1IOSTAT(1) Linux User's Manual IOSTAT(1)
2
6 iostat - Report Central Processing Unit (CPU) statistics and input/out‐
7 put statistics for devices and partitions.
8
11 iostat [ -c ] [ -d ] [ -h ] [ -k | -m ] [ -N ] [ -s ] [ -t ] [ -V ] [
12 -x ] [ -y ] [ -z ] [ --dec={ 0 | 1 | 2 } ] [ { -f | +f } directory ] [
13 -j { ID | LABEL | PATH | UUID | ... } ] [ -o JSON ] [ [ -H ] -g
14 group_name ] [ --human ] [ --pretty ] [ -p [ device[,...] | ALL ] ] [
15 device [...] | ALL ] [ interval [ count ] ]
16
19 The iostat command is used for monitoring system input/output device
20 loading by observing the time the devices are active in relation to
21 their average transfer rates. The iostat command generates reports that
22 can be used to change system configuration to better balance the in‐
23 put/output load between physical disks.
24 The first report generated by the iostat command provides statistics
26 concerning the time since the system was booted, unless the -y option
27 is used (in this case, this first report is omitted). Each subsequent
28 report covers the time since the previous report. All statistics are
29 reported each time the iostat command is run. The report consists of a
30 CPU header row followed by a row of CPU statistics. On multiprocessor
31 systems, CPU statistics are calculated system-wide as averages among
32 all processors. A device header row is displayed followed by a line of
33 statistics for each device that is configured.
34 The interval parameter specifies the amount of time in seconds between
36 each report. The count parameter can be specified in conjunction with
37 the interval parameter. If the count parameter is specified, the value
38 of count determines the number of reports generated at interval seconds
39 apart. If the interval parameter is specified without the count parame‐
40 ter, the iostat command generates reports continuously.
41
44 The iostat command generates two types of reports, the CPU Utilization
45 report and the Device Utilization report.
46 CPU Utilization Report
49 The first report generated by the iostat command is the CPU Uti‐
50 lization Report. For multiprocessor systems, the CPU values are
51 global averages among all processors. The report has the fol‐
52 lowing format:
53 %user Show the percentage of CPU utilization that occurred
55 while executing at the user level (application).
56 %nice Show the percentage of CPU utilization that occurred
58 while executing at the user level with nice priority.
59 %system
61 Show the percentage of CPU utilization that occurred
62 while executing at the system level (kernel).
63 %iowait
65 Show the percentage of time that the CPU or CPUs were
66 idle during which the system had an outstanding disk I/O
67 request.
68 %steal Show the percentage of time spent in involuntary wait by
70 the virtual CPU or CPUs while the hypervisor was servic‐
71 ing another virtual processor.
72 %idle Show the percentage of time that the CPU or CPUs were
74 idle and the system did not have an outstanding disk I/O
75 request.
76 Device Utilization Report
78 The second report generated by the iostat command is the Device
79 Utilization Report. The device report provides statistics on a
80 per physical device or partition basis. Block devices and parti‐
81 tions for which statistics are to be displayed may be entered on
82 the command line. If no device nor partition is entered, then
83 statistics are displayed for every device used by the system,
84 and providing that the kernel maintains statistics for it. If
85 the ALL keyword is given on the command line, then statistics
86 are displayed for every device defined by the system, including
87 those that have never been used. Transfer rates are shown in 1K
88 blocks by default, unless the environment variable POSIXLY_COR‐
89 RECT is set, in which case 512-byte blocks are used. The report
90 may show the following fields, depending on the flags used (e.g.
91 -x, -s and -k or -m):
92 Device:
94 This column gives the device (or partition) name as
95 listed in the /dev directory.
96 tps Indicate the number of transfers per second that were is‐
98 sued to the device. A transfer is an I/O request to the
99 device. Multiple logical requests can be combined into a
100 single I/O request to the device. A transfer is of inde‐
101 terminate size.
