Root/
Source at commit fbf123cd4cc0c097fe9a99c90109ebb2a5e94a50 created 10 years 3 months ago. By Lars-Peter Clausen, dma: jz4740: Dequeue descriptor from active list before completing it | |
---|---|
1 | Say you've got a big slow raid 6, and an X-25E or three. Wouldn't it be |
2 | nice if you could use them as cache... Hence bcache. |
3 | |
4 | Wiki and git repositories are at: |
5 | http://bcache.evilpiepirate.org |
6 | http://evilpiepirate.org/git/linux-bcache.git |
7 | http://evilpiepirate.org/git/bcache-tools.git |
8 | |
9 | It's designed around the performance characteristics of SSDs - it only allocates |
10 | in erase block sized buckets, and it uses a hybrid btree/log to track cached |
11 | extants (which can be anywhere from a single sector to the bucket size). It's |
12 | designed to avoid random writes at all costs; it fills up an erase block |
13 | sequentially, then issues a discard before reusing it. |
14 | |
15 | Both writethrough and writeback caching are supported. Writeback defaults to |
16 | off, but can be switched on and off arbitrarily at runtime. Bcache goes to |
17 | great lengths to protect your data - it reliably handles unclean shutdown. (It |
18 | doesn't even have a notion of a clean shutdown; bcache simply doesn't return |
19 | writes as completed until they're on stable storage). |
20 | |
21 | Writeback caching can use most of the cache for buffering writes - writing |
22 | dirty data to the backing device is always done sequentially, scanning from the |
23 | start to the end of the index. |
24 | |
25 | Since random IO is what SSDs excel at, there generally won't be much benefit |
26 | to caching large sequential IO. Bcache detects sequential IO and skips it; |
27 | it also keeps a rolling average of the IO sizes per task, and as long as the |
28 | average is above the cutoff it will skip all IO from that task - instead of |
29 | caching the first 512k after every seek. Backups and large file copies should |
30 | thus entirely bypass the cache. |
31 | |
32 | In the event of a data IO error on the flash it will try to recover by reading |
33 | from disk or invalidating cache entries. For unrecoverable errors (meta data |
34 | or dirty data), caching is automatically disabled; if dirty data was present |
35 | in the cache it first disables writeback caching and waits for all dirty data |
36 | to be flushed. |
37 | |
38 | Getting started: |
39 | You'll need make-bcache from the bcache-tools repository. Both the cache device |
40 | and backing device must be formatted before use. |
41 | make-bcache -B /dev/sdb |
42 | make-bcache -C /dev/sdc |
43 | |
44 | make-bcache has the ability to format multiple devices at the same time - if |
45 | you format your backing devices and cache device at the same time, you won't |
46 | have to manually attach: |
47 | make-bcache -B /dev/sda /dev/sdb -C /dev/sdc |
48 | |
49 | bcache-tools now ships udev rules, and bcache devices are known to the kernel |
50 | immediately. Without udev, you can manually register devices like this: |
51 | |
52 | echo /dev/sdb > /sys/fs/bcache/register |
53 | echo /dev/sdc > /sys/fs/bcache/register |
54 | |
55 | Registering the backing device makes the bcache device show up in /dev; you can |
56 | now format it and use it as normal. But the first time using a new bcache |
57 | device, it'll be running in passthrough mode until you attach it to a cache. |
58 | See the section on attaching. |
59 | |
60 | The devices show up as: |
61 | |
62 | /dev/bcache<N> |
63 | |
64 | As well as (with udev): |
65 | |
66 | /dev/bcache/by-uuid/<uuid> |
67 | /dev/bcache/by-label/<label> |
68 | |
69 | To get started: |
70 | |
71 | mkfs.ext4 /dev/bcache0 |
72 | mount /dev/bcache0 /mnt |
73 | |
74 | You can control bcache devices through sysfs at /sys/block/bcache<N>/bcache . |
75 | |
76 | Cache devices are managed as sets; multiple caches per set isn't supported yet |
77 | but will allow for mirroring of metadata and dirty data in the future. Your new |
78 | cache set shows up as /sys/fs/bcache/<UUID> |
79 | |
80 | ATTACHING: |
