1=================== 2Block IO Controller 3=================== 4 5Overview 6======== 7cgroup subsys "blkio" implements the block io controller. There seems to be 8a 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 controller 11and based on user options switch IO policies in the background. 12 13One IO control policy is throttling policy which can be used to 14specify upper IO rate limits on devices. This policy is implemented in 15generic block layer and can be used on leaf nodes as well as higher 16level logical devices like device mapper. 17 18HOWTO 19===== 20 21Throttling/Upper Limit policy 22----------------------------- 23Enable Block IO controller:: 24 25 CONFIG_BLK_CGROUP=y 26 27Enable throttling in block layer:: 28 29 CONFIG_BLK_DEV_THROTTLING=y 30 31Mount blkio controller (see cgroups.txt, Why are cgroups needed?):: 32 33 mount -t cgroup -o blkio none /sys/fs/cgroup/blkio 34 35Specify a bandwidth rate on particular device for root group. The format 36for policy is "<major>:<minor> <bytes_per_second>":: 37 38 echo "8:16 1048576" > /sys/fs/cgroup/blkio/blkio.throttle.read_bps_device 39 40This will put a limit of 1MB/second on reads happening for root group 41on 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=1024 46 1024+0 records in 47 1024+0 records out 48 4194304 bytes (4.2 MB) copied, 4.0001 s, 1.0 MB/s 49 50Limits for writes can be put using blkio.throttle.write_bps_device file. 51 52Hierarchical Cgroups 53==================== 54 55Throttling implements hierarchy support; however, 56throttling's hierarchy support is enabled iff "sane_behavior" is 57enabled from cgroup side, which currently is a development option and 58not publicly available. 59 60If somebody created a hierarchy like as follows:: 61 62 root 63 / \ 64 test1 test2 65 | 66 test3 67 68Throttling with "sane_behavior" will handle the 69hierarchy correctly. For throttling, all limits apply 70to the whole subtree while all statistics are local to the IOs 71directly generated by tasks in that cgroup. 72 73Throttling without "sane_behavior" enabled from cgroup side will 74practically treat all groups at same level as if it looks like the 75following:: 76 77 pivot 78 / / \ \ 79 root test1 test2 test3 80 81Various user visible config options 82=================================== 83 84 CONFIG_BLK_CGROUP 85 Block IO controller. 86 87 CONFIG_BFQ_CGROUP_DEBUG 88 Debug help. Right now some additional stats file show up in cgroup 89 if this option is enabled. 90 91 CONFIG_BLK_DEV_THROTTLING 92 Enable block device throttling support in block layer. 93 94Details of cgroup files 95======================= 96 97Proportional weight policy files 98-------------------------------- 99 100 blkio.bfq.weight 101 Specifies per cgroup weight. This is default weight of the group 102 on all the devices until and unless overridden by per device rule 103 (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_device 109 Specifies per cgroup per device weights, overriding the default group 110 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_device 115 116 Configure weight=300 on /dev/sdb (8:16) in this cgroup:: 117 118 # echo 8:16 300 > blkio.bfq.weight_device 119 # cat blkio.bfq.weight_device 120 dev weight 121 8:16 300 122 123 Configure weight=500 on /dev/sda (8:0) in this cgroup:: 124 125 # echo 8:0 500 > blkio.bfq.weight_device 126 # cat blkio.bfq.weight_device 127 dev weight 128 8:0 500 129 8:16 300 130 131 Remove specific weight for /dev/sda in this cgroup:: 132 133 # echo 8:0 0 > blkio.bfq.weight_device 134 # cat blkio.bfq.weight_device 135 dev weight 136 8:16 300 137 138 blkio.time 139 Disk time allocated to cgroup per device in milliseconds. First 140 two fields specify the major and minor number of the device and 141 third field specifies the disk time allocated to group in 142 milliseconds. 143 144 blkio.sectors 145 Number of sectors transferred to/from disk by the group. First 146 two fields specify the major and minor number of the device and 147 third field specifies the number of sectors transferred by the 148 group to/from the device. 149 150 blkio.io_service_bytes 151 Number of bytes transferred to/from the disk by the group. These 152 are further divided by the type of operation - read or write, sync 153 or async. First two fields specify the major and minor number of the 154 device, third field specifies the operation type and the fourth field 155 specifies the number of bytes. 156 157 blkio.io_serviced 158 Number of IOs (bio) issued to the disk by the group. These 159 are further divided by the type of operation - read or write, sync 160 or async. First two fields specify the major and minor number of the 161 device, third field specifies the operation type and the fourth field 162 specifies the number of IOs. 163 164 blkio.io_service_time 165 Total amount of time between request dispatch and request completion 166 for the IOs done by this cgroup. This is in nanoseconds to make it 167 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. This 170 may cause the service time for a given IO to include the service time 171 of multiple IOs when served out of order which may result in total 172 io_service_time > actual time elapsed. This time is further divided by 173 the type of operation - read or write, sync or async. First two fields 174 specify the major and minor number of the device, third field 175 specifies the operation type and the fourth field specifies the 176 io_service_time in ns. 177 178 blkio.io_wait_time 179 Total amount of time the IOs for this cgroup spent waiting in the 180 scheduler queues for service. This can be greater than the total time 181 elapsed