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 \-\-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 \-\-timeout \fR=\fPtimeout
24 Limit run time to \fItimeout\fR seconds.
27 Generate per-job latency logs.
30 Generate per-job bandwidth logs.
33 Print statistics in a terse, semicolon-delimited format.
36 Display version information and exit.
38 .BI \-\-terse\-version \fR=\fPversion
39 Set terse version output format (Current version 3, or older version 2).
42 Display usage information and exit.
44 .BI \-\-cmdhelp \fR=\fPcommand
45 Print help information for \fIcommand\fR. May be `all' for all commands.
47 .BI \-\-showcmd \fR=\fPjobfile
48 Convert \fIjobfile\fR to a set of command-line options.
51 Enable read-only safety checks.
53 .BI \-\-eta \fR=\fPwhen
54 Specifies when real-time ETA estimate should be printed. \fIwhen\fR may
55 be one of `always', `never' or `auto'.
58 Turn on safety read-only checks, preventing any attempted write.
60 .BI \-\-section \fR=\fPsec
61 Only run section \fIsec\fR from job file. Multiple of these options can be given, adding more sections to run.
63 .BI \-\-alloc\-size \fR=\fPkb
64 Set the internal smalloc pool size to \fIkb\fP kilobytes.
66 .BI \-\-warnings\-fatal
67 All fio parser warnings are fatal, causing fio to exit with an error.
69 .BI \-\-max\-jobs \fR=\fPnr
70 Set the maximum allowed number of jobs (threads/processes) to suport.
72 .BI \-\-server \fR=\fPargs
73 Start a backend server, with \fIargs\fP specifying what to listen to. See client/server section.
75 .BI \-\-daemonize \fR=\fPpidfile
76 Background a fio server, writing the pid to the given pid file.
78 .BI \-\-client \fR=\fPhost
79 Instead of running the jobs locally, send and run them on the given host.
81 Job files are in `ini' format. They consist of one or more
82 job definitions, which begin with a job name in square brackets and
83 extend to the next job name. The job name can be any ASCII string
84 except `global', which has a special meaning. Following the job name is
85 a sequence of zero or more parameters, one per line, that define the
86 behavior of the job. Any line starting with a `;' or `#' character is
87 considered a comment and ignored.
89 If \fIjobfile\fR is specified as `-', the job file will be read from
92 The global section contains default parameters for jobs specified in the
93 job file. A job is only affected by global sections residing above it,
94 and there may be any number of global sections. Specific job definitions
95 may override any parameter set in global sections.
98 Some parameters may take arguments of a specific type. The types used are:
101 String: a sequence of alphanumeric characters.
104 SI integer: a whole number, possibly containing a suffix denoting the base unit
105 of the value. Accepted suffixes are `k', 'M', 'G', 'T', and 'P', denoting
106 kilo (1024), mega (1024^2), giga (1024^3), tera (1024^4), and peta (1024^5)
107 respectively. The suffix is not case sensitive. If prefixed with '0x', the
108 value is assumed to be base 16 (hexadecimal). A suffix may include a trailing 'b',
109 for instance 'kb' is identical to 'k'. You can specify a base 10 value
110 by using 'KiB', 'MiB', 'GiB', etc. This is useful for disk drives where
111 values are often given in base 10 values. Specifying '30GiB' will get you
115 Boolean: a true or false value. `0' denotes false, `1' denotes true.
118 Integer range: a range of integers specified in the format
119 \fIlower\fR:\fIupper\fR or \fIlower\fR\-\fIupper\fR. \fIlower\fR and
120 \fIupper\fR may contain a suffix as described above. If an option allows two
121 sets of ranges, they are separated with a `,' or `/' character. For example:
125 List of floating numbers: A list of floating numbers, separated by
130 May be used to override the job name. On the command line, this parameter
131 has the special purpose of signalling the start of a new job.
133 .BI description \fR=\fPstr
134 Human-readable description of the job. It is printed when the job is run, but
135 otherwise has no special purpose.
137 .BI directory \fR=\fPstr
138 Prefix filenames with this directory. Used to place files in a location other
141 .BI filename \fR=\fPstr
143 normally makes up a file name based on the job name, thread number, and file
144 number. If you want to share files between threads in a job or several jobs,
145 specify a \fIfilename\fR for each of them to override the default. If the I/O
146 engine used is `net', \fIfilename\fR is the host and port to connect to in the
147 format \fIhost\fR/\fIport\fR. If the I/O engine is file-based, you can specify
148 a number of files by separating the names with a `:' character. `\-' is a
149 reserved name, meaning stdin or stdout, depending on the read/write direction
152 .BI lockfile \fR=\fPstr
153 Fio defaults to not locking any files before it does IO to them. If a file or
154 file descriptor is shared, fio can serialize IO to that file to make the end
155 result consistent. This is usual for emulating real workloads that share files.
161 No locking. This is the default.
164 Only one thread or process may do IO at the time, excluding all others.
167 Read-write locking on the file. Many readers may access the file at the same
168 time, but writes get exclusive access.
171 The option may be post-fixed with a lock batch number. If set, then each
172 thread/process may do that amount of IOs to the file before giving up the lock.
