Documentation/cgroup-v1/blkio-controller.txt

   1                                 Block IO Controller
   2                                 ===================
   3 Overview
   4 ========
   5 cgroup subsys "blkio" implements the block io controller. There seems to be
   6 a need of various kinds of IO control policies (like proportional BW, max BW)
   7 both at leaf nodes as well as at intermediate nodes in a storage hierarchy.
   8 Plan is to use the same cgroup based management interface for blkio controller
   9 and based on user options switch IO policies in the background.
  10
  11 One IO control policy is throttling policy which can be used to
  12 specify upper IO rate limits on devices. This policy is implemented in
  13 generic block layer and can be used on leaf nodes as well as higher
  14 level logical devices like device mapper.
  15
  16 HOWTO
  17 =====
  18 Throttling/Upper Limit policy
  19 -----------------------------
  20 - Enable Block IO controller
  21         CONFIG_BLK_CGROUP=y
  22
  23 - Enable throttling in block layer
  24         CONFIG_BLK_DEV_THROTTLING=y
  25
  26 - Mount blkio controller (see cgroups.txt, Why are cgroups needed?)
  27         mount -t cgroup -o blkio none /sys/fs/cgroup/blkio
  28
  29 - Specify a bandwidth rate on particular device for root group. The format
  30   for policy is "<major>:<minor>  <bytes_per_second>".
  31
  32         echo "8:16  1048576" > /sys/fs/cgroup/blkio/blkio.throttle.read_bps_device
  33
  34   Above will put a limit of 1MB/second on reads happening for root group
  35   on device having major/minor number 8:16.
  36
  37 - Run dd to read a file and see if rate is throttled to 1MB/s or not.
  38
  39         # dd iflag=direct if=/mnt/common/zerofile of=/dev/null bs=4K count=1024
  40         1024+0 records in
  41         1024+0 records out
  42         4194304 bytes (4.2 MB) copied, 4.0001 s, 1.0 MB/s
  43
  44  Limits for writes can be put using blkio.throttle.write_bps_device file.
  45
  46 Hierarchical Cgroups
  47 ====================
  48
  49 Throttling implements hierarchy support; however,
  50 throttling's hierarchy support is enabled iff "sane_behavior" is
  51 enabled from cgroup side, which currently is a development option and
  52 not publicly available.
  53
  54 If somebody created a hierarchy like as follows.
  55
  56                         root
  57                         /  \
  58                      test1 test2
  59                         |
  60                      test3
  61
  62 Throttling with "sane_behavior" will handle the
  63 hierarchy correctly. For throttling, all limits apply
  64 to the whole subtree while all statistics are local to the IOs
  65 directly generated by tasks in that cgroup.
  66
  67 Throttling without "sane_behavior" enabled from cgroup side will
  68 practically treat all groups at same level as if it looks like the
  69 following.
  70
  71                                 pivot
  72                              /  /   \  \
  73                         root  test1 test2  test3
  74
  75 Various user visible config options
  76 ===================================
  77 CONFIG_BLK_CGROUP
  78         - Block IO controller.
  79
  80 CONFIG_DEBUG_BLK_CGROUP
  81         - Debug help. Right now some additional stats file show up in cgroup
  82           if this option is enabled.
  83
  84 CONFIG_BLK_DEV_THROTTLING
  85         - Enable block device throttling support in block layer.
  86
  87 Details of cgroup files
  88 =======================
  89 Proportional weight policy files
  90 --------------------------------
  91 - blkio.weight
  92         - Specifies per cgroup weight. This is default weight of the group
  93           on all the devices until and unless overridden by per device rule.
  94           (See blkio.weight_device).
  95           Currently allowed range of weights is from 10 to 1000.
  96
  97 - blkio.weight_device
  98         - One can specify per cgroup per device rules using this interface.
  99           These rules override the default value of group weight as specified
 100           by blkio.weight.
