5 Since 2.4.20 (and some versions before, with patches), and 2.5.45,
6 more extensive disk statistics have been introduced to help measure disk
7 activity. Tools such as ``sar`` and ``iostat`` typically interpret these and do
8 the work for you, but in case you are interested in creating your own
9 tools, the fields are explained here.
11 In 2.4 now, the information is found as additional fields in
12 ``/proc/partitions``. In 2.6 and upper, the same information is found in two
13 places: one is in the file ``/proc/diskstats``, and the other is within
14 the sysfs file system, which must be mounted in order to obtain
15 the information. Throughout this document we'll assume that sysfs
16 is mounted on ``/sys``, although of course it may be mounted anywhere.
17 Both ``/proc/diskstats`` and sysfs use the same source for the information
18 and so should not differ.
20 Here are examples of these different formats::
23 3 0 39082680 hda 446216 784926 9550688 4382310 424847 312726 5922052 19310380 0 3376340 23705160
24 3 1 9221278 hda1 35486 0 35496 38030 0 0 0 0 0 38030 38030
27 446216 784926 9550688 4382310 424847 312726 5922052 19310380 0 3376340 23705160
28 35486 38030 38030 38030
31 3 0 hda 446216 784926 9550688 4382310 424847 312726 5922052 19310380 0 3376340 23705160
32 3 1 hda1 35486 38030 38030 38030
35 3 0 hda 446216 784926 9550688 4382310 424847 312726 5922052 19310380 0 3376340 23705160 0 0 0 0
37 On 2.4 you might execute ``grep 'hda ' /proc/partitions``. On 2.6+, you have
38 a choice of ``cat /sys/block/hda/stat`` or ``grep 'hda ' /proc/diskstats``.
40 The advantage of one over the other is that the sysfs choice works well
41 if you are watching a known, small set of disks. ``/proc/diskstats`` may
42 be a better choice if you are watching a large number of disks because
43 you'll avoid the overhead of 50, 100, or 500 or more opens/closes with
44 each snapshot of your disk statistics.
46 In 2.4, the statistics fields are those after the device name. In
47 the above example, the first field of statistics would be 446216.
48 By contrast, in 2.6+ if you look at ``/sys/block/hda/stat``, you'll
49 find just the eleven fields, beginning with 446216. If you look at
50 ``/proc/diskstats``, the eleven fields will be preceded by the major and
51 minor device numbers, and device name. Each of these formats provides
52 eleven fields of statistics, each meaning exactly the same things.
53 All fields except field 9 are cumulative since boot. Field 9 should
54 go to zero as I/Os complete; all others only increase (unless they
55 overflow and wrap). Yes, these are (32-bit or 64-bit) unsigned long
56 (native word size) numbers, and on a very busy or long-lived system they
57 may wrap. Applications should be prepared to deal with that; unless
58 your observations are measured in large numbers of minutes or hours,
59 they should not wrap twice before you notice them.
61 Each set of stats only applies to the indicated device; if you want
62 system-wide stats you'll have to find all the devices and sum them all up.
64 Field 1 -- # of reads completed
65 This is the total number of reads completed successfully.
67 Field 2 -- # of reads merged, field 6 -- # of writes merged
68 Reads and writes which are adjacent to each other may be merged for
69 efficiency. Thus two 4K reads may become one 8K read before it is
70 ultimately handed to the disk, and so it will be counted (and queued)
71 as only one I/O. This field lets you know how often this was done.
73 Field 3 -- # of sectors read
74 This is the total number of sectors read successfully.
76 Field 4 -- # of milliseconds spent reading
77 This is the total number of milliseconds spent by all reads (as
78 measured from __make_request() to end_that_request_last()).
80 Field 5 -- # of writes completed
81 This is the total number of writes completed successfully.
83 Field 6 -- # of writes merged
84 See the description of field 2.
86 Field 7 -- # of sectors written
87 This is the total number of sectors written successfully.
89 Field 8 -- # of milliseconds spent writing
90 This is the total number of milliseconds spent by all writes (as
91 measured from __make_request() to end_that_request_last()).