173 Since lock acquisition is expensive, batching the lock/unlocks will speed up IO.
176 .BI opendir \fR=\fPstr
177 Recursively open any files below directory \fIstr\fR.
179 .BI readwrite \fR=\fPstr "\fR,\fP rw" \fR=\fPstr
180 Type of I/O pattern. Accepted values are:
197 Mixed sequential reads and writes.
200 Mixed random reads and writes.
203 For mixed I/O, the default split is 50/50. For certain types of io the result
204 may still be skewed a bit, since the speed may be different. It is possible to
205 specify a number of IO's to do before getting a new offset, this is one by
206 appending a `:\fI<nr>\fR to the end of the string given. For a random read, it
207 would look like \fBrw=randread:8\fR for passing in an offset modifier with a
208 value of 8. If the postfix is used with a sequential IO pattern, then the value
209 specified will be added to the generated offset for each IO. For instance,
210 using \fBrw=write:4k\fR will skip 4k for every write. It turns sequential IO
211 into sequential IO with holes. See the \fBrw_sequencer\fR option.
214 .BI rw_sequencer \fR=\fPstr
215 If an offset modifier is given by appending a number to the \fBrw=<str>\fR line,
216 then this option controls how that number modifies the IO offset being
217 generated. Accepted values are:
222 Generate sequential offset
225 Generate the same offset
228 \fBsequential\fR is only useful for random IO, where fio would normally
229 generate a new random offset for every IO. If you append eg 8 to randread, you
230 would get a new random offset for every 8 IO's. The result would be a seek for
231 only every 8 IO's, instead of for every IO. Use \fBrw=randread:8\fR to specify
232 that. As sequential IO is already sequential, setting \fBsequential\fR for that
233 would not result in any differences. \fBidentical\fR behaves in a similar
234 fashion, except it sends the same offset 8 number of times before generating a
239 .BI kb_base \fR=\fPint
240 The base unit for a kilobyte. The defacto base is 2^10, 1024. Storage
241 manufacturers like to use 10^3 or 1000 as a base ten unit instead, for obvious
242 reasons. Allow values are 1024 or 1000, with 1024 being the default.
244 .BI randrepeat \fR=\fPbool
245 Seed the random number generator in a predictable way so results are repeatable
246 across runs. Default: true.
248 .BI use_os_rand \fR=\fPbool
249 Fio can either use the random generator supplied by the OS to generator random
250 offsets, or it can use it's own internal generator (based on Tausworthe).
251 Default is to use the internal generator, which is often of better quality and
252 faster. Default: false.
254 .BI fallocate \fR=\fPstr
255 Whether pre-allocation is performed when laying down files. Accepted values
261 Do not pre-allocate space.
264 Pre-allocate via posix_fallocate().
267 Pre-allocate via fallocate() with FALLOC_FL_KEEP_SIZE set.
270 Backward-compatible alias for 'none'.
273 Backward-compatible alias for 'posix'.
276 May not be available on all supported platforms. 'keep' is only
277 available on Linux. If using ZFS on Solaris this must be set to 'none'
278 because ZFS doesn't support it. Default: 'posix'.
281 .BI fadvise_hint \fR=\fPbool
282 Disable use of \fIposix_fadvise\fR\|(2) to advise the kernel what I/O patterns
283 are likely to be issued. Default: true.
286 Total size of I/O for this job. \fBfio\fR will run until this many bytes have
287 been transfered, unless limited by other options (\fBruntime\fR, for instance).
288 Unless \fBnrfiles\fR and \fBfilesize\fR options are given, this amount will be
289 divided between the available files for the job. If not set, fio will use the
290 full size of the given files or devices. If the the files do not exist, size
291 must be given. It is also possible to give size as a percentage between 1 and
292 100. If size=20% is given, fio will use 20% of the full size of the given files
295 .BI fill_device \fR=\fPbool "\fR,\fB fill_fs" \fR=\fPbool
296 Sets size to something really large and waits for ENOSPC (no space left on
297 device) as the terminating condition. Only makes sense with sequential write.
298 For a read workload, the mount point will be filled first then IO started on
299 the result. This option doesn't make sense if operating on a raw device node,
300 since the size of that is already known by the file system. Additionally,
301 writing beyond end-of-device will not return ENOSPC there.
303 .BI filesize \fR=\fPirange
304 Individual file sizes. May be a range, in which case \fBfio\fR will select sizes
305 for files at random within the given range, limited to \fBsize\fR in total (if
306 that is given). If \fBfilesize\fR is not specified, each created file is the
309 .BI blocksize \fR=\fPint[,int] "\fR,\fB bs" \fR=\fPint[,int]
310 Block size for I/O units. Default: 4k. Values for reads and writes can be
311 specified separately in the format \fIread\fR,\fIwrite\fR, either of
312 which may be empty to leave that value at its default.
314 .BI blocksize_range \fR=\fPirange[,irange] "\fR,\fB bsrange" \fR=\fPirange[,irange]
315 Specify a range of I/O block sizes. The issued I/O unit will always be a
316 multiple of the minimum size, unless \fBblocksize_unaligned\fR is set. Applies
317 to both reads and writes if only one range is given, but can be specified
318 separately with a comma seperating the values. Example: bsrange=1k-4k,2k-8k.
319 Also (see \fBblocksize\fR).
321 .BI bssplit \fR=\fPstr
322 This option allows even finer grained control of the block sizes issued,
323 not just even splits between them. With this option, you can weight various
324 block sizes for exact control of the issued IO for a job that has mixed
325 block sizes. The format of the option is bssplit=blocksize/percentage,
326 optionally adding as many definitions as needed separated by a colon.
