| 1 | ============================
|
| 2 | A block layer cache (bcache)
|
| 3 | ============================
|
| 4 |
|
| 5 | Say you've got a big slow raid 6, and an ssd or three. Wouldn't it be
|
| 6 | nice if you could use them as cache... Hence bcache.
|
| 7 |
|
| 8 | The bcache wiki can be found at:
|
| 9 | https://bcache.evilpiepirate.org
|
| 10 |
|
| 11 | This is the git repository of bcache-tools:
|
| 12 | https://git.kernel.org/pub/scm/linux/kernel/git/colyli/bcache-tools.git/
|
| 13 |
|
| 14 | The latest bcache kernel code can be found from mainline Linux kernel:
|
| 15 | https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/
|
| 16 |
|
| 17 | It's designed around the performance characteristics of SSDs - it only allocates
|
| 18 | in erase block sized buckets, and it uses a hybrid btree/log to track cached
|
| 19 | extents (which can be anywhere from a single sector to the bucket size). It's
|
| 20 | designed to avoid random writes at all costs.
|
| 21 |
|
| 22 | Both writethrough and writeback caching are supported. Writeback defaults to
|
| 23 | off, but can be switched on and off arbitrarily at runtime. Bcache goes to
|
| 24 | great lengths to protect your data - it reliably handles unclean shutdown. (It
|
| 25 | doesn't even have a notion of a clean shutdown; bcache simply doesn't return
|
| 26 | writes as completed until they're on stable storage).
|
| 27 |
|
| 28 | Writeback caching can use most of the cache for buffering writes - writing
|
| 29 | dirty data to the backing device is always done sequentially, scanning from the
|
| 30 | start to the end of the index.
|
| 31 |
|
| 32 | Since random IO is what SSDs excel at, there generally won't be much benefit
|
| 33 | to caching large sequential IO. Bcache detects sequential IO and skips it;
|
| 34 | it also keeps a rolling average of the IO sizes per task, and as long as the
|
| 35 | average is above the cutoff it will skip all IO from that task - instead of
|
| 36 | caching the first 512k after every seek. Backups and large file copies should
|
| 37 | thus entirely bypass the cache.
|
| 38 |
|
| 39 | In the event of a data IO error on the flash it will try to recover by reading
|
| 40 | from disk or invalidating cache entries. For unrecoverable errors (meta data
|
| 41 | or dirty data), caching is automatically disabled; if dirty data was present
|
| 42 | in the cache it first disables writeback caching and waits for all dirty data
|
| 43 | to be flushed.
|
| 44 |
|
| 45 | Getting started:
|
| 46 | You'll need bcache util from the bcache-tools repository. Both the cache device
|
| 47 | and backing device must be formatted before use::
|
| 48 |
|
| 49 | bcache make -B /dev/sdb
|
| 50 | bcache make -C /dev/sdc
|
| 51 |
|
| 52 | `bcache make` has the ability to format multiple devices at the same time - if
|
| 53 | you format your backing devices and cache device at the same time, you won't
|
| 54 | have to manually attach::
|
| 55 |
|
| 56 | bcache make -B /dev/sda /dev/sdb -C /dev/sdc
|
| 57 |
|
| 58 | If your bcache-tools is not updated to latest version and does not have the
|
| 59 | unified `bcache` utility, you may use the legacy `make-bcache` utility to format
|
| 60 | bcache device with same -B and -C parameters.
|
| 61 |
|
| 62 | bcache-tools now ships udev rules, and bcache devices are known to the kernel
|
| 63 | immediately. Without udev, you can manually register devices like this::
|
| 64 |
|
| 65 | echo /dev/sdb > /sys/fs/bcache/register
|
| 66 | echo /dev/sdc > /sys/fs/bcache/register
|
| 67 |
|
| 68 | Registering the backing device makes the bcache device show up in /dev; you can
|
| 69 | now format it and use it as normal. But the first time using a new bcache
|
| 70 | device, it'll be running in passthrough mode until you attach it to a cache.
|
| 71 | If you are thinking about using bcache later, it is recommended to setup all your
|
| 72 | slow devices as bcache backing devices without a cache, and you can choose to add
|
| 73 | a caching device later.
|
| 74 | See 'ATTACHING' section below.
