iostats.rst 9.7 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210
  1. =====================
  2. I/O statistics fields
  3. =====================
  4. Since 2.4.20 (and some versions before, with patches), and 2.5.45,
  5. more extensive disk statistics have been introduced to help measure disk
  6. activity. Tools such as ``sar`` and ``iostat`` typically interpret these and do
  7. the work for you, but in case you are interested in creating your own
  8. tools, the fields are explained here.
  9. In 2.4 now, the information is found as additional fields in
  10. ``/proc/partitions``. In 2.6 and upper, the same information is found in two
  11. places: one is in the file ``/proc/diskstats``, and the other is within
  12. the sysfs file system, which must be mounted in order to obtain
  13. the information. Throughout this document we'll assume that sysfs
  14. is mounted on ``/sys``, although of course it may be mounted anywhere.
  15. Both ``/proc/diskstats`` and sysfs use the same source for the information
  16. and so should not differ.
  17. Here are examples of these different formats::
  18. 2.4:
  19. 3 0 39082680 hda 446216 784926 9550688 4382310 424847 312726 5922052 19310380 0 3376340 23705160
  20. 3 1 9221278 hda1 35486 0 35496 38030 0 0 0 0 0 38030 38030
  21. 2.6+ sysfs:
  22. 446216 784926 9550688 4382310 424847 312726 5922052 19310380 0 3376340 23705160
  23. 35486 38030 38030 38030
  24. 2.6+ diskstats:
  25. 3 0 hda 446216 784926 9550688 4382310 424847 312726 5922052 19310380 0 3376340 23705160
  26. 3 1 hda1 35486 38030 38030 38030
  27. 4.18+ diskstats:
  28. 3 0 hda 446216 784926 9550688 4382310 424847 312726 5922052 19310380 0 3376340 23705160 0 0 0 0
  29. On 2.4 you might execute ``grep 'hda ' /proc/partitions``. On 2.6+, you have
  30. a choice of ``cat /sys/block/hda/stat`` or ``grep 'hda ' /proc/diskstats``.
  31. The advantage of one over the other is that the sysfs choice works well
  32. if you are watching a known, small set of disks. ``/proc/diskstats`` may
  33. be a better choice if you are watching a large number of disks because
  34. you'll avoid the overhead of 50, 100, or 500 or more opens/closes with
  35. each snapshot of your disk statistics.
  36. In 2.4, the statistics fields are those after the device name. In
  37. the above example, the first field of statistics would be 446216.
  38. By contrast, in 2.6+ if you look at ``/sys/block/hda/stat``, you'll
  39. find just the 15 fields, beginning with 446216. If you look at
  40. ``/proc/diskstats``, the 15 fields will be preceded by the major and
  41. minor device numbers, and device name. Each of these formats provides
  42. 15 fields of statistics, each meaning exactly the same things.
  43. All fields except field 9 are cumulative since boot. Field 9 should
  44. go to zero as I/Os complete; all others only increase (unless they
  45. overflow and wrap). Wrapping might eventually occur on a very busy
  46. or long-lived system; so applications should be prepared to deal with
  47. it. Regarding wrapping, the types of the fields are either unsigned
  48. int (32 bit) or unsigned long (32-bit or 64-bit, depending on your
  49. machine) as noted per-field below. Unless your observations are very
  50. spread in time, these fields should not wrap twice before you notice it.
  51. Each set of stats only applies to the indicated device; if you want
  52. system-wide stats you'll have to find all the devices and sum them all up.
  53. Field 1 -- # of reads completed (unsigned long)
  54. This is the total number of reads completed successfully.
  55. Field 2 -- # of reads merged, field 6 -- # of writes merged (unsigned long)
  56. Reads and writes which are adjacent to each other may be merged for
  57. efficiency. Thus two 4K reads may become one 8K read before it is
  58. ultimately handed to the disk, and so it will be counted (and queued)
  59. as only one I/O. This field lets you know how often this was done.
  60. Field 3 -- # of sectors read (unsigned long)
  61. This is the total number of sectors read successfully.
  62. Field 4 -- # of milliseconds spent reading (unsigned int)
  63. This is the total number of milliseconds spent by all reads (as
  64. measured from __make_request() to end_that_request_last()).
  65. Field 5 -- # of writes completed (unsigned long)
  66. This is the total number of writes completed successfully.
  67. Field 6 -- # of writes merged (unsigned long)
  68. See the description of field 2.
  69. Field 7 -- # of sectors written (unsigned long)
  70. This is the total number of sectors written successfully.
  71. Field 8 -- # of milliseconds spent writing (unsigned int)
  72. This is the total number of milliseconds spent by all writes (as
  73. measured from __make_request() to end_that_request_last()).
  74. Field 9 -- # of I/Os currently in progress (unsigned int)
  75. The only field that should go to zero. Incremented as requests are
  76. given to appropriate struct request_queue and decremented as they finish.