327 Example: bssplit=4k/10:64k/50:32k/40 would issue 50% 64k blocks, 10% 4k
328 blocks and 40% 32k blocks. \fBbssplit\fR also supports giving separate
329 splits to reads and writes. The format is identical to what the
330 \fBbs\fR option accepts, the read and write parts are separated with a
333 .B blocksize_unaligned\fR,\fP bs_unaligned
334 If set, any size in \fBblocksize_range\fR may be used. This typically won't
335 work with direct I/O, as that normally requires sector alignment.
337 .BI blockalign \fR=\fPint[,int] "\fR,\fB ba" \fR=\fPint[,int]
338 At what boundary to align random IO offsets. Defaults to the same as 'blocksize'
339 the minimum blocksize given. Minimum alignment is typically 512b
340 for using direct IO, though it usually depends on the hardware block size.
341 This option is mutually exclusive with using a random map for files, so it
342 will turn off that option.
345 Initialise buffers with all zeros. Default: fill buffers with random data.
348 If this option is given, fio will refill the IO buffers on every submit. The
349 default is to only fill it at init time and reuse that data. Only makes sense
350 if zero_buffers isn't specified, naturally. If data verification is enabled,
351 refill_buffers is also automatically enabled.
353 .BI scramble_buffers \fR=\fPbool
354 If \fBrefill_buffers\fR is too costly and the target is using data
355 deduplication, then setting this option will slightly modify the IO buffer
356 contents to defeat normal de-dupe attempts. This is not enough to defeat
357 more clever block compression attempts, but it will stop naive dedupe
358 of blocks. Default: true.
360 .BI nrfiles \fR=\fPint
361 Number of files to use for this job. Default: 1.
363 .BI openfiles \fR=\fPint
364 Number of files to keep open at the same time. Default: \fBnrfiles\fR.
366 .BI file_service_type \fR=\fPstr
367 Defines how files to service are selected. The following types are defined:
372 Choose a file at random
375 Round robin over open files (default).
377 Do each file in the set sequentially.
380 The number of I/Os to issue before switching a new file can be specified by
381 appending `:\fIint\fR' to the service type.
384 .BI ioengine \fR=\fPstr
385 Defines how the job issues I/O. The following types are defined:
390 Basic \fIread\fR\|(2) or \fIwrite\fR\|(2) I/O. \fIfseek\fR\|(2) is used to
391 position the I/O location.
394 Basic \fIpread\fR\|(2) or \fIpwrite\fR\|(2) I/O.
397 Basic \fIreadv\fR\|(2) or \fIwritev\fR\|(2) I/O. Will emulate queuing by
398 coalescing adjacents IOs into a single submission.
401 Linux native asynchronous I/O. This engine also has a sub-option,
402 \fBuserspace_reap\fR. To set it, use \fBioengine=libaio:userspace_reap\fR.
403 Normally, with the libaio engine in use, fio will use the
404 \fIio_getevents\fR\|(3) system call to reap newly returned events. With this
405 flag turned on, the AIO ring will be read directly from user-space to reap
406 events. The reaping mode is only enabled when polling for a minimum of \fB0\fR
407 events (eg when \fBiodepth_batch_complete=0\fR).
410 POSIX asynchronous I/O using \fIaio_read\fR\|(3) and \fIaio_write\fR\|(3).
413 Solaris native asynchronous I/O.
416 Windows native asynchronous I/O.
419 File is memory mapped with \fImmap\fR\|(2) and data copied using
423 \fIsplice\fR\|(2) is used to transfer the data and \fIvmsplice\fR\|(2) to
424 transfer data from user-space to the kernel.
427 Use the syslet system calls to make regular read/write asynchronous.
430 SCSI generic sg v3 I/O. May be either synchronous using the SG_IO ioctl, or if
431 the target is an sg character device, we use \fIread\fR\|(2) and
432 \fIwrite\fR\|(2) for asynchronous I/O.
435 Doesn't transfer any data, just pretends to. Mainly used to exercise \fBfio\fR
436 itself and for debugging and testing purposes.
439 Transfer over the network. \fBfilename\fR must be set appropriately to
440 `\fIhost\fR,\fIport\fR,\fItype\fR' regardless of data direction. \fItype\fR
441 is one of \fBtcp\fR, \fBudp\fR, or \fBunix\fR. For UNIX domain sockets,
442 the \fIhost\fR parameter is a file system path.
445 Like \fBnet\fR, but uses \fIsplice\fR\|(2) and \fIvmsplice\fR\|(2) to map data
449 Doesn't transfer any data, but burns CPU cycles according to \fBcpuload\fR and
450 \fBcpucycles\fR parameters.
453 The GUASI I/O engine is the Generic Userspace Asynchronous Syscall Interface
454 approach to asycnronous I/O.
456 See <http://www.xmailserver.org/guasi\-lib.html>.
459 The RDMA I/O engine supports both RDMA memory semantics (RDMA_WRITE/RDMA_READ)
460 and channel semantics (Send/Recv) for the InfiniBand, RoCE and iWARP protocols.
463 Loads an external I/O engine object file. Append the engine filename as
468 .BI iodepth \fR=\fPint
469 Number of I/O units to keep in flight against the file. Note that increasing
470 iodepth beyond 1 will not affect synchronous ioengines (except for small
471 degress when verify_async is in use). Even async engines my impose OS
472 restrictions causing the desired depth not to be achieved. This may happen on
473 Linux when using libaio and not setting \fBdirect\fR=1, since buffered IO is
474 not async on that OS. Keep an eye on the IO depth distribution in the
475 fio output to verify that the achieved depth is as expected. Default: 1.
