302 lines · plain
1===================2Block IO Controller3===================4 5Overview6========7cgroup subsys "blkio" implements the block io controller. There seems to be8a need of various kinds of IO control policies (like proportional BW, max BW)9both at leaf nodes as well as at intermediate nodes in a storage hierarchy.10Plan is to use the same cgroup based management interface for blkio controller11and based on user options switch IO policies in the background.12 13One IO control policy is throttling policy which can be used to14specify upper IO rate limits on devices. This policy is implemented in15generic block layer and can be used on leaf nodes as well as higher16level logical devices like device mapper.17 18HOWTO19=====20 21Throttling/Upper Limit policy22-----------------------------23Enable Block IO controller::24 25 CONFIG_BLK_CGROUP=y26 27Enable throttling in block layer::28 29 CONFIG_BLK_DEV_THROTTLING=y30 31Mount blkio controller (see cgroups.txt, Why are cgroups needed?)::32 33 mount -t cgroup -o blkio none /sys/fs/cgroup/blkio34 35Specify a bandwidth rate on particular device for root group. The format36for policy is "<major>:<minor> <bytes_per_second>"::37 38 echo "8:16 1048576" > /sys/fs/cgroup/blkio/blkio.throttle.read_bps_device39 40This will put a limit of 1MB/second on reads happening for root group41on device having major/minor number 8:16.42 43Run dd to read a file and see if rate is throttled to 1MB/s or not::44 45 # dd iflag=direct if=/mnt/common/zerofile of=/dev/null bs=4K count=102446 1024+0 records in47 1024+0 records out48 4194304 bytes (4.2 MB) copied, 4.0001 s, 1.0 MB/s49 50Limits for writes can be put using blkio.throttle.write_bps_device file.51 52Hierarchical Cgroups53====================54 55Throttling implements hierarchy support; however,56throttling's hierarchy support is enabled iff "sane_behavior" is57enabled from cgroup side, which currently is a development option and58not publicly available.59 60If somebody created a hierarchy like as follows::61 62 root63 / \64 test1 test265 |66 test367 68Throttling with "sane_behavior" will handle the69hierarchy correctly. For throttling, all limits apply70to the whole subtree while all statistics are local to the IOs71directly generated by tasks in that cgroup.72 73Throttling without "sane_behavior" enabled from cgroup side will74practically treat all groups at same level as if it looks like the75following::76 77 pivot78 / / \ \79 root test1 test2 test380 81Various user visible config options82===================================83 84 CONFIG_BLK_CGROUP85 Block IO controller.86 87 CONFIG_BFQ_CGROUP_DEBUG88 Debug help. Right now some additional stats file show up in cgroup89 if this option is enabled.90 91 CONFIG_BLK_DEV_THROTTLING92 Enable block device throttling support in block layer.93 94Details of cgroup files95=======================96 97Proportional weight policy files98--------------------------------99 100 blkio.bfq.weight101 Specifies per cgroup weight. This is default weight of the group102 on all the devices until and unless overridden by per device rule103 (see `blkio.bfq.weight_device` below).104 105 Currently allowed range of weights is from 1 to 1000. For more details,106 see Documentation/block/bfq-iosched.rst.107 108 blkio.bfq.weight_device109 Specifies per cgroup per device weights, overriding the default group110 weight. For more details, see Documentation/block/bfq-iosched.rst.111 112 Following is the format::113 114 # echo dev_maj:dev_minor weight > blkio.bfq.weight_device115 116 Configure weight=300 on /dev/sdb (8:16) in this cgroup::117 118 # echo 8:16 300 > blkio.bfq.weight_device119 # cat blkio.bfq.weight_device120 dev weight121 8:16 300122 123 Configure weight=500 on /dev/sda (8:0) in this cgroup::124 125 # echo 8:0 500 > blkio.bfq.weight_device126 # cat blkio.bfq.weight_device127 dev weight128 8:0 500129 8:16 300130 131 Remove specific weight for /dev/sda in this cgroup::132 133 # echo 8:0 0 > blkio.bfq.weight_device134 # cat blkio.bfq.weight_device135 dev weight136 8:16 300137 138 blkio.time139 Disk time allocated to cgroup per device in milliseconds. First140 two fields specify the major and minor number of the device and141 third field specifies the disk time allocated to group in142 milliseconds.143 144 blkio.sectors145 Number of sectors transferred to/from disk by the group. First146 two fields specify the major and minor number of the device and147 third field specifies the number of sectors transferred by the148 group to/from the device.149 150 blkio.io_service_bytes151 Number of bytes transferred to/from the disk by the group. These152 are further divided by the type of operation - read or write, sync153 or async. First two fields specify the major and minor number of the154 device, third field specifies the operation type and the fourth field155 specifies the number of bytes.156 157 blkio.io_serviced158 Number of IOs (bio) issued to the disk by the group. These159 are further divided by the type of operation - read or write, sync160 or async. First two fields specify the major and minor number of the161 device, third field specifies the operation type and the fourth field162 specifies the number of IOs.163 164 blkio.io_service_time165 Total amount of time between request dispatch and request completion166 for the IOs done by this cgroup. This is in nanoseconds to make it167 meaningful for flash devices too. For devices with queue depth of 1,168 this time represents the actual service time. When queue_depth > 1,169 that is no longer true as requests may be served out of order. This170 may cause the service time for a given IO to include the service time171 of multiple IOs when served out of order which may result in total172 io_service_time > actual time elapsed. This time is further divided by173 the type of operation - read or write, sync or async. First two fields174 specify the major and minor number of the device, third field175 specifies the operation type and the fourth field