|
| 75 |
|
| 76 | The devices show up as::
|
| 77 |
|
| 78 | /dev/bcache<N>
|
| 79 |
|
| 80 | As well as (with udev)::
|
| 81 |
|
| 82 | /dev/bcache/by-uuid/<uuid>
|
| 83 | /dev/bcache/by-label/<label>
|
| 84 |
|
| 85 | To get started::
|
| 86 |
|
| 87 | mkfs.ext4 /dev/bcache0
|
| 88 | mount /dev/bcache0 /mnt
|
| 89 |
|
| 90 | You can control bcache devices through sysfs at /sys/block/bcache<N>/bcache .
|
| 91 | You can also control them through /sys/fs//bcache/<cset-uuid>/ .
|
| 92 |
|
| 93 | Cache devices are managed as sets; multiple caches per set isn't supported yet
|
| 94 | but will allow for mirroring of metadata and dirty data in the future. Your new
|
| 95 | cache set shows up as /sys/fs/bcache/<UUID>
|
| 96 |
|
| 97 | Attaching
|
| 98 | ---------
|
| 99 |
|
| 100 | After your cache device and backing device are registered, the backing device
|
| 101 | must be attached to your cache set to enable caching. Attaching a backing
|
| 102 | device to a cache set is done thusly, with the UUID of the cache set in
|
| 103 | /sys/fs/bcache::
|
| 104 |
|
| 105 | echo <CSET-UUID> > /sys/block/bcache0/bcache/attach
|
| 106 |
|
| 107 | This only has to be done once. The next time you reboot, just reregister all
|
| 108 | your bcache devices. If a backing device has data in a cache somewhere, the
|
| 109 | /dev/bcache<N> device won't be created until the cache shows up - particularly
|
| 110 | important if you have writeback caching turned on.
|
| 111 |
|
| 112 | If you're booting up and your cache device is gone and never coming back, you
|
| 113 | can force run the backing device::
|
| 114 |
|
| 115 | echo 1 > /sys/block/sdb/bcache/running
|
| 116 |
|
| 117 | (You need to use /sys/block/sdb (or whatever your backing device is called), not
|
| 118 | /sys/block/bcache0, because bcache0 doesn't exist yet. If you're using a
|
| 119 | partition, the bcache directory would be at /sys/block/sdb/sdb2/bcache)
|
| 120 |
|
| 121 | The backing device will still use that cache set if it shows up in the future,
|
| 122 | but all the cached data will be invalidated. If there was dirty data in the
|
| 123 | cache, don't expect the filesystem to be recoverable - you will have massive
|
| 124 | filesystem corruption, though ext4's fsck does work miracles.
|
| 125 |
|
| 126 | Error Handling
|
| 127 | --------------
|
| 128 |
|
| 129 | Bcache tries to transparently handle IO errors to/from the cache device without
|
| 130 | affecting normal operation; if it sees too many errors (the threshold is
|
| 131 | configurable, and defaults to 0) it shuts down the cache device and switches all
|
| 132 | the backing devices to passthrough mode.
|
| 133 |
|
| 134 | - For reads from the cache, if they error we just retry the read from the
|
| 135 | backing device.
|
| 136 |
|
| 137 | - For writethrough writes, if the write to the cache errors we just switch to
|
| 138 | invalidating the data at that lba in the cache (i.e. the same thing we do for
|
| 139 | a write that bypasses the cache)
|
| 140 |
|
| 141 | - For writeback writes, we currently pass that error back up to the
|
| 142 | filesystem/userspace. This could be improved - we could retry it as a write
|
| 143 | that skips the cache so we don't have to error the write.
|
| 144 |
|
| 145 | - When we detach, we first try to flush any dirty data (if we were running in
|
| 146 | writeback mode). It currently doesn't do anything intelligent if it fails to
|
| 147 | read some of the dirty data, though.