477 .BI iodepth_batch \fR=\fPint
478 Number of I/Os to submit at once. Default: \fBiodepth\fR.
480 .BI iodepth_batch_complete \fR=\fPint
481 This defines how many pieces of IO to retrieve at once. It defaults to 1 which
482 means that we'll ask for a minimum of 1 IO in the retrieval process from the
483 kernel. The IO retrieval will go on until we hit the limit set by
484 \fBiodepth_low\fR. If this variable is set to 0, then fio will always check for
485 completed events before queuing more IO. This helps reduce IO latency, at the
486 cost of more retrieval system calls.
488 .BI iodepth_low \fR=\fPint
489 Low watermark indicating when to start filling the queue again. Default:
492 .BI direct \fR=\fPbool
493 If true, use non-buffered I/O (usually O_DIRECT). Default: false.
495 .BI buffered \fR=\fPbool
496 If true, use buffered I/O. This is the opposite of the \fBdirect\fR parameter.
499 .BI offset \fR=\fPint
500 Offset in the file to start I/O. Data before the offset will not be touched.
503 How many I/Os to perform before issuing an \fBfsync\fR\|(2) of dirty data. If
504 0, don't sync. Default: 0.
506 .BI fdatasync \fR=\fPint
507 Like \fBfsync\fR, but uses \fBfdatasync\fR\|(2) instead to only sync the
508 data parts of the file. Default: 0.
510 .BI sync_file_range \fR=\fPstr:int
511 Use sync_file_range() for every \fRval\fP number of write operations. Fio will
512 track range of writes that have happened since the last sync_file_range() call.
513 \fRstr\fP can currently be one or more of:
517 SYNC_FILE_RANGE_WAIT_BEFORE
520 SYNC_FILE_RANGE_WRITE
523 SYNC_FILE_RANGE_WRITE
527 So if you do sync_file_range=wait_before,write:8, fio would use
528 \fBSYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE\fP for every 8 writes.
529 Also see the sync_file_range(2) man page. This option is Linux specific.
531 .BI overwrite \fR=\fPbool
532 If writing, setup the file first and do overwrites. Default: false.
534 .BI end_fsync \fR=\fPbool
535 Sync file contents when job exits. Default: false.
537 .BI fsync_on_close \fR=\fPbool
538 If true, sync file contents on close. This differs from \fBend_fsync\fR in that
539 it will happen on every close, not just at the end of the job. Default: false.
541 .BI rwmixcycle \fR=\fPint
542 How many milliseconds before switching between reads and writes for a mixed
543 workload. Default: 500ms.
545 .BI rwmixread \fR=\fPint
546 Percentage of a mixed workload that should be reads. Default: 50.
548 .BI rwmixwrite \fR=\fPint
549 Percentage of a mixed workload that should be writes. If \fBrwmixread\fR and
550 \fBrwmixwrite\fR are given and do not sum to 100%, the latter of the two
551 overrides the first. This may interfere with a given rate setting, if fio is
552 asked to limit reads or writes to a certain rate. If that is the case, then
553 the distribution may be skewed. Default: 50.
556 Normally \fBfio\fR will cover every block of the file when doing random I/O. If
557 this parameter is given, a new offset will be chosen without looking at past
558 I/O history. This parameter is mutually exclusive with \fBverify\fR.
560 .BI softrandommap \fR=\fPbool
561 See \fBnorandommap\fR. If fio runs with the random block map enabled and it
562 fails to allocate the map, if this option is set it will continue without a
563 random block map. As coverage will not be as complete as with random maps, this
564 option is disabled by default.
567 Run job with given nice value. See \fInice\fR\|(2).
570 Set I/O priority value of this job between 0 (highest) and 7 (lowest). See
573 .BI prioclass \fR=\fPint
574 Set I/O priority class. See \fIionice\fR\|(1).
576 .BI thinktime \fR=\fPint
577 Stall job for given number of microseconds between issuing I/Os.
579 .BI thinktime_spin \fR=\fPint
580 Pretend to spend CPU time for given number of microseconds, sleeping the rest
581 of the time specified by \fBthinktime\fR. Only valid if \fBthinktime\fR is set.
583 .BI thinktime_blocks \fR=\fPint
584 Number of blocks to issue before waiting \fBthinktime\fR microseconds.
588 Cap bandwidth used by this job. The number is in bytes/sec, the normal postfix
589 rules apply. You can use \fBrate\fR=500k to limit reads and writes to 500k each,
590 or you can specify read and writes separately. Using \fBrate\fR=1m,500k would
591 limit reads to 1MB/sec and writes to 500KB/sec. Capping only reads or writes
592 can be done with \fBrate\fR=,500k or \fBrate\fR=500k,. The former will only
593 limit writes (to 500KB/sec), the latter will only limit reads.
595 .BI ratemin \fR=\fPint
596 Tell \fBfio\fR to do whatever it can to maintain at least the given bandwidth.
597 Failing to meet this requirement will cause the job to exit. The same format
598 as \fBrate\fR is used for read vs write separation.
600 .BI rate_iops \fR=\fPint
601 Cap the bandwidth to this number of IOPS. Basically the same as rate, just
602 specified independently of bandwidth. The same format as \fBrate\fR is used for
603 read vs write seperation. If \fBblocksize\fR is a range, the smallest block
604 size is used as the metric.