 101
 102           Following is the format.
 103
 104           # echo dev_maj:dev_minor weight > blkio.weight_device
 105           Configure weight=300 on /dev/sdb (8:16) in this cgroup
 106           # echo 8:16 300 > blkio.weight_device
 107           # cat blkio.weight_device
 108           dev     weight
 109           8:16    300
 110
 111           Configure weight=500 on /dev/sda (8:0) in this cgroup
 112           # echo 8:0 500 > blkio.weight_device
 113           # cat blkio.weight_device
 114           dev     weight
 115           8:0     500
 116           8:16    300
 117
 118           Remove specific weight for /dev/sda in this cgroup
 119           # echo 8:0 0 > blkio.weight_device
 120           # cat blkio.weight_device
 121           dev     weight
 122           8:16    300
 123
 124 - blkio.leaf_weight[_device]
 125         - Equivalents of blkio.weight[_device] for the purpose of
 126           deciding how much weight tasks in the given cgroup has while
 127           competing with the cgroup's child cgroups. For details,
 128           please refer to Documentation/block/cfq-iosched.txt.
 129
 130 - blkio.time
 131         - disk time allocated to cgroup per device in milliseconds. First
 132           two fields specify the major and minor number of the device and
 133           third field specifies the disk time allocated to group in
 134           milliseconds.
 135
 136 - blkio.sectors
 137         - number of sectors transferred to/from disk by the group. First
 138           two fields specify the major and minor number of the device and
 139           third field specifies the number of sectors transferred by the
 140           group to/from the device.
 141
 142 - blkio.io_service_bytes
 143         - Number of bytes transferred to/from the disk by the group. These
 144           are further divided by the type of operation - read or write, sync
 145           or async. First two fields specify the major and minor number of the
 146           device, third field specifies the operation type and the fourth field
 147           specifies the number of bytes.
 148
 149 - blkio.io_serviced
 150         - Number of IOs (bio) issued to the disk by the group. These
 151           are further divided by the type of operation - read or write, sync
 152           or async. First two fields specify the major and minor number of the
 153           device, third field specifies the operation type and the fourth field
 154           specifies the number of IOs.
 155
 156 - blkio.io_service_time
 157         - Total amount of time between request dispatch and request completion
 158           for the IOs done by this cgroup. This is in nanoseconds to make it
 159           meaningful for flash devices too. For devices with queue depth of 1,
 160           this time represents the actual service time. When queue_depth > 1,
 161           that is no longer true as requests may be served out of order. This
 162           may cause the service time for a given IO to include the service time
 163           of multiple IOs when served out of order which may result in total
 164           io_service_time > actual time elapsed. This time is further divided by
 165           the type of operation - read or write, sync or async. First two fields
 166           specify the major and minor number of the device, third field
 167           specifies the operation type and the fourth field specifies the
 168           io_service_time in ns.
 169
 170 - blkio.io_wait_time
 171         - Total amount of time the IOs for this cgroup spent waiting in the
 172           scheduler queues for service. This can be greater than the total time
 173           elapsed since it is cumulative io_wait_time for all IOs. It is not a
 174           measure of total time the cgroup spent waiting but rather a measure of
 175           the wait_time for its individual IOs. For devices with queue_depth > 1
 176           this metric does not include the time spent waiting for service once
 177           the IO is dispatched to the device but till it actually gets serviced
 178           (there might be a time lag here due to re-ordering of requests by the
 179           device). This is in nanoseconds to make it meaningful for flash
 180           devices too. This time is further divided by the type of operation -
 181           read or write, sync or async. First two fields specify the major and
 182           minor number of the device, third field specifies the operation type
 183           and the fourth field specifies the io_wait_time in ns.
 184
 185 - blkio.io_merged
 186         - Total number of bios/requests merged into requests belonging to this
 187           cgroup. This is further divided by the type of operation - read or
 188           write, sync or async.
 189
 190 - blkio.io_queued
 191         - Total number of requests queued up at any given instant for this
 192           cgroup. This is further divided by the type of operation - read or
 193           write, sync or async.
 194
 195 - blkio.avg_queue_size
 196         - Debugging aid only enabled if CONFIG_DEBUG_BLK_CGROUP=y.
 197           The average queue size for this cgroup over the entire time of this
 198           cgroup's existence. Queue size samples are taken each time one of the
 199           queues of this cgroup gets a timeslice.
 200
 201 - blkio.group_wait_time
 202         - Debugging aid only enabled if CONFIG_DEBUG_BLK_CGROUP=y.