|
| 148 |
|
| 149 |
|
| 150 | Howto/cookbook
|
| 151 | --------------
|
| 152 |
|
| 153 | A) Starting a bcache with a missing caching device
|
| 154 |
|
| 155 | If registering the backing device doesn't help, it's already there, you just need
|
| 156 | to force it to run without the cache::
|
| 157 |
|
| 158 | host:~# echo /dev/sdb1 > /sys/fs/bcache/register
|
| 159 | [ 119.844831] bcache: register_bcache() error opening /dev/sdb1: device already registered
|
| 160 |
|
| 161 | Next, you try to register your caching device if it's present. However
|
| 162 | if it's absent, or registration fails for some reason, you can still
|
| 163 | start your bcache without its cache, like so::
|
| 164 |
|
| 165 | host:/sys/block/sdb/sdb1/bcache# echo 1 > running
|
| 166 |
|
| 167 | Note that this may cause data loss if you were running in writeback mode.
|
| 168 |
|
| 169 |
|
| 170 | B) Bcache does not find its cache::
|
| 171 |
|
| 172 | host:/sys/block/md5/bcache# echo 0226553a-37cf-41d5-b3ce-8b1e944543a8 > attach
|
| 173 | [ 1933.455082] bcache: bch_cached_dev_attach() Couldn't find uuid for md5 in set
|
| 174 | [ 1933.478179] bcache: __cached_dev_store() Can't attach 0226553a-37cf-41d5-b3ce-8b1e944543a8
|
| 175 | [ 1933.478179] : cache set not found
|
| 176 |
|
| 177 | In this case, the caching device was simply not registered at boot
|
| 178 | or disappeared and came back, and needs to be (re-)registered::
|
| 179 |
|
| 180 | host:/sys/block/md5/bcache# echo /dev/sdh2 > /sys/fs/bcache/register
|
| 181 |
|
| 182 |
|
| 183 | C) Corrupt bcache crashes the kernel at device registration time:
|
| 184 |
|
| 185 | This should never happen. If it does happen, then you have found a bug!
|
| 186 | Please report it to the bcache development list: [email protected]
|
| 187 |
|
| 188 | Be sure to provide as much information that you can including kernel dmesg
|
| 189 | output if available so that we may assist.
|
| 190 |
|
| 191 |
|
| 192 | D) Recovering data without bcache:
|
| 193 |
|
| 194 | If bcache is not available in the kernel, a filesystem on the backing
|
| 195 | device is still available at an 8KiB offset. So either via a loopdev
|
| 196 | of the backing device created with --offset 8K, or any value defined by
|
| 197 | --data-offset when you originally formatted bcache with `bcache make`.
|
| 198 |
|
| 199 | For example::
|
| 200 |
|
| 201 | losetup -o 8192 /dev/loop0 /dev/your_bcache_backing_dev
|
| 202 |
|
| 203 | This should present your unmodified backing device data in /dev/loop0
|
| 204 |
|
| 205 | If your cache is in writethrough mode, then you can safely discard the
|
| 206 | cache device without losing data.
|
| 207 |
|
| 208 |
|
| 209 | E) Wiping a cache device
|
| 210 |
|
| 211 | ::
|
| 212 |
|
| 213 | host:~# wipefs -a /dev/sdh2
|
| 214 | 16 bytes were erased at offset 0x1018 (bcache)
|
| 215 | they were: c6 85 73 f6 4e 1a 45 ca 82 65 f5 7f 48 ba 6d 81
|
| 216 |
|
| 217 | After you boot back with bcache enabled, you recreate the cache and attach it::
|
| 218 |
|
| 219 | host:~# bcache make -C /dev/sdh2
|
| 220 | UUID: 7be7e175-8f4c-4f99-94b2-9c904d227045
|
| 221 | Set UUID: 5bc072a8-ab17-446d-9744-e247949913c1
|
| 222 | version: 0
|
| 223 | nbuckets: 106874
|
| 224 | block_size: 1
|
| 225 | bucket_size: 1024
|
| 226 | nr_in_set: 1
|
| 227 | nr_this_dev: 0