606 .BI rate_iops_min \fR=\fPint
607 If this rate of I/O is not met, the job will exit. The same format as \fBrate\fR
608 is used for read vs write seperation.
610 .BI ratecycle \fR=\fPint
611 Average bandwidth for \fBrate\fR and \fBratemin\fR over this number of
612 milliseconds. Default: 1000ms.
614 .BI cpumask \fR=\fPint
615 Set CPU affinity for this job. \fIint\fR is a bitmask of allowed CPUs the job
616 may run on. See \fBsched_setaffinity\fR\|(2).
618 .BI cpus_allowed \fR=\fPstr
619 Same as \fBcpumask\fR, but allows a comma-delimited list of CPU numbers.
621 .BI startdelay \fR=\fPint
622 Delay start of job for the specified number of seconds.
624 .BI runtime \fR=\fPint
625 Terminate processing after the specified number of seconds.
628 If given, run for the specified \fBruntime\fR duration even if the files are
629 completely read or written. The same workload will be repeated as many times
630 as \fBruntime\fR allows.
632 .BI ramp_time \fR=\fPint
633 If set, fio will run the specified workload for this amount of time before
634 logging any performance numbers. Useful for letting performance settle before
635 logging results, thus minimizing the runtime required for stable results. Note
636 that the \fBramp_time\fR is considered lead in time for a job, thus it will
637 increase the total runtime if a special timeout or runtime is specified.
639 .BI invalidate \fR=\fPbool
640 Invalidate buffer-cache for the file prior to starting I/O. Default: true.
643 Use synchronous I/O for buffered writes. For the majority of I/O engines,
644 this means using O_SYNC. Default: false.
646 .BI iomem \fR=\fPstr "\fR,\fP mem" \fR=\fPstr
647 Allocation method for I/O unit buffer. Allowed values are:
652 Allocate memory with \fImalloc\fR\|(3).
655 Use shared memory buffers allocated through \fIshmget\fR\|(2).
658 Same as \fBshm\fR, but use huge pages as backing.
661 Use \fImmap\fR\|(2) for allocation. Uses anonymous memory unless a filename
662 is given after the option in the format `:\fIfile\fR'.
665 Same as \fBmmap\fR, but use huge files as backing.
668 The amount of memory allocated is the maximum allowed \fBblocksize\fR for the
669 job multiplied by \fBiodepth\fR. For \fBshmhuge\fR or \fBmmaphuge\fR to work,
670 the system must have free huge pages allocated. \fBmmaphuge\fR also needs to
671 have hugetlbfs mounted, and \fIfile\fR must point there. At least on Linux,
672 huge pages must be manually allocated. See \fB/proc/sys/vm/nr_hugehages\fR
673 and the documentation for that. Normally you just need to echo an appropriate
674 number, eg echoing 8 will ensure that the OS has 8 huge pages ready for
678 .BI iomem_align \fR=\fPint "\fR,\fP mem_align" \fR=\fPint
679 This indiciates the memory alignment of the IO memory buffers. Note that the
680 given alignment is applied to the first IO unit buffer, if using \fBiodepth\fR
681 the alignment of the following buffers are given by the \fBbs\fR used. In
682 other words, if using a \fBbs\fR that is a multiple of the page sized in the
683 system, all buffers will be aligned to this value. If using a \fBbs\fR that
684 is not page aligned, the alignment of subsequent IO memory buffers is the
685 sum of the \fBiomem_align\fR and \fBbs\fR used.
687 .BI hugepage\-size \fR=\fPint
688 Defines the size of a huge page. Must be at least equal to the system setting.
689 Should be a multiple of 1MB. Default: 4MB.
692 Terminate all jobs when one finishes. Default: wait for each job to finish.
694 .BI bwavgtime \fR=\fPint
695 Average bandwidth calculations over the given time in milliseconds. Default:
698 .BI iopsavgtime \fR=\fPint
699 Average IOPS calculations over the given time in milliseconds. Default:
702 .BI create_serialize \fR=\fPbool
703 If true, serialize file creation for the jobs. Default: true.
705 .BI create_fsync \fR=\fPbool
706 \fIfsync\fR\|(2) data file after creation. Default: true.
708 .BI create_on_open \fR=\fPbool
709 If true, the files are not created until they are opened for IO by the job.
711 .BI pre_read \fR=\fPbool
712 If this is given, files will be pre-read into memory before starting the given
713 IO operation. This will also clear the \fR \fBinvalidate\fR flag, since it is
714 pointless to pre-read and then drop the cache. This will only work for IO
715 engines that are seekable, since they allow you to read the same data
716 multiple times. Thus it will not work on eg network or splice IO.
718 .BI unlink \fR=\fPbool
719 Unlink job files when done. Default: false.
722 Specifies the number of iterations (runs of the same workload) of this job.
725 .BI do_verify \fR=\fPbool
726 Run the verify phase after a write phase. Only valid if \fBverify\fR is set.
729 .BI verify \fR=\fPstr
730 Method of verifying file contents after each iteration of the job. Allowed
735 .B md5 crc16 crc32 crc32c crc32c-intel crc64 crc7 sha256 sha512 sha1
736 Store appropriate checksum in the header of each block. crc32c-intel is
737 hardware accelerated SSE4.2 driven, falls back to regular crc32c if
738 not supported by the system.
741 Write extra information about each I/O (timestamp, block number, etc.). The
742 block number is verified. See \fBverify_pattern\fR as well.
