You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

2156 lines
82 KiB

7 years ago
7 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
7 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
7 years ago
6 years ago
6 years ago
7 years ago
7 years ago
7 years ago
7 years ago
7 years ago
7 years ago
7 years ago
7 years ago
7 years ago
7 years ago
7 years ago
7 years ago
7 years ago
7 years ago
9 years ago
9 years ago
6 years ago
6 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
8 years ago
9 years ago
7 years ago
7 years ago
9 years ago
9 years ago
9 years ago
  1. .\"t
  2. .\" Automatically generated by Pandoc 1.19.2.4
  3. .\"
  4. .TH "mergerfs" "1" "2019\-06\-03" "mergerfs user manual" ""
  5. .hy
  6. .SH NAME
  7. .PP
  8. mergerfs \- a featureful union filesystem
  9. .SH SYNOPSIS
  10. .PP
  11. mergerfs \-o<options> <branches> <mountpoint>
  12. .SH DESCRIPTION
  13. .PP
  14. \f[B]mergerfs\f[] is a union filesystem geared towards simplifying
  15. storage and management of files across numerous commodity storage
  16. devices.
  17. It is similar to \f[B]mhddfs\f[], \f[B]unionfs\f[], and \f[B]aufs\f[].
  18. .SH FEATURES
  19. .IP \[bu] 2
  20. Runs in userspace (FUSE)
  21. .IP \[bu] 2
  22. Configurable behaviors / file placement
  23. .IP \[bu] 2
  24. Support for extended attributes (xattrs)
  25. .IP \[bu] 2
  26. Support for file attributes (chattr)
  27. .IP \[bu] 2
  28. Runtime configurable (via xattrs)
  29. .IP \[bu] 2
  30. Safe to run as root
  31. .IP \[bu] 2
  32. Opportunistic credential caching
  33. .IP \[bu] 2
  34. Works with heterogeneous filesystem types
  35. .IP \[bu] 2
  36. Handling of writes to full drives (transparently move file to drive with
  37. capacity)
  38. .IP \[bu] 2
  39. Handles pool of read\-only and read/write drives
  40. .IP \[bu] 2
  41. Can turn read\-only files into symlinks to underlying file
  42. .IP \[bu] 2
  43. Hard link copy\-on\-write / CoW
  44. .IP \[bu] 2
  45. supports POSIX ACLs
  46. .SH How it works
  47. .PP
  48. mergerfs logically merges multiple paths together.
  49. Think a union of sets.
  50. The file/s or directory/s acted on or presented through mergerfs are
  51. based on the policy chosen for that particular action.
  52. Read more about policies below.
  53. .IP
  54. .nf
  55. \f[C]
  56. A\ \ \ \ \ \ \ \ \ +\ \ \ \ \ \ B\ \ \ \ \ \ \ \ =\ \ \ \ \ \ \ C
  57. /disk1\ \ \ \ \ \ \ \ \ \ \ /disk2\ \ \ \ \ \ \ \ \ \ \ /merged
  58. |\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ |\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ |
  59. +\-\-\ /dir1\ \ \ \ \ \ \ \ +\-\-\ /dir1\ \ \ \ \ \ \ \ +\-\-\ /dir1
  60. |\ \ \ |\ \ \ \ \ \ \ \ \ \ \ \ |\ \ \ |\ \ \ \ \ \ \ \ \ \ \ \ |\ \ \ |
  61. |\ \ \ +\-\-\ file1\ \ \ \ |\ \ \ +\-\-\ file2\ \ \ \ |\ \ \ +\-\-\ file1
  62. |\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ |\ \ \ +\-\-\ file3\ \ \ \ |\ \ \ +\-\-\ file2
  63. +\-\-\ /dir2\ \ \ \ \ \ \ \ |\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ |\ \ \ +\-\-\ file3
  64. |\ \ \ |\ \ \ \ \ \ \ \ \ \ \ \ +\-\-\ /dir3\ \ \ \ \ \ \ \ |
  65. |\ \ \ +\-\-\ file4\ \ \ \ \ \ \ \ |\ \ \ \ \ \ \ \ \ \ \ \ +\-\-\ /dir2
  66. |\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ +\-\-\ file5\ \ \ |\ \ \ |
  67. +\-\-\ file6\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ |\ \ \ +\-\-\ file4
  68. \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ |
  69. \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ +\-\-\ /dir3
  70. \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ |\ \ \ |
  71. \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ |\ \ \ +\-\-\ file5
  72. \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ |
  73. \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ +\-\-\ file6
  74. \f[]
  75. .fi
  76. .PP
  77. mergerfs does \f[B]not\f[] support the copy\-on\-write (CoW) behavior
  78. found in \f[B]aufs\f[] and \f[B]overlayfs\f[].
  79. You can \f[B]not\f[] mount a read\-only filesystem and write to it.
  80. However, mergerfs will ignore read\-only drives when creating new files
  81. so you can mix read\-write and read\-only drives.
  82. It also does \f[B]not\f[] split data across drives.
  83. It is not RAID0 / striping.
  84. It is simply a union.
  85. .SH OPTIONS
  86. .SS mount options
  87. .IP \[bu] 2
  88. \f[B]allow_other\f[]: A libfuse option which allows users besides the
  89. one which ran mergerfs to see the filesystem.
  90. This is required for most use\-cases.
  91. .IP \[bu] 2
  92. \f[B]minfreespace=SIZE\f[]: The minimum space value used for creation
  93. policies.
  94. Understands \[aq]K\[aq], \[aq]M\[aq], and \[aq]G\[aq] to represent
  95. kilobyte, megabyte, and gigabyte respectively.
  96. (default: 4G)
  97. .IP \[bu] 2
  98. \f[B]moveonenospc=BOOL\f[]: When enabled if a \f[B]write\f[] fails with
  99. \f[B]ENOSPC\f[] or \f[B]EDQUOT\f[] a scan of all drives will be done
  100. looking for the drive with the most free space which is at least the
  101. size of the file plus the amount which failed to write.
  102. An attempt to move the file to that drive will occur (keeping all
  103. metadata possible) and if successful the original is unlinked and the
  104. write retried.
  105. (default: false)
  106. .IP \[bu] 2
  107. \f[B]use_ino\f[]: Causes mergerfs to supply file/directory inodes rather
  108. than libfuse.
  109. While not a default it is recommended it be enabled so that linked files
  110. share the same inode value.