|
| 228 | first_bucket: 1
|
| 229 | [ 650.511912] bcache: run_cache_set() invalidating existing data
|
| 230 | [ 650.549228] bcache: register_cache() registered cache device sdh2
|
| 231 |
|
| 232 | start backing device with missing cache::
|
| 233 |
|
| 234 | host:/sys/block/md5/bcache# echo 1 > running
|
| 235 |
|
| 236 | attach new cache::
|
| 237 |
|
| 238 | host:/sys/block/md5/bcache# echo 5bc072a8-ab17-446d-9744-e247949913c1 > attach
|
| 239 | [ 865.276616] bcache: bch_cached_dev_attach() Caching md5 as bcache0 on set 5bc072a8-ab17-446d-9744-e247949913c1
|
| 240 |
|
| 241 |
|
| 242 | F) Remove or replace a caching device::
|
| 243 |
|
| 244 | host:/sys/block/sda/sda7/bcache# echo 1 > detach
|
| 245 | [ 695.872542] bcache: cached_dev_detach_finish() Caching disabled for sda7
|
| 246 |
|
| 247 | host:~# wipefs -a /dev/nvme0n1p4
|
| 248 | wipefs: error: /dev/nvme0n1p4: probing initialization failed: Device or resource busy
|
| 249 | Ooops, it's disabled, but not unregistered, so it's still protected
|
| 250 |
|
| 251 | We need to go and unregister it::
|
| 252 |
|
| 253 | host:/sys/fs/bcache/b7ba27a1-2398-4649-8ae3-0959f57ba128# ls -l cache0
|
| 254 | lrwxrwxrwx 1 root root 0 Feb 25 18:33 cache0 -> ../../../devices/pci0000:00/0000:00:1d.0/0000:70:00.0/nvme/nvme0/nvme0n1/nvme0n1p4/bcache/
|
| 255 | host:/sys/fs/bcache/b7ba27a1-2398-4649-8ae3-0959f57ba128# echo 1 > stop
|
| 256 | kernel: [ 917.041908] bcache: cache_set_free() Cache set b7ba27a1-2398-4649-8ae3-0959f57ba128 unregistered
|
| 257 |
|
| 258 | Now we can wipe it::
|
| 259 |
|
| 260 | host:~# wipefs -a /dev/nvme0n1p4
|
| 261 | /dev/nvme0n1p4: 16 bytes were erased at offset 0x00001018 (bcache): c6 85 73 f6 4e 1a 45 ca 82 65 f5 7f 48 ba 6d 81
|
| 262 |
|
| 263 |
|
| 264 | G) dm-crypt and bcache
|
| 265 |
|
| 266 | First setup bcache unencrypted and then install dmcrypt on top of
|
| 267 | /dev/bcache<N> This will work faster than if you dmcrypt both the backing
|
| 268 | and caching devices and then install bcache on top. [benchmarks?]
|
| 269 |
|
| 270 |
|
| 271 | H) Stop/free a registered bcache to wipe and/or recreate it
|
| 272 |
|
| 273 | Suppose that you need to free up all bcache references so that you can
|
| 274 | fdisk run and re-register a changed partition table, which won't work
|
| 275 | if there are any active backing or caching devices left on it:
|
| 276 |
|
| 277 | 1) Is it present in /dev/bcache* ? (there are times where it won't be)
|
| 278 |
|
| 279 | If so, it's easy::
|
| 280 |
|
| 281 | host:/sys/block/bcache0/bcache# echo 1 > stop
|
| 282 |
|
| 283 | 2) But if your backing device is gone, this won't work::
|
| 284 |
|
| 285 | host:/sys/block/bcache0# cd bcache
|
| 286 | bash: cd: bcache: No such file or directory
|
| 287 |
|
| 288 | In this case, you may have to unregister the dmcrypt block device that
|
| 289 | references this bcache to free it up::
|
| 290 |
|
| 291 | host:~# dmsetup remove oldds1
|
| 292 | bcache: bcache_device_free() bcache0 stopped
|
| 293 | bcache: cache_set_free() Cache set 5bc072a8-ab17-446d-9744-e247949913c1 unregistered
|
| 294 |
|
| 295 | This causes the backing bcache to be removed from /sys/fs/bcache and
|
| 296 | then it can be reused. This would be true of any block device stacking
|
| 297 | where bcache is a lower device.
