking any files before it does IO to them. If a file or
126 file descriptor is shared, fio can serialize IO to that file to make the end
127 result consistent. This is usual for emulating real workloads that share files.
133 No locking. This is the default.
136 Only one thread or process may do IO at the time, excluding all others.
139 Read-write locking on the file. Many readers may access the file at the same
140 time, but writes get exclusive access.
143 The option may be post-fixed with a lock batch number. If set, then each
144 thread/process may do that amount of IOs to the file before giving up the lock.
145 Since lock acquisition is expensive, batching the lock/unlocks will speed up IO.
148 .BI opendir \fR=\fPstr
149 Recursively open any files below directory \fIstr\fR.
151 .BI readwrite \fR=\fPstr "\fR,\fP rw" \fR=\fPstr
152 Type of I/O pattern. Accepted values are:
169 Mixed sequential reads and writes.
172 Mixed random reads and writes.
175 For mixed I/O, the default split is 50/50. For certain types of io the result
176 may still be skewed a bit, since the speed may be different. It is possible to
177 specify a number of IO's to do before getting a new offset, this is one by
178 appending a `:\fI<nr>\fR to the end of the string given. For a random read, it
179 would look like \fBrw=randread:8\fR for passing in an offset modifier with a
180 value of 8. See the \fBrw_sequencer\fR option.
183 .BI rw_sequencer \fR=\fPstr
184 If an offset modifier is given by appending a number to the \fBrw=<str>\fR line,
185 then this option controls how that number modifies the IO offset being
186 generated. Accepted values are:
191 Generate sequential offset
194 Generate the same offset
197 \fBsequential\fR is only useful for random IO, where fio would normally
198 generate a new random offset for every IO. If you append eg 8 to randread, you
199 would get a new random offset for every 8 IO's. The result would be a seek for
200 only every 8 IO's, instead of for every IO. Use \fBrw=randread:8\fR to specify
201 that. As sequential IO is already sequential, setting \fBsequential\fR for that
202 would not result in any differences. \fBidentical\fR behaves in a similar
203 fashion, except it sends the same offset 8 number of times before generating a
208 .BI kb_base \fR=\fPint
209 The base unit for a kilobyte. The defacto base is 2^10, 1024. Storage
210 manufacturers like to use 10^3 or 1000 as a base ten unit instead, for obvious
211 reasons. Allow values are 1024 or 1000, with 1024 being the default.
213 .BI randrepeat \fR=\fPbool
214 Seed the random number generator in a predictable way so results are repeatable
215 across runs. Default: true.
217 .BI fallocate \fR=\fPbool
218 By default, fio will use fallocate() to advise the system of the size of the
219 file we are going to write. This can be turned off with fallocate=0. May not
220 be available on all supported platforms.
222 .BI fadvise_hint \fR=\fPbool
223 Disable use of \fIposix_fadvise\fR\|(2) to advise the kernel what I/O patterns
224 are likely to be issued. Default: true.
227 Total size of I/O for this job. \fBfio\fR will run until this many bytes have
228 been transfered, unless limited by other options (\fBruntime\fR, for instance).
229 Unless \fBnrfiles\fR and \fBfilesize\fR options are given, this amount will be
230 divided between the available files for the job. If not set, fio will use the
231 full size of the given files or devices. If the the files do not exist, size
234 .BI fill_device \fR=\fPbool
235 Sets size to something really large and waits for ENOSPC (no space left on
236 device) as the terminating condition. Only makes sense with sequential write.
237 For a read workload, the mount point will be filled first then IO started on
240 .BI filesize \fR=\fPirange
241 Individual file sizes. May be a range, in which case \fBfio\fR will select sizes
242 for files at random within the given range, limited to \fBsize\fR in total (if
243 that is given). If \fBfilesize\fR is not specified, each created file is the
246 .BI blocksize \fR=\fPint[,int] "\fR,\fB bs" \fR=\fPint[,int]
247 Block size for I/O units. Default: 4k. Values for reads and writes can be
248 specified separately in the format \fIread\fR,\fIwrite\fR, either of
249 which may be empty to leave that value at its default.