  111. .IP \[bu] 2
  112. \f[B]dropcacheonclose=BOOL\f[]: When a file is requested to be closed
  113. call \f[C]posix_fadvise\f[] on it first to instruct the kernel that we
  114. no longer need the data and it can drop its cache.
  115. Recommended when \f[B]cache.files=partial|full|auto\-full\f[] to limit
  116. double caching.
  117. (default: false)
  118. .IP \[bu] 2
  119. \f[B]symlinkify=BOOL\f[]: When enabled and a file is not writable and
  120. its mtime or ctime is older than \f[B]symlinkify_timeout\f[] files will
  121. be reported as symlinks to the original files.
  122. Please read more below before using.
  123. (default: false)
  124. .IP \[bu] 2
  125. \f[B]symlinkify_timeout=INT\f[]: Time to wait, in seconds, to activate
  126. the \f[B]symlinkify\f[] behavior.
  127. (default: 3600)
  128. .IP \[bu] 2
  129. \f[B]nullrw=BOOL\f[]: Turns reads and writes into no\-ops.
  130. The request will succeed but do nothing.
  131. Useful for benchmarking mergerfs.
  132. (default: false)
  133. .IP \[bu] 2
  134. \f[B]ignorepponrename=BOOL\f[]: Ignore path preserving on rename.
  135. Typically rename and link act differently depending on the policy of
  136. \f[C]create\f[] (read below).
  137. Enabling this will cause rename and link to always use the non\-path
  138. preserving behavior.
  139. This means files, when renamed or linked, will stay on the same drive.
  140. (default: false)
  141. .IP \[bu] 2
  142. \f[B]security_capability=BOOL\f[]: If false return ENOATTR when xattr
  143. security.capability is queried.
  144. (default: true)
  145. .IP \[bu] 2
  146. \f[B]xattr=passthrough|noattr|nosys\f[]: Runtime control of xattrs.
  147. Default is to passthrough xattr requests.
  148. \[aq]noattr\[aq] will short circuit as if nothing exists.
  149. \[aq]nosys\[aq] will respond with ENOSYS as if xattrs are not supported
  150. or disabled.
  151. (default: passthrough)
  152. .IP \[bu] 2
  153. \f[B]link_cow=BOOL\f[]: When enabled if a regular file is opened which
  154. has a link count > 1 it will copy the file to a temporary file and
  155. rename over the original.
  156. Breaking the link and providing a basic copy\-on\-write function similar
  157. to cow\-shell.
  158. (default: false)
  159. .IP \[bu] 2
  160. \f[B]statfs=base|full\f[]: Controls how statfs works.
  161. \[aq]base\[aq] means it will always use all branches in statfs
  162. calculations.
  163. \[aq]full\[aq] is in effect path preserving and only includes drives
  164. where the path exists.
  165. (default: base)
  166. .IP \[bu] 2
  167. \f[B]statfs_ignore=none|ro|nc\f[]: \[aq]ro\[aq] will cause statfs
  168. calculations to ignore available space for branches mounted or tagged as
  169. \[aq]read\-only\[aq] or \[aq]no create\[aq].
  170. \[aq]nc\[aq] will ignore available space for branches tagged as \[aq]no
  171. create\[aq].
  172. (default: none)
  173. .IP \[bu] 2
  174. \f[B]posix_acl=BOOL\f[]: Enable POSIX ACL support (if supported by
  175. kernel and underlying filesystem).
  176. (default: false)
  177. .IP \[bu] 2
  178. \f[B]async_read=BOOL\f[]: Perform reads asynchronously.
  179. If disabled or unavailable the kernel will ensure there is at most one
  180. pending read request per file handle and will attempt to order requests
  181. by offset.
  182. (default: true)
  183. .IP \[bu] 2
  184. \f[B]fuse_msg_size=INT\f[]: Set the max number of pages per FUSE
  185. message.
  186. Only available on Linux >= 4.20 and ignored otherwise.
  187. (min: 1; max: 256; default: 256)
  188. .IP \[bu] 2
  189. \f[B]threads=INT\f[]: Number of threads to use in multithreaded mode.
  190. When set to zero it will attempt to discover and use the number of
  191. logical cores.
  192. If the lookup fails it will fall back to using 4.
  193. If the thread count is set negative it will look up the number of cores
  194. then divide by the absolute value.
  195. ie.
  196. threads=\-2 on an 8 core machine will result in 8 / 2 = 4 threads.
  197. There will always be at least 1 thread.
  198. NOTE: higher number of threads increases parallelism but usually
  199. decreases throughput.
  200. (default: 0)
  201. .IP \[bu] 2
  202. \f[B]fsname=STR\f[]: Sets the name of the filesystem as seen in
  203. \f[B]mount\f[], \f[B]df\f[], etc.
  204. Defaults to a list of the source paths concatenated together with the
  205. longest common prefix removed.
  206. .IP \[bu] 2
  207. \f[B]func.FUNC=POLICY\f[]: Sets the specific FUSE function\[aq]s policy.
  208. See below for the list of value types.
  209. Example: \f[B]func.getattr=newest\f[]
  210. .IP \[bu] 2
  211. \f[B]category.CATEGORY=POLICY\f[]: Sets policy of all FUSE functions in
  212. the provided category.
  213. See POLICIES section for defaults.
  214. Example: \f[B]category.create=mfs\f[]
  215. .IP \[bu] 2
  216. \f[B]cache.open=INT\f[]: \[aq]open\[aq] policy cache timeout in seconds.
  217. (default: 0)
  218. .IP \[bu] 2
  219. \f[B]cache.statfs=INT\f[]: \[aq]statfs\[aq] cache timeout in seconds.
  220. (default: 0)
  221. .IP \[bu] 2
  222. \f[B]cache.attr=INT\f[]: File attribute cache timeout in seconds.
  223. (default: 1)
  224. .IP \[bu] 2
  225. \f[B]cache.entry=INT\f[]: File name lookup cache timeout in seconds.
  226. (default: 1)
  227. .IP \[bu] 2
  228. \f[B]cache.negative_entry=INT\f[]: Negative file name lookup cache
  229. timeout in seconds.
  230. (default: 0)
  231. .IP \[bu] 2
  232. \f[B]cache.files=libfuse|off|partial|full|auto\-full\f[]: File page
  233. caching mode (default: libfuse)
  234. .IP \[bu] 2
  235. \f[B]cache.symlinks=BOOL\f[]: Cache symlinks (if supported by kernel)
  236. (default: false)
  237. .IP \[bu] 2
  238. \f[B]cache.readdir=BOOL\f[]: Cache readdir (if supported by kernel)
  239. (default: false)
  240. .IP \[bu] 2
  241. \f[B]direct_io\f[]: deprecated \- Bypass page cache.