  77. Field 10 -- # of milliseconds spent doing I/Os (unsigned int)
  78. This field increases so long as field 9 is nonzero.
  79. Since 5.0 this field counts jiffies when at least one request was
  80. started or completed. If request runs more than 2 jiffies then some
  81. I/O time might be not accounted in case of concurrent requests.
  82. Field 11 -- weighted # of milliseconds spent doing I/Os (unsigned int)
  83. This field is incremented at each I/O start, I/O completion, I/O
  84. merge, or read of these stats by the number of I/Os in progress
  85. (field 9) times the number of milliseconds spent doing I/O since the
  86. last update of this field. This can provide an easy measure of both
  87. I/O completion time and the backlog that may be accumulating.
  88. Field 12 -- # of discards completed (unsigned long)
  89. This is the total number of discards completed successfully.
  90. Field 13 -- # of discards merged (unsigned long)
  91. See the description of field 2
  92. Field 14 -- # of sectors discarded (unsigned long)
  93. This is the total number of sectors discarded successfully.
  94. Field 15 -- # of milliseconds spent discarding (unsigned int)
  95. This is the total number of milliseconds spent by all discards (as
  96. measured from __make_request() to end_that_request_last()).
  97. Field 16 -- # of flush requests completed
  98. This is the total number of flush requests completed successfully.
  99. Block layer combines flush requests and executes at most one at a time.
  100. This counts flush requests executed by disk. Not tracked for partitions.
  101. Field 17 -- # of milliseconds spent flushing
  102. This is the total number of milliseconds spent by all flush requests.
  103. To avoid introducing performance bottlenecks, no locks are held while
  104. modifying these counters. This implies that minor inaccuracies may be
  105. introduced when changes collide, so (for instance) adding up all the
  106. read I/Os issued per partition should equal those made to the disks ...
  107. but due to the lack of locking it may only be very close.
  108. In 2.6+, there are counters for each CPU, which make the lack of locking
  109. almost a non-issue. When the statistics are read, the per-CPU counters
  110. are summed (possibly overflowing the unsigned long variable they are
  111. summed to) and the result given to the user. There is no convenient
  112. user interface for accessing the per-CPU counters themselves.
  113. Since 4.19 request times are measured with nanoseconds precision and
  114. truncated to milliseconds before showing in this interface.
  115. Disks vs Partitions
  116. -------------------
  117. There were significant changes between 2.4 and 2.6+ in the I/O subsystem.
  118. As a result, some statistic information disappeared. The translation from
  119. a disk address relative to a partition to the disk address relative to
  120. the host disk happens much earlier. All merges and timings now happen
  121. at the disk level rather than at both the disk and partition level as
  122. in 2.4. Consequently, you'll see a different statistics output on 2.6+ for
  123. partitions from that for disks. There are only *four* fields available
  124. for partitions on 2.6+ machines. This is reflected in the examples above.
  125. Field 1 -- # of reads issued
  126. This is the total number of reads issued to this partition.
  127. Field 2 -- # of sectors read
  128. This is the total number of sectors requested to be read from this
  129. partition.
  130. Field 3 -- # of writes issued
  131. This is the total number of writes issued to this partition.
  132. Field 4 -- # of sectors written
  133. This is the total number of sectors requested to be written to
  134. this partition.
  135. Note that since the address is translated to a disk-relative one, and no
  136. record of the partition-relative address is kept, the subsequent success
  137. or failure of the read cannot be attributed to the partition. In other
  138. words, the number of reads for partitions is counted slightly before time
  139. of queuing for partitions, and at completion for whole disks. This is
  140. a subtle distinction that is probably uninteresting for most cases.
  141. More significant is the error induced by counting the numbers of
  142. reads/writes before merges for partitions and after for disks. Since a
  143. typical workload usually contains a lot of successive and adjacent requests,
  144. the number of reads/writes issued can be several times higher than the
  145. number of reads/writes completed.
  146. In 2.6.25, the full statistic set is again available for partitions and
  147. disk and partition statistics are consistent again. Since we still don't
  148. keep record of the partition-relative address, an operation is attributed to
  149. the partition which contains the first sector of the request after the
  150. eventual merges. As requests can be merged across partition, this could lead
  151. to some (probably insignificant) inaccuracy.
  152. Additional notes
  153. ----------------
  154. In 2.6+, sysfs is not mounted by default. If your distribution of
  155. Linux hasn't added it already, here's the line you'll want to add to
  156. your ``/etc/fstab``::
  157. none /sys sysfs defaults 0 0
  158. In 2.6+, all disk statistics were removed from ``/proc/stat``. In 2.4, they
  159. appear in both ``/proc/partitions`` and ``/proc/stat``, although the ones in
  160. ``/proc/stat`` take a very different format from those in ``/proc/partitions``
  161. (see proc(5), if your system has it.)
  162. -- ricklind@us.ibm.com