102 Blk_read/s (kB_read/s, MB_read/s)
104 Indicate the amount of data read from the device ex‐
105 pressed in a number of blocks (kilobytes, megabytes) per
106 second. Blocks are equivalent to sectors and therefore
107 have a size of 512 bytes.
108 Blk_wrtn/s (kB_wrtn/s, MB_wrtn/s)
110 Indicate the amount of data written to the device ex‐
111 pressed in a number of blocks (kilobytes, megabytes) per
112 second.
113 Blk_dscd/s (kB_dscd/s, MB_dscd/s)
115 Indicate the amount of data discarded for the device ex‐
116 pressed in a number of blocks (kilobytes, megabytes) per
117 second.
118 Blk_w+d/s (kB_w+d/s, MB_w+d/s)
120 Indicate the amount of data written to or discarded for
121 the device expressed in a number of blocks (kilobytes,
122 megabytes) per second.
123 Blk_read (kB_read, MB_read)
125 The total number of blocks (kilobytes, megabytes) read.
126 Blk_wrtn (kB_wrtn, MB_wrtn)
128 The total number of blocks (kilobytes, megabytes) writ‐
129 ten.
130 Blk_dscd (kB_dscd, MB_dscd)
132 The total number of blocks (kilobytes, megabytes) dis‐
133 carded.
134 Blk_w+d (kB_w+d, MB_w+d)
136 The total number of blocks (kilobytes, megabytes) written
137 or discarded.
138 r/s The number (after merges) of read requests completed per
140 second for the device.
141 w/s The number (after merges) of write requests completed per
143 second for the device.
144 d/s The number (after merges) of discard requests completed
146 per second for the device.
147 f/s The number (after merges) of flush requests completed per
149 second for the device. This counts flush requests exe‐
150 cuted by disks. Flush requests are not tracked for parti‐
151 tions. Before being merged, flush operations are counted
152 as writes.
153 sec/s (kB/s, MB/s)
155 The number of sectors (kilobytes, megabytes) read from,
156 written to or discarded for the device per second.
157 rsec/s (rkB/s, rMB/s)
159 The number of sectors (kilobytes, megabytes) read from
160 the device per second.
161 wsec/s (wkB/s, wMB/s)
163 The number of sectors (kilobytes, megabytes) written to
164 the device per second.
165 dsec/s (dkB/s, dMB/s)
167 The number of sectors (kilobytes, megabytes) discarded
168 for the device per second.
169 rqm/s The number of I/O requests merged per second that were
171 queued to the device.
172 rrqm/s The number of read requests merged per second that were
174 queued to the device.
175 wrqm/s The number of write requests merged per second that were
177 queued to the device.
178 drqm/s The number of discard requests merged per second that
180 were queued to the device.
181 %rrqm The percentage of read requests merged together before
183 being sent to the device.
184 %wrqm The percentage of write requests merged together before
186 being sent to the device.
187 %drqm The percentage of discard requests merged together before
189 being sent to the device.
190 areq-sz
192 The average size (in kilobytes) of the I/O requests that
193 were issued to the device.
194 Note: In previous versions, this field was known as av‐
195 grq-sz and was expressed in sectors.
196 rareq-sz
198 The average size (in kilobytes) of the read requests that
199 were issued to the device.
200 wareq-sz
202 The average size (in kilobytes) of the write requests
203 that were issued to the device.
204 dareq-sz
206 The average size (in kilobytes) of the discard requests
207 that were issued to the device.
208 await The average time (in milliseconds) for I/O requests is‐
210 sued to the device to be served. This includes the time
211 spent by the requests in queue and the time spent servic‐
212 ing them.
213 r_await
215 The average time (in milliseconds) for read requests is‐
216 sued to the device to be served. This includes the time
217 spent by the requests in queue and the time spent servic‐
218 ing them.