  242. Use \f[C]cache.files=off\f[] instead.
  243. (default: false)
  244. .IP \[bu] 2
  245. \f[B]kernel_cache\f[]: deprecated \- Do not invalidate data cache on
  246. file open.
  247. Use \f[C]cache.files=full\f[] instead.
  248. (default: false)
  249. .IP \[bu] 2
  250. \f[B]auto_cache\f[]: deprecated \- Invalidate data cache if file mtime
  251. or size change.
  252. Use \f[C]cache.files=auto\-full\f[] instead.
  253. (default: false)
  254. .IP \[bu] 2
  255. \f[B]async_read\f[]: deprecated \- Perform reads asynchronously.
  256. Use \f[C]async_read=true\f[] instead.
  257. .IP \[bu] 2
  258. \f[B]sync_read\f[]: deprecated \- Perform reads synchronously.
  259. Use \f[C]async_read=false\f[] instead.
  260. .PP
  261. \f[B]NOTE:\f[] Options are evaluated in the order listed so if the
  262. options are \f[B]func.rmdir=rand,category.action=ff\f[] the
  263. \f[B]action\f[] category setting will override the \f[B]rmdir\f[]
  264. setting.
  265. .SS Value Types
  266. .IP \[bu] 2
  267. BOOL = \[aq]true\[aq] | \[aq]false\[aq]
  268. .IP \[bu] 2
  269. INT = [0,MAX_INT]
  270. .IP \[bu] 2
  271. SIZE = \[aq]NNM\[aq]; NN = INT, M = \[aq]K\[aq] | \[aq]M\[aq] |
  272. \[aq]G\[aq] | \[aq]T\[aq]
  273. .IP \[bu] 2
  274. STR = string
  275. .IP \[bu] 2
  276. FUNC = FUSE function
  277. .IP \[bu] 2
  278. CATEGORY = FUSE function category
  279. .IP \[bu] 2
  280. POLICY = mergerfs function policy
  281. .SS branches
  282. .PP
  283. The \[aq]branches\[aq] (formerly \[aq]srcmounts\[aq]) argument is a
  284. colon (\[aq]:\[aq]) delimited list of paths to be pooled together.
  285. It does not matter if the paths are on the same or different drives nor
  286. does it matter the filesystem.
  287. Used and available space will not be duplicated for paths on the same
  288. device and any features which aren\[aq]t supported by the underlying
  289. filesystem (such as file attributes or extended attributes) will return
  290. the appropriate errors.
  291. .PP
  292. To make it easier to include multiple branches mergerfs supports
  293. globbing (http://linux.die.net/man/7/glob).
  294. \f[B]The globbing tokens MUST be escaped when using via the shell else
  295. the shell itself will apply the glob itself.\f[]
  296. .PP
  297. Each branch can have a suffix of \f[C]=RW\f[] (read / write),
  298. \f[C]=RO\f[] (read\-only), or \f[C]=NC\f[] (no create).
  299. These suffixes work with globs as well and will apply to each path
  300. found.
  301. \f[C]RW\f[] is the default behavior and those paths will be eligible for
  302. all policy categories.
  303. \f[C]RO\f[] will exclude those paths from \f[C]create\f[] and
  304. \f[C]action\f[] policies (just as a filesystem being mounted \f[C]ro\f[]
  305. would).
  306. \f[C]NC\f[] will exclude those paths from \f[C]create\f[] policies (you
  307. can\[aq]t create but you can change / delete).
  308. .IP
  309. .nf
  310. \f[C]
  311. #\ mergerfs\ \-o\ allow_other,use_ino\ /mnt/disk\\*:/mnt/cdrom\ /media/drives
  312. \f[]
  313. .fi
  314. .PP
  315. The above line will use all mount points in /mnt prefixed with
  316. \f[B]disk\f[] and the \f[B]cdrom\f[].
  317. .PP
  318. To have the pool mounted at boot or otherwise accessable from related
  319. tools use \f[B]/etc/fstab\f[].
  320. .IP
  321. .nf
  322. \f[C]
  323. #\ <file\ system>\ \ \ \ \ \ \ \ <mount\ point>\ \ <type>\ \ \ \ \ \ \ \ \ <options>\ \ \ \ \ \ \ \ \ \ \ \ \ <dump>\ \ <pass>
  324. /mnt/disk*:/mnt/cdrom\ \ /media/drives\ \ fuse.mergerfs\ \ allow_other,use_ino\ \ \ 0\ \ \ \ \ \ \ 0
  325. \f[]
  326. .fi
  327. .PP
  328. \f[B]NOTE:\f[] the globbing is done at mount or xattr update time (see
  329. below).
  330. If a new directory is added matching the glob after the fact it will not
  331. be automatically included.
  332. .PP
  333. \f[B]NOTE:\f[] for mounting via \f[B]fstab\f[] to work you must have
  334. \f[B]mount.fuse\f[] installed.
  335. For Ubuntu/Debian it is included in the \f[B]fuse\f[] package.
  336. .SS fuse_msg_size
  337. .PP
  338. FUSE applications communicate with the kernel over a special character
  339. device: \f[C]/dev/fuse\f[].
  340. A large portion of the overhead associated with FUSE is the cost of
  341. going back and forth from user space and kernel space over that device.
  342. Generally speaking the fewer trips needed the better the performance
  343. will be.
  344. Reducing the number of trips can be done a number of ways.
  345. Kernel level caching and increasing message sizes being two significant
  346. ones.
  347. When it comes to reads and writes if the message size is doubled the
  348. number of trips are appoximately halved.
  349. .PP
  350. In Linux 4.20 a new feature was added allowing the negotiation of the
  351. max message size.
  352. Since the size is in multiples of
  353. pages (https://en.wikipedia.org/wiki/Page_(computer_memory)) the feature
  354. is called \f[C]max_pages\f[].
  355. There is a maximum \f[C]max_pages\f[] value of 256 (1MiB) and minimum of
  356. 1 (4KiB).
  357. The default used by Linux >=4.20, and hardcoded value used before 4.20,
  358. is 32 (128KiB).
  359. In mergerfs its referred to as \f[C]fuse_msg_size\f[] to make it clear
  360. what it impacts and provide some abstraction.
  361. .PP
  362. Since there should be no downsides to increasing \f[C]fuse_msg_size\f[]
  363. / \f[C]max_pages\f[], outside a minor bump in RAM usage due to larger
  364. message buffers, mergerfs defaults the value to 256.