|
| 298 |
|
| 299 | 3) In other cases, you can also look in /sys/fs/bcache/::
|
| 300 |
|
| 301 | host:/sys/fs/bcache# ls -l */{cache?,bdev?}
|
| 302 | lrwxrwxrwx 1 root root 0 Mar 5 09:39 0226553a-37cf-41d5-b3ce-8b1e944543a8/bdev1 -> ../../../devices/virtual/block/dm-1/bcache/
|
| 303 | lrwxrwxrwx 1 root root 0 Mar 5 09:39 0226553a-37cf-41d5-b3ce-8b1e944543a8/cache0 -> ../../../devices/virtual/block/dm-4/bcache/
|
| 304 | lrwxrwxrwx 1 root root 0 Mar 5 09:39 5bc072a8-ab17-446d-9744-e247949913c1/cache0 -> ../../../devices/pci0000:00/0000:00:01.0/0000:01:00.0/ata10/host9/target9:0:0/9:0:0:0/block/sdl/sdl2/bcache/
|
| 305 |
|
| 306 | The device names will show which UUID is relevant, cd in that directory
|
| 307 | and stop the cache::
|
| 308 |
|
| 309 | host:/sys/fs/bcache/5bc072a8-ab17-446d-9744-e247949913c1# echo 1 > stop
|
| 310 |
|
| 311 | This will free up bcache references and let you reuse the partition for
|
| 312 | other purposes.
|
| 313 |
|
| 314 |
|
| 315 |
|
| 316 | Troubleshooting performance
|
| 317 | ---------------------------
|
| 318 |
|
| 319 | Bcache has a bunch of config options and tunables. The defaults are intended to
|
| 320 | be reasonable for typical desktop and server workloads, but they're not what you
|
| 321 | want for getting the best possible numbers when benchmarking.
|
| 322 |
|
| 323 | - Backing device alignment
|
| 324 |
|
| 325 | The default metadata size in bcache is 8k. If your backing device is
|
| 326 | RAID based, then be sure to align this by a multiple of your stride
|
| 327 | width using `bcache make --data-offset`. If you intend to expand your
|
| 328 | disk array in the future, then multiply a series of primes by your
|
| 329 | raid stripe size to get the disk multiples that you would like.
|
| 330 |
|
| 331 | For example: If you have a 64k stripe size, then the following offset
|
| 332 | would provide alignment for many common RAID5 data spindle counts::
|
| 333 |
|
| 334 | 64k * 2*2*2*3*3*5*7 bytes = 161280k
|
| 335 |
|
| 336 | That space is wasted, but for only 157.5MB you can grow your RAID 5
|
| 337 | volume to the following data-spindle counts without re-aligning::
|
| 338 |
|
| 339 | 3,4,5,6,7,8,9,10,12,14,15,18,20,21 ...
|
| 340 |
|
| 341 | - Bad write performance
|
| 342 |
|
| 343 | If write performance is not what you expected, you probably wanted to be
|
| 344 | running in writeback mode, which isn't the default (not due to a lack of
|
| 345 | maturity, but simply because in writeback mode you'll lose data if something
|
| 346 | happens to your SSD)::
|
| 347 |
|
| 348 | # echo writeback > /sys/block/bcache0/bcache/cache_mode
|
| 349 |
|
| 350 | - Bad performance, or traffic not going to the SSD that you'd expect
|
| 351 |
|
| 352 | By default, bcache doesn't cache everything. It tries to skip sequential IO -
|
| 353 | because you really want to be caching the random IO, and if you copy a 10
|
| 354 | gigabyte file you probably don't want that pushing 10 gigabytes of randomly
|
| 355 | accessed data out of your cache.
|
| 356 |
|
| 357 | But if you want to benchmark reads from cache, and you start out with fio
|
| 358 | writing an 8 gigabyte test file - so you want to disable that::
|
| 359 |
|
| 360 | # echo 0 > /sys/block/bcache0/bcache/sequential_cutoff
|
| 361 |
|
| 362 | To set it back to the default (4 mb), do::
|
| 363 |
|
| 364 | # echo 4M > /sys/block/bcache0/bcache/sequential_cutoff
|
| 365 |
|
| 366 | - Traffic's still going to the spindle/still getting cache misses
|
| 367 |
|
| 368 | In the real world, SSDs don't always keep up with disks - particularly with
|
| 369 | slower SSDs, many disks being cached by one SSD, or mostly sequential IO. So
|
| 370 | you want to avoid being bottlenecked by the SSD and having it slow everything
|
| 371 | down.
|
| 372 |
|
| 373 | To avoid that bcache tracks latency to the cache device, and gradually
|
| 374 | throttles traffic if the latency exceeds a threshold (it does this by
|
| 375 | cranking down the sequential bypass).