specifies the176 io_service_time in ns.177 178 blkio.io_wait_time179 Total amount of time the IOs for this cgroup spent waiting in the180 scheduler queues for service. This can be greater than the total time181 elapsed since it is cumulative io_wait_time for all IOs. It is not a182 measure of total time the cgroup spent waiting but rather a measure of183 the wait_time for its individual IOs. For devices with queue_depth > 1184 this metric does not include the time spent waiting for service once185 the IO is dispatched to the device but till it actually gets serviced186 (there might be a time lag here due to re-ordering of requests by the187 device). This is in nanoseconds to make it meaningful for flash188 devices too. This time is further divided by the type of operation -189 read or write, sync or async. First two fields specify the major and190 minor number of the device, third field specifies the operation type191 and the fourth field specifies the io_wait_time in ns.192 193 blkio.io_merged194 Total number of bios/requests merged into requests belonging to this195 cgroup. This is further divided by the type of operation - read or196 write, sync or async.197 198 blkio.io_queued199 Total number of requests queued up at any given instant for this200 cgroup. This is further divided by the type of operation - read or201 write, sync or async.202 203 blkio.avg_queue_size204 Debugging aid only enabled if CONFIG_BFQ_CGROUP_DEBUG=y.205 The average queue size for this cgroup over the entire time of this206 cgroup's existence. Queue size samples are taken each time one of the207 queues of this cgroup gets a timeslice.208 209 blkio.group_wait_time210 Debugging aid only enabled if CONFIG_BFQ_CGROUP_DEBUG=y.211 This is the amount of time the cgroup had to wait since it became busy212 (i.e., went from 0 to 1 request queued) to get a timeslice for one of213 its queues. This is different from the io_wait_time which is the214 cumulative total of the amount of time spent by each IO in that cgroup215 waiting in the scheduler queue. This is in nanoseconds. If this is216 read when the cgroup is in a waiting (for timeslice) state, the stat217 will only report the group_wait_time accumulated till the last time it218 got a timeslice and will not include the current delta.219 220 blkio.empty_time221 Debugging aid only enabled if CONFIG_BFQ_CGROUP_DEBUG=y.222 This is the amount of time a cgroup spends without any pending223 requests when not being served, i.e., it does not include any time224 spent idling for one of the queues of the cgroup. This is in225 nanoseconds. If this is read when the cgroup is in an empty state,226 the stat will only report the empty_time accumulated till the last227 time it had a pending request and will not include the current delta.228 229 blkio.idle_time230 Debugging aid only enabled if CONFIG_BFQ_CGROUP_DEBUG=y.231 This is the amount of time spent by the IO scheduler idling for a232 given cgroup in anticipation of a better request than the existing ones233 from other queues/cgroups. This is in nanoseconds. If this is read234 when the cgroup is in an idling state, the stat will only report the235 idle_time accumulated till the last idle period and will not include236 the current delta.237 238 blkio.dequeue239 Debugging aid only enabled if CONFIG_BFQ_CGROUP_DEBUG=y. This240 gives the statistics about how many a times a group was dequeued241 from service tree of the device. First two fields specify the major242 and minor number of the device and third field specifies the number243 of times a group was dequeued from a particular device.244 245 blkio.*_recursive246 Recursive version of various stats. These files show the247 same information as their non-recursive counterparts but248 include stats from all the descendant cgroups.249 250Throttling/Upper limit policy files251-----------------------------------252 blkio.throttle.read_bps_device253 Specifies upper limit on READ rate from the device. IO rate is254 specified in bytes per second. Rules are per device. Following is255 the format::256 257 echo "<major>:<minor> <rate_bytes_per_second>" > /cgrp/blkio.throttle.read_bps_device258 259 blkio.throttle.write_bps_device260 Specifies upper limit on WRITE rate to the device. IO rate is261 specified in bytes per second. Rules are per device. Following is262 the format::263 264 echo "<major>:<minor> <rate_bytes_per_second>" > /cgrp/blkio.throttle.write_bps_device265 266 blkio.throttle.read_iops_device267 Specifies upper limit on READ rate from the device. IO rate is268 specified in IO per second. Rules are per device. Following is269 the format::270 271 echo "<major>:<minor> <rate_io_per_second>" > /cgrp/blkio.throttle.read_iops_device272 273 blkio.throttle.write_iops_device274 Specifies upper limit on WRITE rate to the device. IO rate is275 specified in io per second. Rules are per device. Following is276 the format::277 278 echo "<major>:<minor> <rate_io_per_second>" > /cgrp/blkio.throttle.write_iops_device279 280 Note: If both BW and IOPS rules are specified for a device, then IO is281 subjected to both the constraints.282 283 blkio.throttle.io_serviced284 Number of IOs (bio) issued to the disk by the group. These285 are further divided by the type of operation - read or write, sync286 or async. First two fields specify the major and minor number of the287 device, third field specifies the operation type and the fourth field288 specifies the number of IOs.289 290 blkio.throttle.io_service_bytes291 Number of bytes transferred to/from the disk by the group. These292 are further divided by the type of operation - read or write, sync293 or async. First two fields specify the major and minor number of the294 device, third field specifies the operation type and the fourth field295 specifies the number of bytes.296 297Common files among various policies298-----------------------------------299 blkio.reset_stats300 Writing an int to this file will result in resetting all the stats301 for that cgroup.302