81 | |
82 | After your cache device and backing device are registered, the backing device |
83 | must be attached to your cache set to enable caching. Attaching a backing |
84 | device to a cache set is done thusly, with the UUID of the cache set in |
85 | /sys/fs/bcache: |
86 | |
87 | echo <CSET-UUID> > /sys/block/bcache0/bcache/attach |
88 | |
89 | This only has to be done once. The next time you reboot, just reregister all |
90 | your bcache devices. If a backing device has data in a cache somewhere, the |
91 | /dev/bcache<N> device won't be created until the cache shows up - particularly |
92 | important if you have writeback caching turned on. |
93 | |
94 | If you're booting up and your cache device is gone and never coming back, you |
95 | can force run the backing device: |
96 | |
97 | echo 1 > /sys/block/sdb/bcache/running |
98 | |
99 | (You need to use /sys/block/sdb (or whatever your backing device is called), not |
100 | /sys/block/bcache0, because bcache0 doesn't exist yet. If you're using a |
101 | partition, the bcache directory would be at /sys/block/sdb/sdb2/bcache) |
102 | |
103 | The backing device will still use that cache set if it shows up in the future, |
104 | but all the cached data will be invalidated. If there was dirty data in the |
105 | cache, don't expect the filesystem to be recoverable - you will have massive |
106 | filesystem corruption, though ext4's fsck does work miracles. |
107 | |
108 | ERROR HANDLING: |
109 | |
110 | Bcache tries to transparently handle IO errors to/from the cache device without |
111 | affecting normal operation; if it sees too many errors (the threshold is |
112 | configurable, and defaults to 0) it shuts down the cache device and switches all |
113 | the backing devices to passthrough mode. |
114 | |
115 | - For reads from the cache, if they error we just retry the read from the |
116 | backing device. |
117 | |
118 | - For writethrough writes, if the write to the cache errors we just switch to |
119 | invalidating the data at that lba in the cache (i.e. the same thing we do for |
120 | a write that bypasses the cache) |
121 | |
122 | - For writeback writes, we currently pass that error back up to the |
123 | filesystem/userspace. This could be improved - we could retry it as a write |
124 | that skips the cache so we don't have to error the write. |
125 | |
126 | - When we detach, we first try to flush any dirty data (if we were running in |
127 | writeback mode). It currently doesn't do anything intelligent if it fails to |
128 | read some of the dirty data, though. |
129 | |
130 | TROUBLESHOOTING PERFORMANCE: |
131 | |
132 | Bcache has a bunch of config options and tunables. The defaults are intended to |
133 | be reasonable for typical desktop and server workloads, but they're not what you |
134 | want for getting the best possible numbers when benchmarking. |
135 | |
136 | - Bad write performance |
137 | |
138 | If write performance is not what you expected, you probably wanted to be |
139 | running in writeback mode, which isn't the default (not due to a lack of |
140 | maturity, but simply because in writeback mode you'll lose data if something |
141 | happens to your SSD) |
142 | |
143 | # echo writeback > /sys/block/bcache0/cache_mode |
144 | |
145 | - Bad performance, or traffic not going to the SSD that you'd expect |
146 | |
147 | By default, bcache doesn't cache everything. It tries to skip sequential IO - |
148 | because you really want to be caching the random IO, and if you copy a 10 |
149 | gigabyte file you probably don't want that pushing 10 gigabytes of randomly |
150 | accessed data out of your cache. |
151 | |
152 | But if you want to benchmark reads from cache, and you start out with fio |
153 | writing an 8 gigabyte test file - so you want to disable that. |
154 | |
155 | # echo 0 > /sys/block/bcache0/bcache/sequential_cutoff |
156 | |
157 | To set it back to the default (4 mb), do |
158 | |
159 | # echo 4M > /sys/block/bcache0/bcache/sequential_cutoff |
160 | |
161 | - Traffic's still going to the spindle/still getting cache misses |