251 .BI blocksize_range \fR=\fPirange[,irange] "\fR,\fB bsrange" \fR=\fPirange[,irange]
252 Specify a range of I/O block sizes. The issued I/O unit will always be a
253 multiple of the minimum size, unless \fBblocksize_unaligned\fR is set. Applies
254 to both reads and writes if only one range is given, but can be specified
255 separately with a comma seperating the values. Example: bsrange=1k-4k,2k-8k.
256 Also (see \fBblocksize\fR).
258 .BI bssplit \fR=\fPstr
259 This option allows even finer grained control of the block sizes issued,
260 not just even splits between them. With this option, you can weight various
261 block sizes for exact control of the issued IO for a job that has mixed
262 block sizes. The format of the option is bssplit=blocksize/percentage,
263 optionally adding as many definitions as needed seperated by a colon.
264 Example: bssplit=4k/10:64k/50:32k/40 would issue 50% 64k blocks, 10% 4k
265 blocks and 40% 32k blocks. \fBbssplit\fR also supports giving separate
266 splits to reads and writes. The format is identical to what the
267 \fBbs\fR option accepts, the read and write parts are separated with a
270 .B blocksize_unaligned\fR,\fP bs_unaligned
271 If set, any size in \fBblocksize_range\fR may be used. This typically won't
272 work with direct I/O, as that normally requires sector alignment.
274 .BI blockalign \fR=\fPint[,int] "\fR,\fB ba" \fR=\fPint[,int]
275 At what boundary to align random IO offsets. Defaults to the same as 'blocksize'
276 the minimum blocksize given. Minimum alignment is typically 512b
277 for using direct IO, though it usually depends on the hardware block size.
278 This option is mutually exclusive with using a random map for files, so it
279 will turn off that option.
282 Initialise buffers with all zeros. Default: fill buffers with random data.
285 If this option is given, fio will refill the IO buffers on every submit. The
286 default is to only fill it at init time and reuse that data. Only makes sense
287 if zero_buffers isn't specified, naturally. If data verification is enabled,
288 refill_buffers is also automatically enabled.
290 .BI nrfiles \fR=\fPint
291 Number of files to use for this job. Default: 1.
293 .BI openfiles \fR=\fPint
294 Number of files to keep open at the same time. Default: \fBnrfiles\fR.
296 .BI file_service_type \fR=\fPstr
297 Defines how files to service are selected. The following types are defined:
302 Choose a file at random
305 Round robin over open files (default).
307 Do each file in the set sequentially.
310 The number of I/Os to issue before switching a new file can be specified by
311 appending `:\fIint\fR' to the service type.
314 .BI ioengine \fR=\fPstr
315 Defines how the job issues I/O. The following types are defined:
320 Basic \fIread\fR\|(2) or \fIwrite\fR\|(2) I/O. \fIfseek\fR\|(2) is used to
321 position the I/O location.
324 Basic \fIpread\fR\|(2) or \fIpwrite\fR\|(2) I/O.
327 Basic \fIreadv\fR\|(2) or \fIwritev\fR\|(2) I/O. Will emulate queuing by
328 coalescing adjacents IOs into a single submission.
331 Linux native asynchronous I/O.
334 glibc POSIX asynchronous I/O using \fIaio_read\fR\|(3) and \fIaio_write\fR\|(3).
337 File is memory mapped with \fImmap\fR\|(2) and data copied using
341 \fIsplice\fR\|(2) is used to transfer the data and \fIvmsplice\fR\|(2) to
342 transfer data from user-space to the kernel.
345 Use the syslet system calls to make regular read/write asynchronous.
348 SCSI generic sg v3 I/O. May be either synchronous using the SG_IO ioctl, or if
349 the target is an sg character device, we use \fIread\fR\|(2) and
350 \fIwrite\fR\|(2) for asynchronous I/O.
353 Doesn't transfer any data, just pretends to. Mainly used to exercise \fBfio\fR
354 itself and for debugging and testing purposes.
357 Transfer over the network. \fBfilename\fR must be set appropriately to
358 `\fIhost\fR/\fIport\fR' regardless of data direction. If receiving, only the
359 \fIport\fR argument is used.