since it is cumulative io_wait_time for all IOs. It is not a 182 measure of total time the cgroup spent waiting but rather a measure of 183 the wait_time for its individual IOs. For devices with queue_depth > 1 184 this metric does not include the time spent waiting for service once 185 the IO is dispatched to the device but till it actually gets serviced 186 (there might be a time lag here due to re-ordering of requests by the 187 device). This is in nanoseconds to make it meaningful for flash 188 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 and 190 minor number of the device, third field specifies the operation type 191 and the fourth field specifies the io_wait_time in ns. 192 193 blkio.io_merged 194 Total number of bios/requests merged into requests belonging to this 195 cgroup. This is further divided by the type of operation - read or 196 write, sync or async. 197 198 blkio.io_queued 199 Total number of requests queued up at any given instant for this 200 cgroup. This is further divided by the type of operation - read or 201 write, sync or async. 202 203 blkio.avg_queue_size 204 Debugging aid only enabled if CONFIG_BFQ_CGROUP_DEBUG=y. 205 The average queue size for this cgroup over the entire time of this 206 cgroup's existence. Queue size samples are taken each time one of the 207 queues of this cgroup gets a timeslice. 208 209 blkio.group_wait_time 210 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 busy 212 (i.e., went from 0 to 1 request queued) to get a timeslice for one of 213 its queues. This is different from the io_wait_time which is the 214 cumulative total of the amount of time spent by each IO in that cgroup 215 waiting in the scheduler queue. This is in nanoseconds. If this is 216 read when the cgroup is in a waiting (for timeslice) state, the stat 217 will only report the group_wait_time accumulated till the last time it 218 got a timeslice and will not include the current delta. 219 220 blkio.empty_time 221 Debugging aid only enabled if CONFIG_BFQ_CGROUP_DEBUG=y. 222 This is the amount of time a cgroup spends without any pending 223 requests when not being served, i.e., it does not include any time 224 spent idling for one of the queues of the cgroup. This is in 225 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 last 227 time it had a pending request and will not include the current delta. 228 229 blkio.idle_time 230 Debugging aid only enabled if CONFIG_BFQ_CGROUP_DEBUG=y. 231 This is the amount of time spent by the IO scheduler idling for a 232 given cgroup in anticipation of a better request than the existing ones 233 from other queues/cgroups. This is in nanoseconds. If this is read 234 when the cgroup is in an idling state, the stat will only report the 235 idle_time accumulated till the last idle period and will not include 236 the current delta. 237 238 blkio.dequeue 239 Debugging aid only enabled if CONFIG_BFQ_CGROUP_DEBUG=y. This 240 gives the statistics about how many a times a group was dequeued 241 from service tree of the device. First two fields specify the major 242 and minor number of the device and third field specifies the number 243 of times a group was dequeued from a particular device. 244 245 blkio.*_recursive 246 Recursive version of various stats. These files show the 247 same information as their non-recursive counterparts but 248 include stats from all the descendant cgroups. 249 250Throttling/Upper limit policy files 251----------------------------------- 252 blkio.throttle.read_bps_device 253 Specifies upper limit on READ rate from the device. IO rate is 254 specified in bytes per second. Rules are per device. Following is 255 the format:: 256 257 echo "<major>:<minor> <rate_bytes_per_second>" > /cgrp/blkio.throttle.read_bps_device 258 259 blkio.throttle.write_bps_device 260 Specifies upper limit on WRITE rate to the device. IO rate is 261 specified in bytes per second. Rules are per device. Following is 262 the format:: 263 264 echo "<major>:<minor> <rate_bytes_per_second>" > /cgrp/blkio.throttle.write_bps_device 265 266 blkio.throttle.read_iops_device 267 Specifies upper limit on READ rate from the device. IO rate is 268 specified in IO per second. Rules are per device. Following is 269 the format:: 270 271 echo "<major>:<minor> <rate_io_per_second>" > /cgrp/blkio.throttle.read_iops_device 272 273 blkio.throttle.write_iops_device 274 Specifies upper limit on WRITE rate to the device. IO rate is 275 specified in io per second. Rules are per device. Following is 276 the format:: 277 278 echo "<major>:<minor> <rate_io_per_second>" > /cgrp/blkio.throttle.write_iops_device 279 280 Note: If both BW and IOPS rules are specified for a device, then IO is 281 subjected to both the constraints. 282 283 blkio.throttle.io_serviced 284 Number of IOs (bio) issued to the disk by the group. These 285 are further divided by the type of operation - read or write, sync 286 or async. First two fields specify the major and minor number of the 287 device, third field specifies the operation type and the fourth field 288 specifies the number of IOs. 289 290 blkio.throttle.io_service_bytes 291 Number of bytes transferred to/from the disk by the group. These 292 are further divided by the type of operation - read or write, sync 293 or async. First two fields specify the major and minor number of the 294 device, third field specifies the operation type and the fourth field 295 specifies the number of bytes. 296 297Common files among various policies 298----------------------------------- 299 blkio.reset_stats 300 Writing an int to this file will result in resetting all the stats 301 for that cgroup. 302