  365. On kernels before 4.20 the value has no effect.
  366. The reason the value is configurable is to enable experimentation and
  367. benchmarking.
  368. See the \f[C]nullrw\f[] section for benchmarking examples.
  369. .SS symlinkify
  370. .PP
  371. Due to the levels of indirection introduced by mergerfs and the
  372. underlying technology FUSE there can be varying levels of performance
  373. degredation.
  374. This feature will turn non\-directories which are not writable into
  375. symlinks to the original file found by the \f[C]readlink\f[] policy
  376. after the mtime and ctime are older than the timeout.
  377. .PP
  378. \f[B]WARNING:\f[] The current implementation has a known issue in which
  379. if the file is open and being used when the file is converted to a
  380. symlink then the application which has that file open will receive an
  381. error when using it.
  382. This is unlikely to occur in practice but is something to keep in mind.
  383. .PP
  384. \f[B]WARNING:\f[] Some backup solutions, such as CrashPlan, do not
  385. backup the target of a symlink.
  386. If using this feature it will be necessary to point any backup software
  387. to the original drives or configure the software to follow symlinks if
  388. such an option is available.
  389. Alternatively create two mounts.
  390. One for backup and one for general consumption.
  391. .SS nullrw
  392. .PP
  393. Due to how FUSE works there is an overhead to all requests made to a
  394. FUSE filesystem.
  395. Meaning that even a simple passthrough will have some slowdown.
  396. However, generally the overhead is minimal in comparison to the cost of
  397. the underlying I/O.
  398. By disabling the underlying I/O we can test the theoretical performance
  399. boundries.
  400. .PP
  401. By enabling \f[C]nullrw\f[] mergerfs will work as it always does
  402. \f[B]except\f[] that all reads and writes will be no\-ops.
  403. A write will succeed (the size of the write will be returned as if it
  404. were successful) but mergerfs does nothing with the data it was given.
  405. Similarly a read will return the size requested but won\[aq]t touch the
  406. buffer.
  407. .PP
  408. Example:
  409. .IP
  410. .nf
  411. \f[C]
  412. $\ dd\ if=/dev/zero\ of=/path/to/mergerfs/mount/benchmark\ ibs=1M\ obs=512\ count=1024\ iflag=dsync,nocache\ oflag=dsync,nocache\ conv=fdatasync\ status=progress
  413. 1024+0\ records\ in
  414. 2097152+0\ records\ out
  415. 1073741824\ bytes\ (1.1\ GB,\ 1.0\ GiB)\ copied,\ 15.4067\ s,\ 69.7\ MB/s
  416. $\ dd\ if=/dev/zero\ of=/path/to/mergerfs/mount/benchmark\ ibs=1M\ obs=1M\ count=1024\ iflag=dsync,nocache\ oflag=dsync,nocache\ conv=fdatasync\ status=progress
  417. 1024+0\ records\ in
  418. 1024+0\ records\ out
  419. 1073741824\ bytes\ (1.1\ GB,\ 1.0\ GiB)\ copied,\ 0.219585\ s,\ 4.9\ GB/s
  420. $\ dd\ if=/path/to/mergerfs/mount/benchmark\ of=/dev/null\ bs=512\ count=102400\ iflag=dsync,nocache\ oflag=dsync,nocache\ conv=fdatasync\ status=progress
  421. 102400+0\ records\ in
  422. 102400+0\ records\ out
  423. 52428800\ bytes\ (52\ MB,\ 50\ MiB)\ copied,\ 0.757991\ s,\ 69.2\ MB/s
  424. $\ dd\ if=/path/to/mergerfs/mount/benchmark\ of=/dev/null\ bs=1M\ count=1024\ iflag=dsync,nocache\ oflag=dsync,nocache\ conv=fdatasync\ status=progress
  425. 1024+0\ records\ in
  426. 1024+0\ records\ out
  427. 1073741824\ bytes\ (1.1\ GB,\ 1.0\ GiB)\ copied,\ 0.18405\ s,\ 5.8\ GB/s
  428. \f[]
  429. .fi
  430. .PP
  431. It\[aq]s important to test with different \f[C]obs\f[] (output block
  432. size) values since the relative overhead is greater with smaller values.
  433. As you can see above the size of a read or write can massively impact
  434. theoretical performance.
  435. If an application performs much worse through mergerfs it could very
  436. well be that it doesn\[aq]t optimally size its read and write requests.
  437. In such cases contact the mergerfs author so it can be investigated.
  438. .SS xattr
  439. .PP
  440. Runtime extended attribute support can be managed via the \f[C]xattr\f[]
  441. option.
  442. By default it will passthrough any xattr calls.
  443. Given xattr support is rarely used and can have significant performance
  444. implications mergerfs allows it to be disabled at runtime.
  445. .PP
  446. \f[C]noattr\f[] will cause mergerfs to short circuit all xattr calls and
  447. return ENOATTR where appropriate.
  448. mergerfs still gets all the requests but they will not be forwarded on
  449. to the underlying filesystems.
  450. The runtime control will still function in this mode.
  451. .PP
  452. \f[C]nosys\f[] will cause mergerfs to return ENOSYS for any xattr call.
  453. The difference with \f[C]noattr\f[] is that the kernel will cache this
  454. fact and itself short circuit future calls.
  455. This will be more efficient than \f[C]noattr\f[] but will cause
  456. mergerfs\[aq] runtime control via the hidden file to stop working.
  457. .SH FUNCTIONS / POLICIES / CATEGORIES
  458. .PP
  459. The POSIX filesystem API is made up of a number of functions.
  460. \f[B]creat\f[], \f[B]stat\f[], \f[B]chown\f[], etc.
  461. In mergerfs most of the core functions are grouped into 3 categories:
  462. \f[B]action\f[], \f[B]create\f[], and \f[B]search\f[].
  463. These functions and categories can be assigned a policy which dictates
  464. what file or directory is chosen when performing that behavior.
  465. Any policy can be assigned to a function or category though some may not
  466. be very useful in practice.
  467. For instance: \f[B]rand\f[] (random) may be useful for file creation
  468. (create) but could lead to very odd behavior if used for \f[C]chmod\f[]
  469. if there were more than one copy of the file.
  470. .PP
  471. Some functions, listed in the category \f[C]N/A\f[] below, can not be
  472. assigned the normal policies.
  473. All functions which work on file handles use the handle which was
  474. acquired by \f[C]open\f[] or \f[C]create\f[].