362 Like \fBnet\fR, but uses \fIsplice\fR\|(2) and \fIvmsplice\fR\|(2) to map data
366 Doesn't transfer any data, but burns CPU cycles according to \fBcpuload\fR and
367 \fBcpucycles\fR parameters.
370 The GUASI I/O engine is the Generic Userspace Asynchronous Syscall Interface
371 approach to asycnronous I/O.
373 See <http://www.xmailserver.org/guasi\-lib.html>.
376 Loads an external I/O engine object file. Append the engine filename as
381 .BI iodepth \fR=\fPint
382 Number of I/O units to keep in flight against the file. Note that increasing
383 iodepth beyond 1 will not affect synchronous ioengines (except for small
384 degress when verify_async is in use). Even async engines my impose OS
385 restrictions causing the desired depth not to be achieved. This may happen on
386 Linux when using libaio and not setting \fBdirect\fR=1, since buffered IO is
387 not async on that OS. Keep an eye on the IO depth distribution in the
388 fio output to verify that the achieved depth is as expected. Default: 1.
390 .BI iodepth_batch \fR=\fPint
391 Number of I/Os to submit at once. Default: \fBiodepth\fR.
393 .BI iodepth_batch_complete \fR=\fPint
394 This defines how many pieces of IO to retrieve at once. It defaults to 1 which
395 means that we'll ask for a minimum of 1 IO in the retrieval process from the
396 kernel. The IO retrieval will go on until we hit the limit set by
397 \fBiodepth_low\fR. If this variable is set to 0, then fio will always check for
398 completed events before queuing more IO. This helps reduce IO latency, at the
399 cost of more retrieval system calls.
401 .BI iodepth_low \fR=\fPint
402 Low watermark indicating when to start filling the queue again. Default:
405 .BI direct \fR=\fPbool
406 If true, use non-buffered I/O (usually O_DIRECT). Default: false.
408 .BI buffered \fR=\fPbool
409 If true, use buffered I/O. This is the opposite of the \fBdirect\fR parameter.
412 .BI offset \fR=\fPint
413 Offset in the file to start I/O. Data before the offset will not be touched.
416 How many I/Os to perform before issuing an \fBfsync\fR\|(2) of dirty data. If
417 0, don't sync. Default: 0.
419 .BI fdatasync \fR=\fPint
420 Like \fBfsync\fR, but uses \fBfdatasync\fR\|(2) instead to only sync the
421 data parts of the file. Default: 0.
423 .BI sync_file_range \fR=\fPstr:int
424 Use sync_file_range() for every \fRval\fP number of write operations. Fio will
425 track range of writes that have happened since the last sync_file_range() call.
426 \fRstr\fP can currently be one or more of:
430 SYNC_FILE_RANGE_WAIT_BEFORE
433 SYNC_FILE_RANGE_WRITE
436 SYNC_FILE_RANGE_WRITE
440 So if you do sync_file_range=wait_before,write:8, fio would use
441 \fBSYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE\fP for every 8 writes.
442 Also see the sync_file_range(2) man page. This option is Linux specific.
444 .BI overwrite \fR=\fPbool
445 If writing, setup the file first and do overwrites. Default: false.
447 .BI end_fsync \fR=\fPbool
448 Sync file contents when job exits. Default: false.
450 .BI fsync_on_close \fR=\fPbool
451 If true, sync file contents on close. This differs from \fBend_fsync\fR in that
452 it will happen on every close, not just at the end of the job. Default: false.
454 .BI rwmixcycle \fR=\fPint
455 How many milliseconds before switching between reads and writes for a mixed
456 workload. Default: 500ms.
458 .BI rwmixread \fR=\fPint
459 Percentage of a mixed workload that should be reads. Default: 50.
461 .BI rwmixwrite \fR=\fPint
462 Percentage of a mixed workload that should be writes. If \fBrwmixread\fR and
463 \fBrwmixwrite\fR are given and do not sum to 100%, the latter of the two
464 overrides the first. This may interfere with a given rate setting, if fio is
465 asked to limit reads or writes to a certain rate. If that is the case, then
466 the distribution may be skewed. Default: 50.