745 Pretend to verify. Used for testing internals.
748 This option can be used for repeated burn-in tests of a system to make sure
749 that the written data is also correctly read back. If the data direction given
750 is a read or random read, fio will assume that it should verify a previously
751 written file. If the data direction includes any form of write, the verify will
752 be of the newly written data.
755 .BI verify_sort \fR=\fPbool
756 If true, written verify blocks are sorted if \fBfio\fR deems it to be faster to
757 read them back in a sorted manner. Default: true.
759 .BI verify_offset \fR=\fPint
760 Swap the verification header with data somewhere else in the block before
761 writing. It is swapped back before verifying.
763 .BI verify_interval \fR=\fPint
764 Write the verification header for this number of bytes, which should divide
765 \fBblocksize\fR. Default: \fBblocksize\fR.
767 .BI verify_pattern \fR=\fPstr
768 If set, fio will fill the io buffers with this pattern. Fio defaults to filling
769 with totally random bytes, but sometimes it's interesting to fill with a known
770 pattern for io verification purposes. Depending on the width of the pattern,
771 fio will fill 1/2/3/4 bytes of the buffer at the time(it can be either a
772 decimal or a hex number). The verify_pattern if larger than a 32-bit quantity
773 has to be a hex number that starts with either "0x" or "0X". Use with
776 .BI verify_fatal \fR=\fPbool
777 If true, exit the job on the first observed verification failure. Default:
780 .BI verify_dump \fR=\fPbool
781 If set, dump the contents of both the original data block and the data block we
782 read off disk to files. This allows later analysis to inspect just what kind of
783 data corruption occurred. On by default.
785 .BI verify_async \fR=\fPint
786 Fio will normally verify IO inline from the submitting thread. This option
787 takes an integer describing how many async offload threads to create for IO
788 verification instead, causing fio to offload the duty of verifying IO contents
789 to one or more separate threads. If using this offload option, even sync IO
790 engines can benefit from using an \fBiodepth\fR setting higher than 1, as it
791 allows them to have IO in flight while verifies are running.
793 .BI verify_async_cpus \fR=\fPstr
794 Tell fio to set the given CPU affinity on the async IO verification threads.
795 See \fBcpus_allowed\fP for the format used.
797 .BI verify_backlog \fR=\fPint
798 Fio will normally verify the written contents of a job that utilizes verify
799 once that job has completed. In other words, everything is written then
800 everything is read back and verified. You may want to verify continually
801 instead for a variety of reasons. Fio stores the meta data associated with an
802 IO block in memory, so for large verify workloads, quite a bit of memory would
803 be used up holding this meta data. If this option is enabled, fio will write
804 only N blocks before verifying these blocks.
806 .BI verify_backlog_batch \fR=\fPint
807 Control how many blocks fio will verify if verify_backlog is set. If not set,
808 will default to the value of \fBverify_backlog\fR (meaning the entire queue is
809 read back and verified). If \fBverify_backlog_batch\fR is less than
810 \fBverify_backlog\fR then not all blocks will be verified, if
811 \fBverify_backlog_batch\fR is larger than \fBverify_backlog\fR, some blocks
812 will be verified more than once.
814 .B stonewall "\fR,\fP wait_for_previous"
815 Wait for preceding jobs in the job file to exit before starting this one.
816 \fBstonewall\fR implies \fBnew_group\fR.
819 Start a new reporting group. If not given, all jobs in a file will be part
820 of the same reporting group, unless separated by a stonewall.
822 .BI numjobs \fR=\fPint
823 Number of clones (processes/threads performing the same workload) of this job.
827 If set, display per-group reports instead of per-job when \fBnumjobs\fR is
831 Use threads created with \fBpthread_create\fR\|(3) instead of processes created
832 with \fBfork\fR\|(2).
834 .BI zonesize \fR=\fPint
835 Divide file into zones of the specified size in bytes. See \fBzoneskip\fR.
837 .BI zoneskip \fR=\fPint
838 Skip the specified number of bytes when \fBzonesize\fR bytes of data have been
841 .BI write_iolog \fR=\fPstr
842 Write the issued I/O patterns to the specified file. Specify a separate file
843 for each job, otherwise the iologs will be interspersed and the file may be
846 .BI read_iolog \fR=\fPstr
847 Replay the I/O patterns contained in the specified file generated by
848 \fBwrite_iolog\fR, or may be a \fBblktrace\fR binary file.
850 .BI replay_no_stall \fR=\fPint
851 While replaying I/O patterns using \fBread_iolog\fR the default behavior
852 attempts to respect timing information between I/Os. Enabling
853 \fBreplay_no_stall\fR causes I/Os to be replayed as fast as possible while
854 still respecting ordering.
856 .BI replay_redirect \fR=\fPstr
857 While replaying I/O patterns using \fBread_iolog\fR the default behavior
858 is to replay the IOPS onto the major/minor device that each IOP was recorded
859 from. Setting \fBreplay_redirect\fR causes all IOPS to be replayed onto the
860 single specified device regardless of the device it was recorded from.
862 .BI write_bw_log \fR=\fPstr
863 If given, write a bandwidth log of the jobs in this job file. Can be used to
864 store data of the bandwidth of the jobs in their lifetime. The included
865 fio_generate_plots script uses gnuplot to turn these text files into nice
866 graphs. See \fBwrite_log_log\fR for behaviour of given filename. For this
867 option, the postfix is _bw.log.