|
| 376 |
|
| 377 | You can disable this if you need to by setting the thresholds to 0::
|
| 378 |
|
| 379 | # echo 0 > /sys/fs/bcache/<cache set>/congested_read_threshold_us
|
| 380 | # echo 0 > /sys/fs/bcache/<cache set>/congested_write_threshold_us
|
| 381 |
|
| 382 | The default is 2000 us (2 milliseconds) for reads, and 20000 for writes.
|
| 383 |
|
| 384 | - Still getting cache misses, of the same data
|
| 385 |
|
| 386 | One last issue that sometimes trips people up is actually an old bug, due to
|
| 387 | the way cache coherency is handled for cache misses. If a btree node is full,
|
| 388 | a cache miss won't be able to insert a key for the new data and the data
|
| 389 | won't be written to the cache.
|
| 390 |
|
| 391 | In practice this isn't an issue because as soon as a write comes along it'll
|
| 392 | cause the btree node to be split, and you need almost no write traffic for
|
| 393 | this to not show up enough to be noticeable (especially since bcache's btree
|
| 394 | nodes are huge and index large regions of the device). But when you're
|
| 395 | benchmarking, if you're trying to warm the cache by reading a bunch of data
|
| 396 | and there's no other traffic - that can be a problem.
|
| 397 |
|
| 398 | Solution: warm the cache by doing writes, or use the testing branch (there's
|
| 399 | a fix for the issue there).
|
| 400 |
|
| 401 |
|
| 402 | Sysfs - backing device
|
| 403 | ----------------------
|
| 404 |
|
| 405 | Available at /sys/block/<bdev>/bcache, /sys/block/bcache*/bcache and
|
| 406 | (if attached) /sys/fs/bcache/<cset-uuid>/bdev*
|
| 407 |
|
| 408 | attach
|
| 409 | Echo the UUID of a cache set to this file to enable caching.
|
| 410 |
|
| 411 | cache_mode
|
| 412 | Can be one of either writethrough, writeback, writearound or none.
|
| 413 |
|
| 414 | clear_stats
|
| 415 | Writing to this file resets the running total stats (not the day/hour/5 minute
|
| 416 | decaying versions).
|
| 417 |
|
| 418 | detach
|
| 419 | Write to this file to detach from a cache set. If there is dirty data in the
|
| 420 | cache, it will be flushed first.
|
| 421 |
|
| 422 | dirty_data
|
| 423 | Amount of dirty data for this backing device in the cache. Continuously
|
| 424 | updated unlike the cache set's version, but may be slightly off.
|
| 425 |
|
| 426 | label
|
| 427 | Name of underlying device.
|
| 428 |
|
| 429 | readahead
|
| 430 | Size of readahead that should be performed. Defaults to 0. If set to e.g.
|
| 431 | 1M, it will round cache miss reads up to that size, but without overlapping
|
| 432 | existing cache entries.
|
| 433 |
|
| 434 | running
|
| 435 | 1 if bcache is running (i.e. whether the /dev/bcache device exists, whether
|
| 436 | it's in passthrough mode or caching).
|
| 437 |
|
| 438 | sequential_cutoff
|
| 439 | A sequential IO will bypass the cache once it passes this threshold; the
|
| 440 | most recent 128 IOs are tracked so sequential IO can be detected even when
|
| 441 | it isn't all done at once.
|
| 442 |
|
| 443 | sequential_merge
|
| 444 | If non zero, bcache keeps a list of the last 128 requests submitted to compare
|
| 445 | against all new requests to determine which new requests are sequential
|
| 446 | continuations of previous requests for the purpose of determining sequential
|
| 447 | cutoff. This is necessary if the sequential cutoff value is greater than the
|
| 448 | maximum acceptable sequential size for any single request.
|
| 449 |
|
| 450 | state
|
| 451 | The backing device can be in one of four different states:
|
| 452 |
|
| 453 | no cache: Has never been attached to a cache set.
|
| 454 |
|
| 455 | clean: Part of a cache set, and there is no cached dirty data.
|
| 456 |
|
| 457 | dirty: Part of a cache set, and there is cached dirty data.