162 | |
163 | In the real world, SSDs don't always keep up with disks - particularly with |
164 | slower SSDs, many disks being cached by one SSD, or mostly sequential IO. So |
165 | you want to avoid being bottlenecked by the SSD and having it slow everything |
166 | down. |
167 | |
168 | To avoid that bcache tracks latency to the cache device, and gradually |
169 | throttles traffic if the latency exceeds a threshold (it does this by |
170 | cranking down the sequential bypass). |
171 | |
172 | You can disable this if you need to by setting the thresholds to 0: |
173 | |
174 | # echo 0 > /sys/fs/bcache/<cache set>/congested_read_threshold_us |
175 | # echo 0 > /sys/fs/bcache/<cache set>/congested_write_threshold_us |
176 | |
177 | The default is 2000 us (2 milliseconds) for reads, and 20000 for writes. |
178 | |
179 | - Still getting cache misses, of the same data |
180 | |
181 | One last issue that sometimes trips people up is actually an old bug, due to |
182 | the way cache coherency is handled for cache misses. If a btree node is full, |
183 | a cache miss won't be able to insert a key for the new data and the data |
184 | won't be written to the cache. |
185 | |
186 | In practice this isn't an issue because as soon as a write comes along it'll |
187 | cause the btree node to be split, and you need almost no write traffic for |
188 | this to not show up enough to be noticeable (especially since bcache's btree |
189 | nodes are huge and index large regions of the device). But when you're |
190 | benchmarking, if you're trying to warm the cache by reading a bunch of data |
191 | and there's no other traffic - that can be a problem. |
192 | |
193 | Solution: warm the cache by doing writes, or use the testing branch (there's |
194 | a fix for the issue there). |
195 | |
196 | SYSFS - BACKING DEVICE: |
197 | |
198 | Available at /sys/block/<bdev>/bcache, /sys/block/bcache*/bcache and |
199 | (if attached) /sys/fs/bcache/<cset-uuid>/bdev* |
200 | |
201 | attach |
202 | Echo the UUID of a cache set to this file to enable caching. |
203 | |
204 | cache_mode |
205 | Can be one of either writethrough, writeback, writearound or none. |
206 | |
207 | clear_stats |
208 | Writing to this file resets the running total stats (not the day/hour/5 minute |
209 | decaying versions). |
210 | |
211 | detach |
212 | Write to this file to detach from a cache set. If there is dirty data in the |
213 | cache, it will be flushed first. |
214 | |
215 | dirty_data |
216 | Amount of dirty data for this backing device in the cache. Continuously |
217 | updated unlike the cache set's version, but may be slightly off. |
218 | |
219 | label |
220 | Name of underlying device. |
221 | |
222 | readahead |
223 | Size of readahead that should be performed. Defaults to 0. If set to e.g. |
224 | 1M, it will round cache miss reads up to that size, but without overlapping |
225 | existing cache entries. |
226 | |
227 | running |
228 | 1 if bcache is running (i.e. whether the /dev/bcache device exists, whether |
229 | it's in passthrough mode or caching). |
230 | |
231 | sequential_cutoff |
232 | A sequential IO will bypass the cache once it passes this threshold; the |
233 | most recent 128 IOs are tracked so sequential IO can be detected even when |
234 | it isn't all done at once. |
235 | |
236 | sequential_merge |
237 | If non zero, bcache keeps a list of the last 128 requests submitted to compare |
238 | against all new requests to determine which new requests are sequential |
239 | continuations of previous requests for the purpose of determining sequential |
240 | cutoff. This is necessary if the sequential cutoff value is greater than the |
241 | maximum acceptable sequential size for any single request. |
242 | |
243 | state |
244 | The backing device can be in one of four different states: |
245 | |
246 | no cache: Has never been attached to a cache set. |
247 | |
248 | clean: Part of a cache set, and there is no cached dirty data. |
249 | |
250 | dirty: Part of a cache set, and there is cached dirty data. |