  475. \f[C]readdir\f[] has no real need for a policy given the purpose is
  476. merely to return a list of entries in a directory.
  477. \f[C]statfs\f[]\[aq]s behavior can be modified via other options.
  478. That said many times the current FUSE kernel driver will not always
  479. provide the file handle when a client calls \f[C]fgetattr\f[],
  480. \f[C]fchown\f[], \f[C]fchmod\f[], \f[C]futimens\f[], \f[C]ftruncate\f[],
  481. etc.
  482. This means it will call the regular, path based, versions.
  483. .PP
  484. When using policies which are based on a branch\[aq]s available space
  485. the base path provided is used.
  486. Not the full path to the file in question.
  487. Meaning that sub mounts won\[aq]t be considered in the space
  488. calculations.
  489. The reason is that it doesn\[aq]t really work for non\-path preserving
  490. policies and can lead to non\-obvious behaviors.
  491. .SS Function / Category classifications
  492. .PP
  493. .TS
  494. tab(@);
  495. lw(7.9n) lw(62.1n).
  496. T{
  497. Category
  498. T}@T{
  499. FUSE Functions
  500. T}
  501. _
  502. T{
  503. action
  504. T}@T{
  505. chmod, chown, link, removexattr, rename, rmdir, setxattr, truncate,
  506. unlink, utimens
  507. T}
  508. T{
  509. create
  510. T}@T{
  511. create, mkdir, mknod, symlink
  512. T}
  513. T{
  514. search
  515. T}@T{
  516. access, getattr, getxattr, ioctl (directories), listxattr, open,
  517. readlink
  518. T}
  519. T{
  520. N/A
  521. T}@T{
  522. fchmod, fchown, futimens, ftruncate, fallocate, fgetattr, fsync, ioctl
  523. (files), read, readdir, release, statfs, write, copy_file_range
  524. T}
  525. .TE
  526. .PP
  527. In cases where something may be searched (to confirm a directory exists
  528. across all source mounts) \f[B]getattr\f[] will be used.
  529. .SS Path Preservation
  530. .PP
  531. Policies, as described below, are of two basic types.
  532. \f[C]path\ preserving\f[] and \f[C]non\-path\ preserving\f[].
  533. .PP
  534. All policies which start with \f[C]ep\f[] (\f[B]epff\f[],
  535. \f[B]eplfs\f[], \f[B]eplus\f[], \f[B]epmfs\f[], \f[B]eprand\f[]) are
  536. \f[C]path\ preserving\f[].
  537. \f[C]ep\f[] stands for \f[C]existing\ path\f[].
  538. .PP
  539. A path preserving policy will only consider drives where the relative
  540. path being accessed already exists.
  541. .PP
  542. When using non\-path preserving policies paths will be cloned to target
  543. drives as necessary.
  544. .SS Filters
  545. .PP
  546. Policies basically search branches and create a list of files / paths
  547. for functions to work on.
  548. The policy is responsible for filtering and sorting.
  549. The policy type defines the sorting but filtering is mostly uniform as
  550. described below.
  551. .IP \[bu] 2
  552. No \f[B]search\f[] policies filter.
  553. .IP \[bu] 2
  554. All \f[B]action\f[] policies will filter out branches which are mounted
  555. \f[B]read\-only\f[] or tagged as \f[B]RO (read\-only)\f[].
  556. .IP \[bu] 2
  557. All \f[B]create\f[] policies will filter out branches which are mounted
  558. \f[B]read\-only\f[], tagged \f[B]RO (read\-only)\f[] or \f[B]NC (no
  559. create)\f[], or has available space less than \f[C]minfreespace\f[].
  560. .PP
  561. If all branches are filtered an error will be returned.
  562. Typically \f[B]EROFS\f[] or \f[B]ENOSPC\f[] depending on the reasons.
  563. .SS Policy descriptions
  564. .PP
  565. .TS
  566. tab(@);
  567. lw(16.6n) lw(53.4n).
  568. T{
  569. Policy
  570. T}@T{
  571. Description
  572. T}
  573. _
  574. T{
  575. all
  576. T}@T{
  577. Search category: same as \f[B]epall\f[].
  578. Action category: same as \f[B]epall\f[].
  579. Create category: for \f[B]mkdir\f[], \f[B]mknod\f[], and
  580. \f[B]symlink\f[] it will apply to all branches.
  581. \f[B]create\f[] works like \f[B]ff\f[].
  582. T}
  583. T{
  584. epall (existing path, all)
  585. T}@T{
  586. Search category: same as \f[B]epff\f[] (but more expensive because it
  587. doesn\[aq]t stop after finding a valid branch).
  588. Action category: apply to all found.
  589. Create category: for \f[B]mkdir\f[], \f[B]mknod\f[], and
  590. \f[B]symlink\f[] it will apply to all found.
  591. \f[B]create\f[] works like \f[B]epff\f[] (but more expensive because it
  592. doesn\[aq]t stop after finding a valid branch).
  593. T}
  594. T{
  595. epff (existing path, first found)
  596. T}@T{
  597. Given the order of the branches, as defined at mount time or configured
  598. at runtime, act on the first one found where the relative path exists.
  599. T}
  600. T{
  601. eplfs (existing path, least free space)
  602. T}@T{
  603. Of all the branches on which the relative path exists choose the drive
  604. with the least free space.
  605. T}
  606. T{
  607. eplus (existing path, least used space)
  608. T}@T{
  609. Of all the branches on which the relative path exists choose the drive
  610. with the least used space.
  611. T}
  612. T{
  613. epmfs (existing path, most free space)
  614. T}@T{
  615. Of all the branches on which the relative path exists choose the drive
  616. with the most free space.
  617. T}
  618. T{
  619. eprand (existing path, random)
  620. T}@T{
  621. Calls \f[B]epall\f[] and then randomizes.
  622. T}
  623. T{
  624. erofs
  625. T}@T{
  626. Exclusively return \f[B]\-1\f[] with \f[B]errno\f[] set to
  627. \f[B]EROFS\f[] (read\-only filesystem).
  628. T}
  629. T{
  630. ff (first found)
  631. T}@T{
  632. Search category: same as \f[B]epff\f[].
  633. Action category: same as \f[B]epff\f[].
  634. Create category: Given the order of the drives, as defined at mount time
  635. or configured at runtime, act on the first one found.