|
| 458 |
|
| 459 | inconsistent: The backing device was forcibly run by the user when there was
|
| 460 | dirty data cached but the cache set was unavailable; whatever data was on the
|
| 461 | backing device has likely been corrupted.
|
| 462 |
|
| 463 | stop
|
| 464 | Write to this file to shut down the bcache device and close the backing
|
| 465 | device.
|
| 466 |
|
| 467 | writeback_delay
|
| 468 | When dirty data is written to the cache and it previously did not contain
|
| 469 | any, waits some number of seconds before initiating writeback. Defaults to
|
| 470 | 30.
|
| 471 |
|
| 472 | writeback_percent
|
| 473 | If nonzero, bcache tries to keep around this percentage of the cache dirty by
|
| 474 | throttling background writeback and using a PD controller to smoothly adjust
|
| 475 | the rate.
|
| 476 |
|
| 477 | writeback_rate
|
| 478 | Rate in sectors per second - if writeback_percent is nonzero, background
|
| 479 | writeback is throttled to this rate. Continuously adjusted by bcache but may
|
| 480 | also be set by the user.
|
| 481 |
|
| 482 | writeback_running
|
| 483 | If off, writeback of dirty data will not take place at all. Dirty data will
|
| 484 | still be added to the cache until it is mostly full; only meant for
|
| 485 | benchmarking. Defaults to on.
|
| 486 |
|
| 487 | Sysfs - backing device stats
|
| 488 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
| 489 |
|
| 490 | There are directories with these numbers for a running total, as well as
|
| 491 | versions that decay over the past day, hour and 5 minutes; they're also
|
| 492 | aggregated in the cache set directory as well.
|
| 493 |
|
| 494 | bypassed
|
| 495 | Amount of IO (both reads and writes) that has bypassed the cache
|
| 496 |
|
| 497 | cache_hits, cache_misses, cache_hit_ratio
|
| 498 | Hits and misses are counted per individual IO as bcache sees them; a
|
| 499 | partial hit is counted as a miss.
|
| 500 |
|
| 501 | cache_bypass_hits, cache_bypass_misses
|
| 502 | Hits and misses for IO that is intended to skip the cache are still counted,
|
| 503 | but broken out here.
|
| 504 |
|
| 505 | cache_miss_collisions
|
| 506 | Counts instances where data was going to be inserted into the cache from a
|
| 507 | cache miss, but raced with a write and data was already present (usually 0
|
| 508 | since the synchronization for cache misses was rewritten)
|
| 509 |
|
| 510 | Sysfs - cache set
|
| 511 | ~~~~~~~~~~~~~~~~~
|
| 512 |
|
| 513 | Available at /sys/fs/bcache/<cset-uuid>
|
| 514 |
|
| 515 | average_key_size
|
| 516 | Average data per key in the btree.
|
| 517 |
|
| 518 | bdev<0..n>
|
| 519 | Symlink to each of the attached backing devices.
|
| 520 |
|
| 521 | block_size
|
| 522 | Block size of the cache devices.
|
| 523 |
|
| 524 | btree_cache_size
|
| 525 | Amount of memory currently used by the btree cache
|
| 526 |
|
| 527 | bucket_size
|
| 528 | Size of buckets
|
| 529 |
|
| 530 | cache<0..n>
|
| 531 | Symlink to each of the cache devices comprising this cache set.
|
| 532 |
|
| 533 | cache_available_percent
|
| 534 | Percentage of cache device which doesn't contain dirty data, and could
|
| 535 | potentially be used for writeback. This doesn't mean this space isn't used
|
| 536 | for clean cached data; the unused statistic (in priority_stats) is typically
|
| 537 | much lower.
|
| 538 |
|
| 539 | clear_stats
|
| 540 | Clears the statistics associated with this cache
|
| 541 |
|
| 542 | dirty_data
|
| 543 | Amount of dirty data is in the cache (updated when garbage collection runs).
|
| 544 |
|
| 545 | flash_vol_create
|
| 546 | Echoing a size to this file (in human readable units, k/M/G) creates a thinly
|
| 547 | provisioned volume backed by the cache set.
|
| 548 |
|
| 549 | io_error_halflife, io_error_limit
|
| 550 | These determines how many errors we accept before disabling the cache.
|
| 551 | Each error is decayed by the half life (in # ios). If the decaying count
|
| 552 | reaches io_error_limit dirty data is written out and the cache is disabled.