 203           This is the amount of time the cgroup had to wait since it became busy
 204           (i.e., went from 0 to 1 request queued) to get a timeslice for one of
 205           its queues. This is different from the io_wait_time which is the
 206           cumulative total of the amount of time spent by each IO in that cgroup
 207           waiting in the scheduler queue. This is in nanoseconds. If this is
 208           read when the cgroup is in a waiting (for timeslice) state, the stat
 209           will only report the group_wait_time accumulated till the last time it
 210           got a timeslice and will not include the current delta.
 211
 212 - blkio.empty_time
 213         - Debugging aid only enabled if CONFIG_DEBUG_BLK_CGROUP=y.
 214           This is the amount of time a cgroup spends without any pending
 215           requests when not being served, i.e., it does not include any time
 216           spent idling for one of the queues of the cgroup. This is in
 217           nanoseconds. If this is read when the cgroup is in an empty state,
 218           the stat will only report the empty_time accumulated till the last
 219           time it had a pending request and will not include the current delta.
 220
 221 - blkio.idle_time
 222         - Debugging aid only enabled if CONFIG_DEBUG_BLK_CGROUP=y.
 223           This is the amount of time spent by the IO scheduler idling for a
 224           given cgroup in anticipation of a better request than the existing ones
 225           from other queues/cgroups. This is in nanoseconds. If this is read
 226           when the cgroup is in an idling state, the stat will only report the
 227           idle_time accumulated till the last idle period and will not include
 228           the current delta.
 229
 230 - blkio.dequeue
 231         - Debugging aid only enabled if CONFIG_DEBUG_BLK_CGROUP=y. This
 232           gives the statistics about how many a times a group was dequeued
 233           from service tree of the device. First two fields specify the major
 234           and minor number of the device and third field specifies the number
 235           of times a group was dequeued from a particular device.
 236
 237 - blkio.*_recursive
 238         - Recursive version of various stats. These files show the
 239           same information as their non-recursive counterparts but
 240           include stats from all the descendant cgroups.
 241
 242 Throttling/Upper limit policy files
 243 -----------------------------------
 244 - blkio.throttle.read_bps_device
 245         - Specifies upper limit on READ rate from the device. IO rate is
 246           specified in bytes per second. Rules are per device. Following is
 247           the format.
 248
 249   echo "<major>:<minor>  <rate_bytes_per_second>" > /cgrp/blkio.throttle.read_bps_device
 250
 251 - blkio.throttle.write_bps_device
 252         - Specifies upper limit on WRITE rate to the device. IO rate is
 253           specified in bytes per second. Rules are per device. Following is
 254           the format.
 255
 256   echo "<major>:<minor>  <rate_bytes_per_second>" > /cgrp/blkio.throttle.write_bps_device
 257
 258 - blkio.throttle.read_iops_device
 259         - Specifies upper limit on READ rate from the device. IO rate is
 260           specified in IO per second. Rules are per device. Following is
 261           the format.
 262
 263   echo "<major>:<minor>  <rate_io_per_second>" > /cgrp/blkio.throttle.read_iops_device
 264
 265 - blkio.throttle.write_iops_device
 266         - Specifies upper limit on WRITE rate to the device. IO rate is
 267           specified in io per second. Rules are per device. Following is
 268           the format.
 269
 270   echo "<major>:<minor>  <rate_io_per_second>" > /cgrp/blkio.throttle.write_iops_device
 271
 272 Note: If both BW and IOPS rules are specified for a device, then IO is
 273       subjected to both the constraints.
 274
 275 - blkio.throttle.io_serviced
 276         - Number of IOs (bio) issued to the disk by the group. These
 277           are further divided by the type of operation - read or write, sync
 278           or async. First two fields specify the major and minor number of the
 279           device, third field specifies the operation type and the fourth field
 280           specifies the number of IOs.
 281
 282 - blkio.throttle.io_service_bytes
 283         - Number of bytes transferred to/from the disk by the group. These
 284           are further divided by the type of operation - read or write, sync
 285           or async. First two fields specify the major and minor number of the
 286           device, third field specifies the operation type and the fourth field
 287           specifies the number of bytes.
 288
 289 Common files among various policies
 290 -----------------------------------
 291 - blkio.reset_stats
 292         - Writing an int to this file will result in resetting all the stats
 293           for that cgroup.