  636. T}
  637. T{
  638. lfs (least free space)
  639. T}@T{
  640. Search category: same as \f[B]eplfs\f[].
  641. Action category: same as \f[B]eplfs\f[].
  642. Create category: Pick the drive with the least available free space.
  643. T}
  644. T{
  645. lus (least used space)
  646. T}@T{
  647. Search category: same as \f[B]eplus\f[].
  648. Action category: same as \f[B]eplus\f[].
  649. Create category: Pick the drive with the least used space.
  650. T}
  651. T{
  652. mfs (most free space)
  653. T}@T{
  654. Search category: same as \f[B]epmfs\f[].
  655. Action category: same as \f[B]epmfs\f[].
  656. Create category: Pick the drive with the most available free space.
  657. T}
  658. T{
  659. newest
  660. T}@T{
  661. Pick the file / directory with the largest mtime.
  662. T}
  663. T{
  664. rand (random)
  665. T}@T{
  666. Calls \f[B]all\f[] and then randomizes.
  667. T}
  668. .TE
  669. .SS Defaults
  670. .PP
  671. .TS
  672. tab(@);
  673. l l.
  674. T{
  675. Category
  676. T}@T{
  677. Policy
  678. T}
  679. _
  680. T{
  681. action
  682. T}@T{
  683. epall
  684. T}
  685. T{
  686. create
  687. T}@T{
  688. epmfs
  689. T}
  690. T{
  691. search
  692. T}@T{
  693. ff
  694. T}
  695. .TE
  696. .SS ioctl
  697. .PP
  698. When \f[C]ioctl\f[] is used with an open file then it will use the file
  699. handle which was created at the original \f[C]open\f[] call.
  700. However, when using \f[C]ioctl\f[] with a directory mergerfs will use
  701. the \f[C]open\f[] policy to find the directory to act on.
  702. .SS unlink
  703. .PP
  704. In FUSE there is an opaque "file handle" which is created by
  705. \f[C]open\f[], \f[C]create\f[], or \f[C]opendir\f[], passed to the
  706. kernel, and then is passed back to the FUSE userland application by the
  707. kernel.
  708. Unfortunately, the FUSE kernel driver does not always send the file
  709. handle when it theoretically could/should.
  710. This complicates certain behaviors / workflows particularly in the high
  711. level API.
  712. As a result mergerfs is currently doing a few hacky things.
  713. .PP
  714. libfuse2 and libfuse3, when using the high level API, will rename names
  715. to \f[C]\&.fuse_hiddenXXXXXX\f[] if the file is open when unlinked or
  716. renamed over.
  717. It does this so the file is still available when a request referencing
  718. the now missing file is made.
  719. This file however keeps a \f[C]rmdir\f[] from succeeding and can be
  720. picked up by software reading directories.
  721. .PP
  722. The change mergerfs has done is that if a file is open when an unlink or
  723. rename happens it will open the file and keep it open till closed by all
  724. those who opened it prior.
  725. When a request comes in referencing that file and it doesn\[aq]t include
  726. a file handle it will instead use the file handle created at
  727. unlink/rename time.
  728. .PP
  729. This won\[aq]t result in technically proper behavior but close enough
  730. for many usecases.
  731. .PP
  732. The plan is to rewrite mergerfs to use the low level API so these
  733. invasive libfuse changes are no longer necessary.
  734. .SS rename & link
  735. .PP
  736. \f[B]NOTE:\f[] If you\[aq]re receiving errors from software when files
  737. are moved / renamed / linked then you should consider changing the
  738. create policy to one which is \f[B]not\f[] path preserving, enabling
  739. \f[C]ignorepponrename\f[], or contacting the author of the offending
  740. software and requesting that \f[C]EXDEV\f[] be properly handled.
  741. .PP
  742. \f[C]rename\f[] and \f[C]link\f[] are tricky functions in a union
  743. filesystem.
  744. \f[C]rename\f[] only works within a single filesystem or device.
  745. If a rename can\[aq]t be done atomically due to the source and
  746. destination paths existing on different mount points it will return
  747. \f[B]\-1\f[] with \f[B]errno = EXDEV\f[] (cross device).
  748. So if a \f[C]rename\f[]\[aq]s source and target are on different drives
  749. within the pool it creates an issue.
  750. .PP
  751. Originally mergerfs would return EXDEV whenever a rename was requested
  752. which was cross directory in any way.
  753. This made the code simple and was technically complient with POSIX
  754. requirements.
  755. However, many applications fail to handle EXDEV at all and treat it as a
  756. normal error or otherwise handle it poorly.
  757. Such apps include: gvfsd\-fuse v1.20.3 and prior, Finder / CIFS/SMB
  758. client in Apple OSX 10.9+, NZBGet, Samba\[aq]s recycling bin feature.
  759. .PP
  760. As a result a compromise was made in order to get most software to work
  761. while still obeying mergerfs\[aq] policies.
  762. Below is the basic logic.
  763. .IP \[bu] 2
  764. If using a \f[B]create\f[] policy which tries to preserve directory
  765. paths (epff,eplfs,eplus,epmfs)
  766. .IP \[bu] 2
  767. Using the \f[B]rename\f[] policy get the list of files to rename
  768. .IP \[bu] 2
  769. For each file attempt rename:
  770. .RS 2
  771. .IP \[bu] 2
  772. If failure with ENOENT run \f[B]create\f[] policy
  773. .IP \[bu] 2
  774. If create policy returns the same drive as currently evaluating then
  775. clone the path
  776. .IP \[bu] 2
  777. Re\-attempt rename
  778. .RE
  779. .IP \[bu] 2
  780. If \f[B]any\f[] of the renames succeed the higher level rename is
  781. considered a success
  782. .IP \[bu] 2
  783. If \f[B]no\f[] renames succeed the first error encountered will be
  784. returned
  785. .IP \[bu] 2
  786. On success:
  787. .RS 2
  788. .IP \[bu] 2
  789. Remove the target from all drives with no source file
  790. .IP \[bu] 2
  791. Remove the source from all drives which failed to rename
  792. .RE
  793. .IP \[bu] 2
  794. If using a \f[B]create\f[] policy which does \f[B]not\f[] try to
  795. preserve directory paths
  796. .IP \[bu] 2
  797. Using the \f[B]rename\f[] policy get the list of files to rename
  798. .IP \[bu] 2
  799. Using the \f[B]getattr\f[] policy get the target path
  800. .IP \[bu] 2
  801. For each file attempt rename:
  802. .RS 2
  803. .IP \[bu] 2
  804. If the source drive != target drive:
  805. .IP \[bu] 2
  806. Clone target path from target drive to source drive
  807. .IP \[bu] 2
  808. Rename
  809. .RE
  810. .IP \[bu] 2
  811. If \f[B]any\f[] of the renames succeed the higher level rename is
  812. considered a success
  813. .IP \[bu] 2
  814. If \f[B]no\f[] renames succeed the first error encountered will be
  815. returned
  816. .IP \[bu] 2
  817. On success:
  818. .RS 2
  819. .IP \[bu] 2
  820. Remove the target from all drives with no source file
  821. .IP \[bu] 2
  822. Remove the source from all drives which failed to rename
  823. .RE
  824. .PP
  825. The the removals are subject to normal entitlement checks.
  826. .PP
  827. The above behavior will help minimize the likelihood of EXDEV being
  828. returned but it will still be possible.