869 .BI write_lat_log \fR=\fPstr
870 Same as \fBwrite_bw_log\fR, but writes I/O completion latencies. If no
871 filename is given with this option, the default filename of "jobname_type.log"
872 is used. Even if the filename is given, fio will still append the type of log.
874 .BI write_iops_log \fR=\fPstr
875 Same as \fBwrite_bw_log\fR, but writes IOPS. If no filename is given with this
876 option, the default filename of "jobname_type.log" is used. Even if the
877 filename is given, fio will still append the type of log.
879 .BI disable_lat \fR=\fPbool
880 Disable measurements of total latency numbers. Useful only for cutting
881 back the number of calls to gettimeofday, as that does impact performance at
882 really high IOPS rates. Note that to really get rid of a large amount of these
883 calls, this option must be used with disable_slat and disable_bw as well.
885 .BI disable_clat \fR=\fPbool
886 Disable measurements of completion latency numbers. See \fBdisable_lat\fR.
888 .BI disable_slat \fR=\fPbool
889 Disable measurements of submission latency numbers. See \fBdisable_lat\fR.
891 .BI disable_bw_measurement \fR=\fPbool
892 Disable measurements of throughput/bandwidth numbers. See \fBdisable_lat\fR.
894 .BI lockmem \fR=\fPint
895 Pin the specified amount of memory with \fBmlock\fR\|(2). Can be used to
896 simulate a smaller amount of memory.
898 .BI exec_prerun \fR=\fPstr
899 Before running the job, execute the specified command with \fBsystem\fR\|(3).
901 .BI exec_postrun \fR=\fPstr
902 Same as \fBexec_prerun\fR, but the command is executed after the job completes.
904 .BI ioscheduler \fR=\fPstr
905 Attempt to switch the device hosting the file to the specified I/O scheduler.
907 .BI cpuload \fR=\fPint
908 If the job is a CPU cycle-eater, attempt to use the specified percentage of
911 .BI cpuchunks \fR=\fPint
912 If the job is a CPU cycle-eater, split the load into cycles of the
913 given time in milliseconds.
915 .BI disk_util \fR=\fPbool
916 Generate disk utilization statistics if the platform supports it. Default: true.
918 .BI gtod_reduce \fR=\fPbool
919 Enable all of the gettimeofday() reducing options (disable_clat, disable_slat,
920 disable_bw) plus reduce precision of the timeout somewhat to really shrink the
921 gettimeofday() call count. With this option enabled, we only do about 0.4% of
922 the gtod() calls we would have done if all time keeping was enabled.
924 .BI gtod_cpu \fR=\fPint
925 Sometimes it's cheaper to dedicate a single thread of execution to just getting
926 the current time. Fio (and databases, for instance) are very intensive on
927 gettimeofday() calls. With this option, you can set one CPU aside for doing
928 nothing but logging current time to a shared memory location. Then the other
929 threads/processes that run IO workloads need only copy that segment, instead of
930 entering the kernel with a gettimeofday() call. The CPU set aside for doing
931 these time calls will be excluded from other uses. Fio will manually clear it
932 from the CPU mask of other jobs.
934 .BI cgroup \fR=\fPstr
935 Add job to this control group. If it doesn't exist, it will be created.
936 The system must have a mounted cgroup blkio mount point for this to work. If
937 your system doesn't have it mounted, you can do so with:
939 # mount \-t cgroup \-o blkio none /cgroup
941 .BI cgroup_weight \fR=\fPint
942 Set the weight of the cgroup to this value. See the documentation that comes
943 with the kernel, allowed values are in the range of 100..1000.
945 .BI cgroup_nodelete \fR=\fPbool
946 Normally fio will delete the cgroups it has created after the job completion.
947 To override this behavior and to leave cgroups around after the job completion,
948 set cgroup_nodelete=1. This can be useful if one wants to inspect various
949 cgroup files after job completion. Default: false
952 Instead of running as the invoking user, set the user ID to this value before
953 the thread/process does any work.
956 Set group ID, see \fBuid\fR.
958 .BI clat_percentiles \fR=\fPbool
959 Enable the reporting of percentiles of completion latencies.
961 .BI percentile_list \fR=\fPfloat_list
962 Overwrite the default list of percentiles for completion
963 latencies. Each number is a floating number in the range (0,100], and
964 the maximum length of the list is 20. Use ':' to separate the
965 numbers. For example, \-\-percentile_list=99.5:99.9 will cause fio to
966 report the values of completion latency below which 99.5% and 99.9% of
967 the observed latencies fell, respectively.
969 While running, \fBfio\fR will display the status of the created jobs. For
973 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
976 The characters in the first set of brackets denote the current status of each
977 threads. The possible values are:
983 Setup but not started.
989 Initialized, waiting.
992 Running, doing sequential reads.
995 Running, doing random reads.
998 Running, doing sequential writes.
1001 Running, doing random writes.
1004 Running, doing mixed sequential reads/writes.
1007 Running, doing mixed random reads/writes.
1010 Running, currently waiting for \fBfsync\fR\|(2).
1013 Running, verifying written data.
1016 Exited, not reaped by main thread.
1019 Exited, thread reaped.
1023 The second set of brackets shows the estimated completion percentage of
1024 the current group. The third set shows the read and write I/O rate,
1025 respectively. Finally, the estimated run time of the job is displayed.