219 w_await
221 The average time (in milliseconds) for write requests is‐
222 sued to the device to be served. This includes the time
223 spent by the requests in queue and the time spent servic‐
224 ing them.
225 d_await
227 The average time (in milliseconds) for discard requests
228 issued to the device to be served. This includes the time
229 spent by the requests in queue and the time spent servic‐
230 ing them.
231 f_await
233 The average time (in milliseconds) for flush requests is‐
234 sued to the device to be served. The block layer com‐
235 bines flush requests and executes at most one at a time.
236 Thus flush operations could be twice as long: Wait for
237 current flush request, then execute it, then wait for the
238 next one.
239 aqu-sz The average queue length of the requests that were issued
241 to the device.
242 Note: In previous versions, this field was known as
243 avgqu-sz.
244 %util Percentage of elapsed time during which I/O requests were
246 issued to the device (bandwidth utilization for the de‐
247 vice). Device saturation occurs when this value is close
248 to 100% for devices serving requests serially. But for
249 devices serving requests in parallel, such as RAID arrays
250 and modern SSDs, this number does not reflect their per‐
251 formance limits.
252
255 -c Display the CPU utilization report.
256 -d Display the device utilization report.
258 --dec={ 0 | 1 | 2 }
260 Specify the number of decimal places to use (0 to 2, default
261 value is 2).
262 -f directory
264 +f directory
265 Specify an alternative directory for iostat to read devices sta‐
266 tistics. Option -f tells iostat to use only the files located in
267 the alternative directory, whereas option +f tells it to use
268 both the standard kernel files and the files located in the al‐
269 ternative directory to read device statistics.
270 directory is a directory containing files with statistics for
272 devices managed in userspace. It may contain:
273 - a "diskstats" file whose format is compliant with that located
275 in "/proc",
276 - statistics for individual devices contained in files whose
277 format is compliant with that of files located in "/sys".
278 In particular, the following files located in directory may be
280 used by iostat:
281 directory/block/device/stat
283 directory/block/device/partition/stat
284 partition files must have an entry in directory/dev/block/ di‐
286 rectory, e.g.:
287 directory/dev/block/major:minor --> ../../block/device/partition
289 -g group_name { device [...] | ALL }
291 Display statistics for a group of devices. The iostat command
292 reports statistics for each individual device in the list then a
293 line of global statistics for the group displayed as group_name
294 and made up of all the devices in the list. The ALL keyword
295 means that all the block devices defined by the system shall be
296 included in the group.
297 -H This option must be used with option -g and indicates that only
299 global statistics for the group are to be displayed, and not
300 statistics for individual devices in the group.
301 -h This option is equivalent to specifying --human --pretty.
303 --human
305 Print sizes in human readable format (e.g. 1.0k, 1.2M, etc.)
306 The units displayed with this option supersede any other default
307 units (e.g. kilobytes, sectors...) associated with the metrics.
308 -j { ID | LABEL | PATH | UUID | ... } [ device [...] | ALL ]
310 Display persistent device names. Keywords ID, LABEL, etc. spec‐
311 ify the type of the persistent name. These keywords are not lim‐
312 ited, only prerequisite is that directory with required persis‐
313 tent names is present in /dev/disk. Optionally, multiple de‐
314 vices can be specified in the chosen persistent name type. Be‐
315 cause persistent device names are usually long, option --pretty
316 is implicitly set with this option.
317 -k Display statistics in kilobytes per second.
319 -m Display statistics in megabytes per second.
321 -N Display the registered device mapper names for any device mapper
323 devices. Useful for viewing LVM2 statistics.
324 -o JSON
326 Display the statistics in JSON (Javascript Object Notation) for‐
327 mat. JSON output field order is undefined, and new fields may
328 be added in the future.