  829. .PP
  830. \f[B]link\f[] uses the same strategy but without the removals.
  831. .SS readdir
  832. .PP
  833. readdir (http://linux.die.net/man/3/readdir) is different from all other
  834. filesystem functions.
  835. While it could have its own set of policies to tweak its behavior at
  836. this time it provides a simple union of files and directories found.
  837. Remember that any action or information queried about these files and
  838. directories come from the respective function.
  839. For instance: an \f[B]ls\f[] is a \f[B]readdir\f[] and for each
  840. file/directory returned \f[B]getattr\f[] is called.
  841. Meaning the policy of \f[B]getattr\f[] is responsible for choosing the
  842. file/directory which is the source of the metadata you see in an
  843. \f[B]ls\f[].
  844. .SS statfs / statvfs
  845. .PP
  846. statvfs (http://linux.die.net/man/2/statvfs) normalizes the source
  847. drives based on the fragment size and sums the number of adjusted blocks
  848. and inodes.
  849. This means you will see the combined space of all sources.
  850. Total, used, and free.
  851. The sources however are dedupped based on the drive so multiple sources
  852. on the same drive will not result in double counting its space.
  853. Filesystems mounted further down the tree of the branch will not be
  854. included when checking the mount\[aq]s stats.
  855. .PP
  856. The options \f[C]statfs\f[] and \f[C]statfs_ignore\f[] can be used to
  857. modify \f[C]statfs\f[] behavior.
  858. .SH BUILDING
  859. .PP
  860. \f[B]NOTE:\f[] Prebuilt packages can be found at:
  861. https://github.com/trapexit/mergerfs/releases
  862. .PP
  863. First get the code from github (https://github.com/trapexit/mergerfs).
  864. .IP
  865. .nf
  866. \f[C]
  867. $\ git\ clone\ https://github.com/trapexit/mergerfs.git
  868. $\ #\ or
  869. $\ wget\ https://github.com/trapexit/mergerfs/releases/download/<ver>/mergerfs\-<ver>.tar.gz
  870. \f[]
  871. .fi
  872. .SS Debian / Ubuntu
  873. .IP
  874. .nf
  875. \f[C]
  876. $\ cd\ mergerfs
  877. $\ sudo\ tools/install\-build\-pkgs
  878. $\ make\ deb
  879. $\ sudo\ dpkg\ \-i\ ../mergerfs_version_arch.deb
  880. \f[]
  881. .fi
  882. .SS Fedora
  883. .IP
  884. .nf
  885. \f[C]
  886. $\ su\ \-
  887. #\ cd\ mergerfs
  888. #\ tools/install\-build\-pkgs
  889. #\ make\ rpm
  890. #\ rpm\ \-i\ rpmbuild/RPMS/<arch>/mergerfs\-<verion>.<arch>.rpm
  891. \f[]
  892. .fi
  893. .SS Generically
  894. .PP
  895. Have git, g++, make, python installed.
  896. .IP
  897. .nf
  898. \f[C]
  899. $\ cd\ mergerfs
  900. $\ make
  901. $\ sudo\ make\ install
  902. \f[]
  903. .fi
  904. .SS Build options
  905. .IP
  906. .nf
  907. \f[C]
  908. $\ make\ help
  909. usage:\ make
  910. make\ USE_XATTR=0\ \ \ \ \ \ \-\ build\ program\ without\ xattrs\ functionality
  911. make\ STATIC=1\ \ \ \ \ \ \ \ \ \-\ build\ static\ binary
  912. make\ LTO=1\ \ \ \ \ \ \ \ \ \ \ \ \-\ build\ with\ link\ time\ optimization
  913. \f[]
  914. .fi
  915. .SH RUNTIME CONFIG
  916. .SS .mergerfs pseudo file
  917. .IP
  918. .nf
  919. \f[C]
  920. <mountpoint>/.mergerfs
  921. \f[]
  922. .fi
  923. .PP
  924. There is a pseudo file available at the mount point which allows for the
  925. runtime modification of certain \f[B]mergerfs\f[] options.
  926. The file will not show up in \f[B]readdir\f[] but can be
  927. \f[B]stat\f[]\[aq]ed and manipulated via
  928. {list,get,set}xattrs (http://linux.die.net/man/2/listxattr) calls.
  929. .PP
  930. Any changes made at runtime are \f[B]not\f[] persisted.
  931. If you wish for values to persist they must be included as options
  932. wherever you configure the mounting of mergerfs (/etc/fstab).
  933. .SS Keys
  934. .PP
  935. Use \f[C]xattr\ \-l\ /mountpoint/.mergerfs\f[] to see all supported
  936. keys.
  937. Some are informational and therefore read\-only.
  938. \f[C]setxattr\f[] will return EINVAL on read\-only keys.
  939. .SS Values
  940. .PP
  941. Same as the command line.
  942. .SS user.mergerfs.branches
  943. .PP
  944. \f[B]NOTE:\f[] formerly \f[C]user.mergerfs.srcmounts\f[] but said key is
  945. still supported.
  946. .PP
  947. Used to query or modify the list of branches.
  948. When modifying there are several shortcuts to easy manipulation of the
  949. list.
  950. .PP
  951. .TS
  952. tab(@);
  953. l l.
  954. T{
  955. Value
  956. T}@T{
  957. Description
  958. T}
  959. _
  960. T{
  961. [list]
  962. T}@T{
  963. set
  964. T}
  965. T{
  966. +<[list]
  967. T}@T{
  968. prepend
  969. T}
  970. T{
  971. +>[list]
  972. T}@T{
  973. append
  974. T}
  975. T{
  976. \-[list]
  977. T}@T{
  978. remove all values provided
  979. T}
  980. T{
  981. \-<
  982. T}@T{
  983. remove first in list
  984. T}
  985. T{
  986. \->
  987. T}@T{
  988. remove last in list
  989. T}
  990. .TE
  991. .PP
  992. \f[C]xattr\ \-w\ user.mergerfs.branches\ +</mnt/drive3\ /mnt/pool/.mergerfs\f[]
  993. .PP
  994. The \f[C]=NC\f[], \f[C]=RO\f[], \f[C]=RW\f[] syntax works just as on the
  995. command line.