469 Normally \fBfio\fR will cover every block of the file when doing random I/O. If
470 this parameter is given, a new offset will be chosen without looking at past
471 I/O history. This parameter is mutually exclusive with \fBverify\fR.
474 See \fBnorandommap\fR. If fio runs with the random block map enabled and it
475 fails to allocate the map, if this option is set it will continue without a
476 random block map. As coverage will not be as complete as with random maps, this
477 option is disabled by default.
480 Run job with given nice value. See \fInice\fR\|(2).
483 Set I/O priority value of this job between 0 (highest) and 7 (lowest). See
486 .BI prioclass \fR=\fPint
487 Set I/O priority class. See \fIionice\fR\|(1).
489 .BI thinktime \fR=\fPint
490 Stall job for given number of microseconds between issuing I/Os.
492 .BI thinktime_spin \fR=\fPint
493 Pretend to spend CPU time for given number of microseconds, sleeping the rest
494 of the time specified by \fBthinktime\fR. Only valid if \fBthinktime\fR is set.
496 .BI thinktime_blocks \fR=\fPint
497 Number of blocks to issue before waiting \fBthinktime\fR microseconds.
501 Cap bandwidth used by this job. The number is in bytes/sec, the normal postfix
502 rules apply. You can use \fBrate\fR=500k to limit reads and writes to 500k each,
503 or you can specify read and writes separately. Using \fBrate\fR=1m,500k would
504 limit reads to 1MB/sec and writes to 500KB/sec. Capping only reads or writes
505 can be done with \fBrate\fR=,500k or \fBrate\fR=500k,. The former will only
506 limit writes (to 500KB/sec), the latter will only limit reads.
508 .BI ratemin \fR=\fPint
509 Tell \fBfio\fR to do whatever it can to maintain at least the given bandwidth.
510 Failing to meet this requirement will cause the job to exit. The same format
511 as \fBrate\fR is used for read vs write separation.
513 .BI rate_iops \fR=\fPint
514 Cap the bandwidth to this number of IOPS. Basically the same as rate, just
515 specified independently of bandwidth. The same format as \fBrate\fR is used for
516 read vs write seperation. If \fBblocksize\fR is a range, the smallest block
517 size is used as the metric.
519 .BI rate_iops_min \fR=\fPint
520 If this rate of I/O is not met, the job will exit. The same format as \fBrate\fR
521 is used for read vs write seperation.
523 .BI ratecycle \fR=\fPint
524 Average bandwidth for \fBrate\fR and \fBratemin\fR over this number of
525 milliseconds. Default: 1000ms.
527 .BI cpumask \fR=\fPint
528 Set CPU affinity for this job. \fIint\fR is a bitmask of allowed CPUs the job
529 may run on. See \fBsched_setaffinity\fR\|(2).
531 .BI cpus_allowed \fR=\fPstr
532 Same as \fBcpumask\fR, but allows a comma-delimited list of CPU numbers.
534 .BI startdelay \fR=\fPint
535 Delay start of job for the specified number of seconds.
537 .BI runtime \fR=\fPint
538 Terminate processing after the specified number of seconds.
541 If given, run for the specified \fBruntime\fR duration even if the files are
542 completely read or written. The same workload will be repeated as many times
543 as \fBruntime\fR allows.
545 .BI ramp_time \fR=\fPint
546 If set, fio will run the specified workload for this amount of time before
547 logging any performance numbers. Useful for letting performance settle before
548 logging results, thus minimizing the runtime required for stable results. Note
549 that the \fBramp_time\fR is considered lead in time for a job, thus it will
550 increase the total runtime if a special timeout or runtime is specified.
552 .BI invalidate \fR=\fPbool
553 Invalidate buffer-cache for the file prior to starting I/O. Default: true.
556 Use synchronous I/O for buffered writes. For the majority of I/O engines,
557 this means using O_SYNC. Default: false.