251 | |
252 | inconsistent: The backing device was forcibly run by the user when there was |
253 | dirty data cached but the cache set was unavailable; whatever data was on the |
254 | backing device has likely been corrupted. |
255 | |
256 | stop |
257 | Write to this file to shut down the bcache device and close the backing |
258 | device. |
259 | |
260 | writeback_delay |
261 | When dirty data is written to the cache and it previously did not contain |
262 | any, waits some number of seconds before initiating writeback. Defaults to |
263 | 30. |
264 | |
265 | writeback_percent |
266 | If nonzero, bcache tries to keep around this percentage of the cache dirty by |
267 | throttling background writeback and using a PD controller to smoothly adjust |
268 | the rate. |
269 | |
270 | writeback_rate |
271 | Rate in sectors per second - if writeback_percent is nonzero, background |
272 | writeback is throttled to this rate. Continuously adjusted by bcache but may |
273 | also be set by the user. |
274 | |
275 | writeback_running |
276 | If off, writeback of dirty data will not take place at all. Dirty data will |
277 | still be added to the cache until it is mostly full; only meant for |
278 | benchmarking. Defaults to on. |
279 | |
280 | SYSFS - BACKING DEVICE STATS: |
281 | |
282 | There are directories with these numbers for a running total, as well as |
283 | versions that decay over the past day, hour and 5 minutes; they're also |
284 | aggregated in the cache set directory as well. |
285 | |
286 | bypassed |
287 | Amount of IO (both reads and writes) that has bypassed the cache |
288 | |
289 | cache_hits |
290 | cache_misses |
291 | cache_hit_ratio |
292 | Hits and misses are counted per individual IO as bcache sees them; a |
293 | partial hit is counted as a miss. |
294 | |
295 | cache_bypass_hits |
296 | cache_bypass_misses |
297 | Hits and misses for IO that is intended to skip the cache are still counted, |
298 | but broken out here. |
299 | |
300 | cache_miss_collisions |
301 | Counts instances where data was going to be inserted into the cache from a |
302 | cache miss, but raced with a write and data was already present (usually 0 |
303 | since the synchronization for cache misses was rewritten) |
304 | |
305 | cache_readaheads |
306 | Count of times readahead occurred. |
307 | |
308 | SYSFS - CACHE SET: |
309 | |
310 | Available at /sys/fs/bcache/<cset-uuid> |
311 | |
312 | average_key_size |
313 | Average data per key in the btree. |
314 | |
315 | bdev<0..n> |
316 | Symlink to each of the attached backing devices. |
317 | |
318 | block_size |
319 | Block size of the cache devices. |
320 | |
321 | btree_cache_size |
322 | Amount of memory currently used by the btree cache |
323 | |
324 | bucket_size |
325 | Size of buckets |
326 | |
327 | cache<0..n> |
328 | Symlink to each of the cache devices comprising this cache set. |
329 | |
330 | cache_available_percent |
331 | Percentage of cache device which doesn't contain dirty data, and could |
332 | potentially be used for writeback. This doesn't mean this space isn't used |
333 | for clean cached data; the unused statistic (in priority_stats) is typically |
334 | much lower. |
335 | |
336 | clear_stats |
337 | Clears the statistics associated with this cache |
338 | |
339 | dirty_data |
340 | Amount of dirty data is in the cache (updated when garbage collection runs). |
341 | |
342 | flash_vol_create |
343 | Echoing a size to this file (in human readable units, k/M/G) creates a thinly |
344 | provisioned volume backed by the cache set. |
345 | |
346 | io_error_halflife |
347 | io_error_limit |
348 | These determines how many errors we accept before disabling the cache. |
349 | Each error is decayed by the half life (in # ios). If the decaying count |
350 | reaches io_error_limit dirty data is written out and the cache is disabled. |
351 | |
352 | journal_delay_ms |
353 | Journal writes will delay for up to this many milliseconds, unless a cache |
354 | flush happens sooner. Defaults to 100. |
355 | |
356 | root_usage_percent |