93 Field 9 -- # of I/Os currently in progress
94 The only field that should go to zero. Incremented as requests are
95 given to appropriate struct request_queue and decremented as they finish.
97 Field 10 -- # of milliseconds spent doing I/Os
98 This field increases so long as field 9 is nonzero.
100 Since 5.0 this field counts jiffies when at least one request was
101 started or completed. If request runs more than 2 jiffies then some
102 I/O time will not be accounted unless there are other requests.
104 Field 11 -- weighted # of milliseconds spent doing I/Os
105 This field is incremented at each I/O start, I/O completion, I/O
106 merge, or read of these stats by the number of I/Os in progress
107 (field 9) times the number of milliseconds spent doing I/O since the
108 last update of this field. This can provide an easy measure of both
109 I/O completion time and the backlog that may be accumulating.
111 Field 12 -- # of discards completed
112 This is the total number of discards completed successfully.
114 Field 13 -- # of discards merged
115 See the description of field 2
117 Field 14 -- # of sectors discarded
118 This is the total number of sectors discarded successfully.
120 Field 15 -- # of milliseconds spent discarding
121 This is the total number of milliseconds spent by all discards (as
122 measured from __make_request() to end_that_request_last()).
124 To avoid introducing performance bottlenecks, no locks are held while
125 modifying these counters. This implies that minor inaccuracies may be
126 introduced when changes collide, so (for instance) adding up all the
127 read I/Os issued per partition should equal those made to the disks ...
128 but due to the lack of locking it may only be very close.
130 In 2.6+, there are counters for each CPU, which make the lack of locking
131 almost a non-issue. When the statistics are read, the per-CPU counters
132 are summed (possibly overflowing the unsigned long variable they are
133 summed to) and the result given to the user. There is no convenient
134 user interface for accessing the per-CPU counters themselves.
139 There were significant changes between 2.4 and 2.6+ in the I/O subsystem.
140 As a result, some statistic information disappeared. The translation from
141 a disk address relative to a partition to the disk address relative to
142 the host disk happens much earlier. All merges and timings now happen
143 at the disk level rather than at both the disk and partition level as
144 in 2.4. Consequently, you'll see a different statistics output on 2.6+ for
145 partitions from that for disks. There are only *four* fields available
146 for partitions on 2.6+ machines. This is reflected in the examples above.
148 Field 1 -- # of reads issued
149 This is the total number of reads issued to this partition.
151 Field 2 -- # of sectors read
152 This is the total number of sectors requested to be read from this
155 Field 3 -- # of writes issued
156 This is the total number of writes issued to this partition.
158 Field 4 -- # of sectors written
159 This is the total number of sectors requested to be written to
162 Note that since the address is translated to a disk-relative one, and no
163 record of the partition-relative address is kept, the subsequent success
164 or failure of the read cannot be attributed to the partition. In other
165 words, the number of reads for partitions is counted slightly before time
166 of queuing for partitions, and at completion for whole disks. This is
167 a subtle distinction that is probably uninteresting for most cases.
169 More significant is the error induced by counting the numbers of
170 reads/writes before merges for partitions and after for disks. Since a
171 typical workload usually contains a lot of successive and adjacent requests,
172 the number of reads/writes issued can be several times higher than the
173 number of reads/writes completed.
175 In 2.6.25, the full statistic set is again available for partitions and
176 disk and partition statistics are consistent again. Since we still don't
177 keep record of the partition-relative address, an operation is attributed to
178 the partition which contains the first sector of the request after the
179 eventual merges. As requests can be merged across partition, this could lead
180 to some (probably insignificant) inaccuracy.
185 In 2.6+, sysfs is not mounted by default. If your distribution of
186 Linux hasn't added it already, here's the line you'll want to add to
187 your ``/etc/fstab``::
189 none /sys sysfs defaults 0 0
192 In 2.6+, all disk statistics were removed from ``/proc/stat``. In 2.4, they
193 appear in both ``/proc/partitions`` and ``/proc/stat``, although the ones in
194 ``/proc/stat`` take a very different format from those in ``/proc/partitions``
195 (see proc(5), if your system has it.)
197 -- ricklind@us.ibm.com