559 .BI iomem \fR=\fPstr "\fR,\fP mem" \fR=\fPstr
560 Allocation method for I/O unit buffer. Allowed values are:
565 Allocate memory with \fImalloc\fR\|(3).
568 Use shared memory buffers allocated through \fIshmget\fR\|(2).
571 Same as \fBshm\fR, but use huge pages as backing.
574 Use \fImmap\fR\|(2) for allocation. Uses anonymous memory unless a filename
575 is given after the option in the format `:\fIfile\fR'.
578 Same as \fBmmap\fR, but use huge files as backing.
581 The amount of memory allocated is the maximum allowed \fBblocksize\fR for the
582 job multiplied by \fBiodepth\fR. For \fBshmhuge\fR or \fBmmaphuge\fR to work,
583 the system must have free huge pages allocated. \fBmmaphuge\fR also needs to
584 have hugetlbfs mounted, and \fIfile\fR must point there. At least on Linux,
585 huge pages must be manually allocated. See \fB/proc/sys/vm/nr_hugehages\fR
586 and the documentation for that. Normally you just need to echo an appropriate
587 number, eg echoing 8 will ensure that the OS has 8 huge pages ready for
591 .BI iomem_align \fR=\fPint
592 This indiciates the memory alignment of the IO memory buffers. Note that the
593 given alignment is applied to the first IO unit buffer, if using \fBiodepth\fR
594 the alignment of the following buffers are given by the \fBbs\fR used. In
595 other words, if using a \fBbs\fR that is a multiple of the page sized in the
596 system, all buffers will be aligned to this value. If using a \fBbs\fR that
597 is not page aligned, the alignment of subsequent IO memory buffers is the
598 sum of the \fBiomem_align\fR and \fBbs\fR used.
600 .BI hugepage\-size \fR=\fPint
601 Defines the size of a huge page. Must be at least equal to the system setting.
602 Should be a multiple of 1MB. Default: 4MB.
605 Terminate all jobs when one finishes. Default: wait for each job to finish.
607 .BI bwavgtime \fR=\fPint
608 Average bandwidth calculations over the given time in milliseconds. Default:
611 .BI create_serialize \fR=\fPbool
612 If true, serialize file creation for the jobs. Default: true.
614 .BI create_fsync \fR=\fPbool
615 \fIfsync\fR\|(2) data file after creation. Default: true.
617 .BI create_on_open \fR=\fPbool
618 If true, the files are not created until they are opened for IO by the job.
620 .BI pre_read \fR=\fPbool
621 If this is given, files will be pre-read into memory before starting the given
622 IO operation. This will also clear the \fR \fBinvalidate\fR flag, since it is
623 pointless to pre-read and then drop the cache. This will only work for IO
624 engines that are seekable, since they allow you to read the same data
625 multiple times. Thus it will not work on eg network or splice IO.
627 .BI unlink \fR=\fPbool
628 Unlink job files when done. Default: false.
631 Specifies the number of iterations (runs of the same workload) of this job.
634 .BI do_verify \fR=\fPbool
635 Run the verify phase after a write phase. Only valid if \fBverify\fR is set.
638 .BI verify \fR=\fPstr
639 Method of verifying file contents after each iteration of the job. Allowed
644 .B md5 crc16 crc32 crc32c crc32c-intel crc64 crc7 sha256 sha512 sha1
645 Store appropriate checksum in the header of each block. crc32c-intel is
646 hardware accelerated SSE4.2 driven, falls back to regular crc32c if
647 not supported by the system.
650 Write extra information about each I/O (timestamp, block number, etc.). The
651 block number is verified. See \fBverify_pattern\fR as well.
654 Pretend to verify. Used for testing internals.
657 This option can be used for repeated burn-in tests of a system to make sure
658 that the written data is also correctly read back. If the data direction given
659 is a read or random read, fio will assume that it should verify a previously
660 written file. If the data direction includes any form of write, the verify will
661 be of the newly written data.
664 .BI verify_sort \fR=\fPbool
665 If true, written verify blocks are sorted if \fBfio\fR deems it to be faster to
666 read them back in a sorted manner. Default: true.
668 .BI verify_offset \fR=\fPint
669 Swap the verification header with data somewhere else in the block before
670 writing. It is swapped back before verifying.