357 | Percentage of the root btree node in use. If this gets too high the node |
358 | will split, increasing the tree depth. |
359 | |
360 | stop |
361 | Write to this file to shut down the cache set - waits until all attached |
362 | backing devices have been shut down. |
363 | |
364 | tree_depth |
365 | Depth of the btree (A single node btree has depth 0). |
366 | |
367 | unregister |
368 | Detaches all backing devices and closes the cache devices; if dirty data is |
369 | present it will disable writeback caching and wait for it to be flushed. |
370 | |
371 | SYSFS - CACHE SET INTERNAL: |
372 | |
373 | This directory also exposes timings for a number of internal operations, with |
374 | separate files for average duration, average frequency, last occurrence and max |
375 | duration: garbage collection, btree read, btree node sorts and btree splits. |
376 | |
377 | active_journal_entries |
378 | Number of journal entries that are newer than the index. |
379 | |
380 | btree_nodes |
381 | Total nodes in the btree. |
382 | |
383 | btree_used_percent |
384 | Average fraction of btree in use. |
385 | |
386 | bset_tree_stats |
387 | Statistics about the auxiliary search trees |
388 | |
389 | btree_cache_max_chain |
390 | Longest chain in the btree node cache's hash table |
391 | |
392 | cache_read_races |
393 | Counts instances where while data was being read from the cache, the bucket |
394 | was reused and invalidated - i.e. where the pointer was stale after the read |
395 | completed. When this occurs the data is reread from the backing device. |
396 | |
397 | trigger_gc |
398 | Writing to this file forces garbage collection to run. |
399 | |
400 | SYSFS - CACHE DEVICE: |
401 | |
402 | Available at /sys/block/<cdev>/bcache |
403 | |
404 | block_size |
405 | Minimum granularity of writes - should match hardware sector size. |
406 | |
407 | btree_written |
408 | Sum of all btree writes, in (kilo/mega/giga) bytes |
409 | |
410 | bucket_size |
411 | Size of buckets |
412 | |
413 | cache_replacement_policy |
414 | One of either lru, fifo or random. |
415 | |
416 | discard |
417 | Boolean; if on a discard/TRIM will be issued to each bucket before it is |
418 | reused. Defaults to off, since SATA TRIM is an unqueued command (and thus |
419 | slow). |
420 | |
421 | freelist_percent |
422 | Size of the freelist as a percentage of nbuckets. Can be written to to |
423 | increase the number of buckets kept on the freelist, which lets you |
424 | artificially reduce the size of the cache at runtime. Mostly for testing |
425 | purposes (i.e. testing how different size caches affect your hit rate), but |
426 | since buckets are discarded when they move on to the freelist will also make |
427 | the SSD's garbage collection easier by effectively giving it more reserved |
428 | space. |
429 | |
430 | io_errors |
431 | Number of errors that have occurred, decayed by io_error_halflife. |
432 | |
433 | metadata_written |
434 | Sum of all non data writes (btree writes and all other metadata). |
435 | |
436 | nbuckets |
437 | Total buckets in this cache |
438 | |
439 | priority_stats |
440 | Statistics about how recently data in the cache has been accessed. |
441 | This can reveal your working set size. Unused is the percentage of |
442 | the cache that doesn't contain any data. Metadata is bcache's |
443 | metadata overhead. Average is the average priority of cache buckets. |
444 | Next is a list of quantiles with the priority threshold of each. |
445 | |
446 | written |
447 | Sum of all data that has been written to the cache; comparison with |
448 | btree_written gives the amount of write inflation in bcache. |
449 |
Branches:
ben-wpan
ben-wpan-stefan
javiroman/ks7010
jz-2.6.34
jz-2.6.34-rc5
jz-2.6.34-rc6
jz-2.6.34-rc7
jz-2.6.35
jz-2.6.36
jz-2.6.37
jz-2.6.38
jz-2.6.39
jz-3.0
jz-3.1
jz-3.11
jz-3.12
jz-3.13
jz-3.15
jz-3.16
jz-3.18-dt
jz-3.2
jz-3.3
jz-3.4
jz-3.5
jz-3.6
jz-3.6-rc2-pwm
jz-3.9
jz-3.9-clk
jz-3.9-rc8
jz47xx
jz47xx-2.6.38
master
Tags:
od-2011-09-04
od-2011-09-18
v2.6.34-rc5
v2.6.34-rc6
v2.6.34-rc7
v3.9