1027 When \fBfio\fR completes (or is interrupted by Ctrl-C), it will show data
1028 for each thread, each group of threads, and each disk, in that order.
1030 Per-thread statistics first show the threads client number, group-id, and
1031 error code. The remaining figures are as follows:
1035 Number of megabytes of I/O performed.
1038 Average data rate (bandwidth).
1044 Submission latency minimum, maximum, average and standard deviation. This is
1045 the time it took to submit the I/O.
1048 Completion latency minimum, maximum, average and standard deviation. This
1049 is the time between submission and completion.
1052 Bandwidth minimum, maximum, percentage of aggregate bandwidth received, average
1053 and standard deviation.
1056 CPU usage statistics. Includes user and system time, number of context switches
1057 this thread went through and number of major and minor page faults.
1060 Distribution of I/O depths. Each depth includes everything less than (or equal)
1061 to it, but greater than the previous depth.
1064 Number of read/write requests issued, and number of short read/write requests.
1067 Distribution of I/O completion latencies. The numbers follow the same pattern
1071 The group statistics show:
1076 Number of megabytes I/O performed.
1079 Aggregate bandwidth of threads in the group.
1082 Minimum average bandwidth a thread saw.
1085 Maximum average bandwidth a thread saw.
1088 Shortest runtime of threads in the group.
1091 Longest runtime of threads in the group.
1095 Finally, disk statistics are printed with reads first:
1100 Number of I/Os performed by all groups.
1103 Number of merges in the I/O scheduler.
1106 Number of ticks we kept the disk busy.
1109 Total time spent in the disk queue.
1116 If the \fB\-\-minimal\fR option is given, the results will be printed in a
1117 semicolon-delimited format suitable for scripted use - a job description
1118 (if provided) follows on a new line. Note that the first
1119 number in the line is the version number. If the output has to be changed
1120 for some reason, this number will be incremented by 1 to signify that
1121 change. The fields are:
1124 .B terse version, fio version, jobname, groupid, error
1128 .B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, IOPS, runtime \fR(ms)\fP
1132 .B min, max, mean, standard deviation
1136 .B min, max, mean, standard deviation
1138 Completion latency percentiles (20 fields):
1140 .B Xth percentile=usec
1144 .B min, max, mean, standard deviation
1148 .B min, max, aggregate percentage of total, mean, standard deviation
1154 .B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, IOPS, runtime \fR(ms)\fP
1158 .B min, max, mean, standard deviation
1162 .B min, max, mean, standard deviation
1164 Completion latency percentiles (20 fields):
1166 .B Xth percentile=usec
1170 .B min, max, mean, standard deviation
1174 .B min, max, aggregate percentage of total, mean, standard deviation
1180 .B user, system, context switches, major page faults, minor page faults
1183 IO depth distribution:
1185 .B <=1, 2, 4, 8, 16, 32, >=64
1188 IO latency distribution:
1192 .B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
1196 .B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
1200 Disk utilization (1 for each disk used):
1202 .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
1205 Error Info (dependent on continue_on_error, default off):
1207 .B total # errors, first error code
1210 .B text description (if provided in config - appears on newline)
1213 Normally you would run fio as a stand-alone application on the machine
1214 where the IO workload should be generated. However, it is also possible to
1215 run the frontend and backend of fio separately. This makes it possible to
1216 have a fio server running on the machine(s) where the IO workload should
1217 be running, while controlling it from another machine.
1219 To start the server, you would do:
1221 \fBfio \-\-server=args\fR
1223 on that machine, where args defines what fio listens to. The arguments
1224 are of the form 'type:hostname or IP:port'. 'type' is either 'ip' for
1225 TCP/IP, or 'sock' for a local unix domain socket. 'hostname' is either
1226 a hostname or IP address, and 'port' is the port to listen to (only valid
1227 for TCP/IP, not a local socket). Some examples:
1231 Start a fio server, listening on all interfaces on the default port (8765).
1233 2) fio --server=ip:hostname:4444
1235 Start a fio server, listening on IP belonging to hostname and on port 4444.
1237 3) fio --server=:4444
1239 Start a fio server, listening on all interfaces on port 4444.
1241 4) fio --server=1.2.3.4
1243 Start a fio server, listening on IP 1.2.3.4 on the default port.
1245 5) fio --server=sock:/tmp/fio.sock
1247 Start a fio server, listening on the local socket /tmp/fio.sock.
1249 When a server is running, you can connect to it from a client. The client
1252 fio --local-args --client=server --remote-args <job file(s)>
1254 where --local-args are arguments that are local to the client where it is
1255 running, 'server' is the connect string, and --remote-args and <job file(s)>
1256 are sent to the server. The 'server' string follows the same format as it
1257 does on the server side, to allow IP/hostname/socket and port strings.
1258 You can connect to multiple clients as well, to do that you could run:
1260 fio --client=server2 --client=server2 <job file(s)>
1264 was written by Jens Axboe <jens.axboe@oracle.com>,
1265 now Jens Axboe <jaxboe@fusionio.com>.
1267 This man page was written by Aaron Carroll <aaronc@cse.unsw.edu.au> based
1268 on documentation by Jens Axboe.
1269 .SH "REPORTING BUGS"
1270 Report bugs to the \fBfio\fR mailing list <fio@vger.kernel.org>.
1273 For further documentation see \fBHOWTO\fR and \fBREADME\fR.
1275 Sample jobfiles are available in the \fBexamples\fR directory.
projects / fio.git / blob