672 .BI verify_interval \fR=\fPint
673 Write the verification header for this number of bytes, which should divide
674 \fBblocksize\fR. Default: \fBblocksize\fR.
676 .BI verify_pattern \fR=\fPstr
677 If set, fio will fill the io buffers with this pattern. Fio defaults to filling
678 with totally random bytes, but sometimes it's interesting to fill with a known
679 pattern for io verification purposes. Depending on the width of the pattern,
680 fio will fill 1/2/3/4 bytes of the buffer at the time(it can be either a
681 decimal or a hex number). The verify_pattern if larger than a 32-bit quantity
682 has to be a hex number that starts with either "0x" or "0X". Use with
685 .BI verify_fatal \fR=\fPbool
686 If true, exit the job on the first observed verification failure. Default:
689 .BI verify_async \fR=\fPint
690 Fio will normally verify IO inline from the submitting thread. This option
691 takes an integer describing how many async offload threads to create for IO
692 verification instead, causing fio to offload the duty of verifying IO contents
693 to one or more separate threads. If using this offload option, even sync IO
694 engines can benefit from using an \fBiodepth\fR setting higher than 1, as it
695 allows them to have IO in flight while verifies are running.
697 .BI verify_async_cpus \fR=\fPstr
698 Tell fio to set the given CPU affinity on the async IO verification threads.
699 See \fBcpus_allowed\fP for the format used.
701 .BI verify_backlog \fR=\fPint
702 Fio will normally verify the written contents of a job that utilizes verify
703 once that job has completed. In other words, everything is written then
704 everything is read back and verified. You may want to verify continually
705 instead for a variety of reasons. Fio stores the meta data associated with an
706 IO block in memory, so for large verify workloads, quite a bit of memory would
707 be used up holding this meta data. If this option is enabled, fio will write
708 only N blocks before verifying these blocks.
710 .BI verify_backlog_batch \fR=\fPint
711 Control how many blocks fio will verify if verify_backlog is set. If not set,
712 will default to the value of \fBverify_backlog\fR (meaning the entire queue is
713 read back and verified). If \fBverify_backlog_batch\fR is less than
714 \fBverify_backlog\fR then not all blocks will be verified, if
715 \fBverify_backlog_batch\fR is larger than \fBverify_backlog\fR, some blocks
716 will be verified more than once.
719 Wait for preceeding jobs in the job file to exit before starting this one.
720 \fBstonewall\fR implies \fBnew_group\fR.
723 Start a new reporting group. If not given, all jobs in a file will be part
724 of the same reporting group, unless separated by a stonewall.
726 .BI numjobs \fR=\fPint
727 Number of clones (processes/threads performing the same workload) of this job.
731 If set, display per-group reports instead of per-job when \fBnumjobs\fR is
735 Use threads created with \fBpthread_create\fR\|(3) instead of processes created
736 with \fBfork\fR\|(2).
738 .BI zonesize \fR=\fPint
739 Divide file into zones of the specified size in bytes. See \fBzoneskip\fR.
741 .BI zoneskip \fR=\fPint
742 Skip the specified number of bytes when \fBzonesize\fR bytes of data have been
745 .BI write_iolog \fR=\fPstr
746 Write the issued I/O patterns to the specified file.
748 .BI read_iolog \fR=\fPstr
749 Replay the I/O patterns contained in the specified file generated by
750 \fBwrite_iolog\fR, or may be a \fBblktrace\fR binary file.
752 .BI replay_no_stall \fR=\fPint
753 While replaying I/O patterns using \fBread_iolog\fR the default behavior
754 attempts to respect timing information between I/Os. Enabling
755 \fBreplay_no_stall\fR causes I/Os to be replayed as fast as possible while
756 still respecting ordering.
758 .BI replay_redirect \fR=\fPstr
759 While replaying I/O patterns using \fBread_iolog\fR the default behavior
760 is to replay the IOPS onto the major/minor device that each IOP was recorded
761 from. Setting \fBreplay_redirect\fR causes all IOPS to be replayed onto the
762 single specified device regardless of the device it was recorded from.