329 -p [ { device[,...] | ALL } ]
331 Display statistics for block devices and all their partitions
332 that are used by the system. If a device name is entered on the
333 command line, then statistics for it and all its partitions are
334 displayed. Last, the ALL keyword indicates that statistics have
335 to be displayed for all the block devices and partitions defined
336 by the system, including those that have never been used. If op‐
337 tion -j is defined before this option, devices entered on the
338 command line can be specified with the chosen persistent name
339 type.
340 --pretty
342 Make the Device Utilization Report easier to read by a human.
343 -s Display a short (narrow) version of the report that should fit
345 in 80 characters wide screens.
346 -t Print the time for each report displayed. The timestamp format
348 may depend on the value of the S_TIME_FORMAT environment vari‐
349 able (see below).
350 -V Print version number then exit.
352 -x Display extended statistics.
354 -y Omit first report with statistics since system boot, if display‐
356 ing multiple records at given interval.
357 -z Tell iostat to omit output for any devices for which there was
359 no activity during the sample period.
360
363 The iostat command takes into account the following environment vari‐
364 ables:
365 POSIXLY_CORRECT
367 When this variable is set, transfer rates are shown in 512-byte
368 blocks instead of the default 1K blocks.
369 S_COLORS
371 By default statistics are displayed in color when the output is
372 connected to a terminal. Use this variable to change the set‐
373 tings. Possible values for this variable are never, always or
374 auto (the latter is equivalent to the default settings).
375 Please note that the color (being red, yellow, or some other
376 color) used to display a value is not indicative of any kind of
377 issue simply because of the color. It only indicates different
378 ranges of values.
379 S_COLORS_SGR
381 Specify the colors and other attributes used to display statis‐
382 tics on the terminal. Its value is a colon-separated list of
383 capabilities that defaults to
384 H=31;1:I=32;22:M=35;1:N=34;1:Z=34;22. Supported capabilities
385 are:
386 H= SGR (Select Graphic Rendition) substring for percentage
388 values greater than or equal to 75%.
389 I= SGR substring for device names.
391 M= SGR substring for percentage values in the range from 50%
393 to 75%.
394 N= SGR substring for non-zero statistics values.
396 Z= SGR substring for zero values.
398 S_TIME_FORMAT
400 If this variable exists and its value is ISO then the current
401 locale will be ignored when printing the date in the report
402 header. The iostat command will use the ISO 8601 format (YYYY-
403 MM-DD) instead. The timestamp displayed with option -t will
404 also be compliant with ISO 8601 format.
405
408 iostat Display a single history since boot report for all CPU and De‐
409 vices.
410 iostat -d 2
412 Display a continuous device report at two second intervals.
413 iostat -d 2 6
415 Display six reports at two second intervals for all devices.
416 iostat -x sda sdb 2 6
418 Display six reports of extended statistics at two second inter‐
419 vals for devices sda and sdb.
420 iostat -p sda 2 6
422 Display six reports at two second intervals for device sda and
423 all its partitions (sda1, etc.)
424
427 /proc filesystem must be mounted for iostat to work.
428 Kernels older than 2.6.x are no longer supported.
430 Although iostat speaks of kilobytes (kB), megabytes (MB)..., it actu‐
432 ally uses kibibytes (kiB), mebibytes (MiB)... A kibibyte is equal to
433 1024 bytes, and a mebibyte is equal to 1024 kibibytes.
434
437 /proc/stat contains system statistics.
438 /proc/uptime contains system uptime.
439 /proc/diskstats contains disks statistics.
440 /sys contains statistics for block devices.
441 /proc/self/mountstats contains statistics for network filesystems.
442 /dev/disk contains persistent device names.
443
446 Sebastien Godard (sysstat <at> orange.fr)
447
450 sar(1), pidstat(1), mpstat(1), vmstat(8), tapestat(1), nfsiostat(1),
451 cifsiostat(1)
452 https://github.com/sysstat/sysstat
454 http://pagesperso-orange.fr/sebastien.godard/
455Linux OCTOBER 2020 IOSTAT(1)