|
| 553 |
|
| 554 | journal_delay_ms
|
| 555 | Journal writes will delay for up to this many milliseconds, unless a cache
|
| 556 | flush happens sooner. Defaults to 100.
|
| 557 |
|
| 558 | root_usage_percent
|
| 559 | Percentage of the root btree node in use. If this gets too high the node
|
| 560 | will split, increasing the tree depth.
|
| 561 |
|
| 562 | stop
|
| 563 | Write to this file to shut down the cache set - waits until all attached
|
| 564 | backing devices have been shut down.
|
| 565 |
|
| 566 | tree_depth
|
| 567 | Depth of the btree (A single node btree has depth 0).
|
| 568 |
|
| 569 | unregister
|
| 570 | Detaches all backing devices and closes the cache devices; if dirty data is
|
| 571 | present it will disable writeback caching and wait for it to be flushed.
|
| 572 |
|
| 573 | Sysfs - cache set internal
|
| 574 | ~~~~~~~~~~~~~~~~~~~~~~~~~~
|
| 575 |
|
| 576 | This directory also exposes timings for a number of internal operations, with
|
| 577 | separate files for average duration, average frequency, last occurrence and max
|
| 578 | duration: garbage collection, btree read, btree node sorts and btree splits.
|
| 579 |
|
| 580 | active_journal_entries
|
| 581 | Number of journal entries that are newer than the index.
|
| 582 |
|
| 583 | btree_nodes
|
| 584 | Total nodes in the btree.
|
| 585 |
|
| 586 | btree_used_percent
|
| 587 | Average fraction of btree in use.
|
| 588 |
|
| 589 | bset_tree_stats
|
| 590 | Statistics about the auxiliary search trees
|
| 591 |
|
| 592 | btree_cache_max_chain
|
| 593 | Longest chain in the btree node cache's hash table
|
| 594 |
|
| 595 | cache_read_races
|
| 596 | Counts instances where while data was being read from the cache, the bucket
|
| 597 | was reused and invalidated - i.e. where the pointer was stale after the read
|
| 598 | completed. When this occurs the data is reread from the backing device.
|
| 599 |
|
| 600 | trigger_gc
|
| 601 | Writing to this file forces garbage collection to run.
|
| 602 |
|
| 603 | Sysfs - Cache device
|
| 604 | ~~~~~~~~~~~~~~~~~~~~
|
| 605 |
|
| 606 | Available at /sys/block/<cdev>/bcache
|
| 607 |
|
| 608 | block_size
|
| 609 | Minimum granularity of writes - should match hardware sector size.
|
| 610 |
|
| 611 | btree_written
|
| 612 | Sum of all btree writes, in (kilo/mega/giga) bytes
|
| 613 |
|
| 614 | bucket_size
|
| 615 | Size of buckets
|
| 616 |
|
| 617 | cache_replacement_policy
|
| 618 | One of either lru, fifo or random.
|
| 619 |
|
| 620 | freelist_percent
|
| 621 | Size of the freelist as a percentage of nbuckets. Can be written to
|
| 622 | increase the number of buckets kept on the freelist, which lets you
|
| 623 | artificially reduce the size of the cache at runtime. Mostly for testing
|
| 624 | purposes (i.e. testing how different size caches affect your hit rate).
|
| 625 |
|
| 626 | io_errors
|
| 627 | Number of errors that have occurred, decayed by io_error_halflife.
|
| 628 |
|
| 629 | metadata_written
|
| 630 | Sum of all non data writes (btree writes and all other metadata).
|
| 631 |
|
| 632 | nbuckets
|
| 633 | Total buckets in this cache
|
| 634 |
|
| 635 | priority_stats
|
| 636 | Statistics about how recently data in the cache has been accessed.
|
| 637 | This can reveal your working set size. Unused is the percentage of
|
| 638 | the cache that doesn't contain any data. Metadata is bcache's
|
| 639 | metadata overhead. Average is the average priority of cache buckets.
|
| 640 | Next is a list of quantiles with the priority threshold of each.
|
| 641 |
|
| 642 | written
|
| 643 | Sum of all data that has been written to the cache; comparison with
|
| 644 | btree_written gives the amount of write inflation in bcache.
|