764 .B write_bw_log \fR=\fPstr
765 If given, write a bandwidth log of the jobs in this job file. Can be used to
766 store data of the bandwidth of the jobs in their lifetime. The included
767 fio_generate_plots script uses gnuplot to turn these text files into nice
768 graphs. See \fBwrite_log_log\fR for behaviour of given filename. For this
769 option, the postfix is _bw.log.
771 .B write_lat_log \fR=\fPstr
772 Same as \fBwrite_bw_log\fR, but writes I/O completion latencies. If no
773 filename is given with this option, the default filename of "jobname_type.log"
774 is used. Even if the filename is given, fio will still append the type of log.
776 .B disable_lat \fR=\fPbool
777 Disable measurements of total latency numbers. Useful only for cutting
778 back the number of calls to gettimeofday, as that does impact performance at
779 really high IOPS rates. Note that to really get rid of a large amount of these
780 calls, this option must be used with disable_slat and disable_bw as well.
782 .B disable_clat \fR=\fPbool
783 Disable measurements of submission latency numbers. See \fBdisable_lat\fR.
785 .B disable_slat \fR=\fPbool
786 Disable measurements of submission latency numbers. See \fBdisable_lat\fR.
788 .B disable_bw_measurement \fR=\fPbool
789 Disable measurements of throughput/bandwidth numbers. See \fBdisable_lat\fR.
791 .BI lockmem \fR=\fPint
792 Pin the specified amount of memory with \fBmlock\fR\|(2). Can be used to
793 simulate a smaller amount of memory.
795 .BI exec_prerun \fR=\fPstr
796 Before running the job, execute the specified command with \fBsystem\fR\|(3).
798 .BI exec_postrun \fR=\fPstr
799 Same as \fBexec_prerun\fR, but the command is executed after the job completes.
801 .BI ioscheduler \fR=\fPstr
802 Attempt to switch the device hosting the file to the specified I/O scheduler.
804 .BI cpuload \fR=\fPint
805 If the job is a CPU cycle-eater, attempt to use the specified percentage of
808 .BI cpuchunks \fR=\fPint
809 If the job is a CPU cycle-eater, split the load into cycles of the
810 given time in milliseconds.
812 .BI disk_util \fR=\fPbool
813 Generate disk utilization statistics if the platform supports it. Default: true.
815 .BI gtod_reduce \fR=\fPbool
816 Enable all of the gettimeofday() reducing options (disable_clat, disable_slat,
817 disable_bw) plus reduce precision of the timeout somewhat to really shrink the
818 gettimeofday() call count. With this option enabled, we only do about 0.4% of
819 the gtod() calls we would have done if all time keeping was enabled.
821 .BI gtod_cpu \fR=\fPint
822 Sometimes it's cheaper to dedicate a single thread of execution to just getting
823 the current time. Fio (and databases, for instance) are very intensive on
824 gettimeofday() calls. With this option, you can set one CPU aside for doing
825 nothing but logging current time to a shared memory location. Then the other
826 threads/processes that run IO workloads need only copy that segment, instead of
827 entering the kernel with a gettimeofday() call. The CPU set aside for doing
828 these time calls will be excluded from other uses. Fio will manually clear it
829 from the CPU mask of other jobs.
831 .BI cgroup \fR=\fPstr
832 Add job to this control group. If it doesn't exist, it will be created.
833 The system must have a mounted cgroup blkio mount point for this to work. If
834 your system doesn't have it mounted, you can do so with:
836 # mount -t cgroup -o blkio none /cgroup
838 .BI cgroup_weight \fR=\fPint
839 Set the weight of the cgroup to this value. See the documentation that comes
840 with the kernel, allowed values are in the range of 100..1000.
842 .BI cgroup_nodelete \fR=\fPbool
843 Normally fio will delete the cgroups it has created after the job completion.
844 To override this behavior and to leave cgroups around after the job completion,
845 set cgroup_nodelete=1. This can be useful if one wants to inspect various
846 cgroup files after job completion. Default: false
849 Instead of running as the invoking user, set the user ID to this value before
850 the thread/process does any work.
853 Set group ID, see \fBuid\fR.
855 While running, \fBfio\fR will display the status of the created jobs. For
859 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
862 The characters in the first set of brackets denote the current status of each
863 threads. The possible values are:
869 Setup but not started.
875 Initialized, waiting.
878 Running, doing sequential reads.
881 Running, doing random reads.
884 Running, doing sequential writes.
887 Running, doing random writes.
890 Running, doing mixed sequential reads/writes.
893 Running, doing mixed random reads/writes.
896 Running, currently waiting for \fBfsync\fR\|(2).
899 Running, verifying written data.
902 Exited, not reaped by main thread.
905 Exited, thread reaped.
909 The second set of brackets shows the estimated completion percentage of
910 the current group. The third set shows the read and write I/O rate,
911 respectively. Finally, the estimated run time of the job is displayed.
913 When \fBfio\fR completes (or is interrupted by Ctrl-C), it will show data
914 for each thread, each group of threads, and each disk, in that order.
916 Per-thread statistics first show the threads client number, group-id, and
917 error code. The remaining figures are as follows:
921 Number of megabytes of I/O performed.
924 Average data rate (bandwidth).
930 Submission latency minimum, maximum, average and standard deviation. This is
931 the time it took to submit the I/O.
934 Completion latency minimum, maximum, average and standard deviation. This
935 is the time between submission and completion.
938 Bandwidth minimum, maximum, percentage of aggregate bandwidth received, average
939 and standard deviation.
942 CPU usage statistics. Includes user and system time, number of context switches
943 this thread went through and number of major and minor page faults.
946 Distribution of I/O depths. Each depth includes everything less than (or equal)
947 to it, but greater than the previous depth.
950 Number of read/write requests issued, and number of short read/write requests.
953 Distribution of I/O completion latencies. The numbers follow the same pattern
957 The group statistics show:
962 Number of megabytes I/O performed.
965 Aggregate bandwidth of threads in the group.
968 Minimum average bandwidth a thread saw.
971 Maximum average bandwidth a thread saw.
974 Shortest runtime of threads in the group.
977 Longest runtime of threads in the group.
981 Finally, disk statistics are printed with reads first:
986 Number of I/Os performed by all groups.
989 Number of merges in the I/O scheduler.
992 Number of ticks we kept the disk busy.
995 Total time spent in the disk queue.
1002 If the \fB\-\-minimal\fR option is given, the results will be printed in a
1003 semicolon-delimited format suitable for scripted use - a job description
1004 (if provided) follows on a new line. Note that the first
1005 number in the line is the version number. If the output has to be changed
1006 for some reason, this number will be incremented by 1 to signify that
1007 change. The fields are:
1010 .B version, jobname, groupid, error
1014 .B KB I/O, bandwidth \fR(KB/s)\fP, runtime \fR(ms)\fP
1018 .B min, max, mean, standard deviation
1022 .B min, max, mean, standard deviation
1026 .B min, max, mean, standard deviation
1030 .B min, max, aggregate percentage of total, mean, standard deviation
1036 .B KB I/O, bandwidth \fR(KB/s)\fP, runtime \fR(ms)\fP
1040 .B min, max, mean, standard deviation
1044 .B min, max, mean, standard deviation
1048 .B min, max, mean, standard deviation
1052 .B min, max, aggregate percentage of total, mean, standard deviation
1058 .B user, system, context switches, major page faults, minor page faults
1061 IO depth distribution:
1063 .B <=1, 2, 4, 8, 16, 32, >=64
1066 IO latency distribution:
1070 .B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
1074 .B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
1078 Error Info (dependant on continue_on_error, default off):
1080 .B total # errors, first error code
1083 .B text description (if provided in config - appears on newline)
1087 was written by Jens Axboe <jens.axboe@oracle.com>,
1088 now Jens Axboe <jaxboe@fusionio.com>.
1090 This man page was written by Aaron Carroll <aaronc@cse.unsw.edu.au> based
1091 on documentation by Jens Axboe.
1092 .SH "REPORTING BUGS"
1093 Report bugs to the \fBfio\fR mailing list <fio@vger.kernel.org>.
1096 For further documentation see \fBHOWTO\fR and \fBREADME\fR.
1098 Sample jobfiles are available in the \fBexamples\fR directory.