  996. .SS Example
  997. .IP
  998. .nf
  999. \f[C]
  1000. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-l\ .mergerfs
  1001. user.mergerfs.branches:\ /mnt/a=RW:/mnt/b=RW
  1002. user.mergerfs.minfreespace:\ 4294967295
  1003. user.mergerfs.moveonenospc:\ false
  1004. \&...
  1005. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-p\ user.mergerfs.category.search\ .mergerfs
  1006. ff
  1007. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-w\ user.mergerfs.category.search\ newest\ .mergerfs
  1008. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-p\ user.mergerfs.category.search\ .mergerfs
  1009. newest
  1010. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-w\ user.mergerfs.branches\ +/mnt/c\ .mergerfs
  1011. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-p\ user.mergerfs.branches\ .mergerfs
  1012. /mnt/a:/mnt/b:/mnt/c
  1013. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-w\ user.mergerfs.branches\ =/mnt/c\ .mergerfs
  1014. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-p\ user.mergerfs.branches\ .mergerfs
  1015. /mnt/c
  1016. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-w\ user.mergerfs.branches\ \[aq]+</mnt/a:/mnt/b\[aq]\ .mergerfs
  1017. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-p\ user.mergerfs.branches\ .mergerfs
  1018. /mnt/a:/mnt/b:/mnt/c
  1019. \f[]
  1020. .fi
  1021. .SS file / directory xattrs
  1022. .PP
  1023. While they won\[aq]t show up when using
  1024. listxattr (http://linux.die.net/man/2/listxattr) \f[B]mergerfs\f[]
  1025. offers a number of special xattrs to query information about the files
  1026. served.
  1027. To access the values you will need to issue a
  1028. getxattr (http://linux.die.net/man/2/getxattr) for one of the following:
  1029. .IP \[bu] 2
  1030. \f[B]user.mergerfs.basepath\f[]: the base mount point for the file given
  1031. the current getattr policy
  1032. .IP \[bu] 2
  1033. \f[B]user.mergerfs.relpath\f[]: the relative path of the file from the
  1034. perspective of the mount point
  1035. .IP \[bu] 2
  1036. \f[B]user.mergerfs.fullpath\f[]: the full path of the original file
  1037. given the getattr policy
  1038. .IP \[bu] 2
  1039. \f[B]user.mergerfs.allpaths\f[]: a NUL (\[aq]\[aq]) separated list of
  1040. full paths to all files found
  1041. .IP
  1042. .nf
  1043. \f[C]
  1044. [trapexit:/mnt/mergerfs]\ $\ ls
  1045. A\ B\ C
  1046. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-p\ user.mergerfs.fullpath\ A
  1047. /mnt/a/full/path/to/A
  1048. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-p\ user.mergerfs.basepath\ A
  1049. /mnt/a
  1050. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-p\ user.mergerfs.relpath\ A
  1051. /full/path/to/A
  1052. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-p\ user.mergerfs.allpaths\ A\ |\ tr\ \[aq]\\0\[aq]\ \[aq]\\n\[aq]
  1053. /mnt/a/full/path/to/A
  1054. /mnt/b/full/path/to/A
  1055. \f[]
  1056. .fi
  1057. .SH TOOLING
  1058. .IP \[bu] 2
  1059. https://github.com/trapexit/mergerfs\-tools
  1060. .IP \[bu] 2
  1061. mergerfs.ctl: A tool to make it easier to query and configure mergerfs
  1062. at runtime
  1063. .IP \[bu] 2
  1064. mergerfs.fsck: Provides permissions and ownership auditing and the
  1065. ability to fix them
  1066. .IP \[bu] 2
  1067. mergerfs.dedup: Will help identify and optionally remove duplicate files
  1068. .IP \[bu] 2
  1069. mergerfs.dup: Ensure there are at least N copies of a file across the
  1070. pool
  1071. .IP \[bu] 2
  1072. mergerfs.balance: Rebalance files across drives by moving them from the
  1073. most filled to the least filled
  1074. .IP \[bu] 2
  1075. mergerfs.mktrash: Creates FreeDesktop.org Trash specification compatible
  1076. directories on a mergerfs mount
  1077. .IP \[bu] 2
  1078. https://github.com/trapexit/scorch
  1079. .IP \[bu] 2
  1080. scorch: A tool to help discover silent corruption of files and keep
  1081. track of files
  1082. .IP \[bu] 2
  1083. https://github.com/trapexit/bbf
  1084. .IP \[bu] 2
  1085. bbf (bad block finder): a tool to scan for and \[aq]fix\[aq] hard drive
  1086. bad blocks and find the files using those blocks
  1087. .SH CACHING
  1088. .SS page caching
  1089. .PP
  1090. https://en.wikipedia.org/wiki/Page_cache
  1091. .PP
  1092. tl;dr: * cache.files=off: Disables page caching.
  1093. Underlying files cached, mergerfs files are not.
  1094. * cache.files=partial: Enables page caching.
  1095. Underlying files cached, mergerfs files cached while open.
  1096. * cache.files=full: Enables page caching.
  1097. Underlying files cached, mergerfs files cached across opens.
  1098. * cache.files=auto\-full: Enables page caching.
  1099. Underlying files cached, mergerfs files cached across opens if mtime and
  1100. size are unchanged since previous open.
  1101. * cache.files=libfuse: follow traditional libfuse \f[C]direct_io\f[],
  1102. \f[C]kernel_cache\f[], and \f[C]auto_cache\f[] arguments.
  1103. .PP
  1104. FUSE, which mergerfs uses, offers a number of page caching modes.
  1105. mergerfs tries to simplify their use via the \f[C]cache.files\f[]
  1106. option.
  1107. It can and should replace usage of \f[C]direct_io\f[],
  1108. \f[C]kernel_cache\f[], and \f[C]auto_cache\f[].
  1109. .PP
  1110. Due to mergerfs using FUSE and therefore being a userland process
  1111. proxying existing filesystems the kernel will double cache the content
  1112. being read and written through mergerfs.
  1113. Once from the underlying filesystem and once from mergerfs (it sees them
  1114. as two separate entities).
  1115. Using \f[C]cache.files=off\f[] will keep the double caching from
  1116. happening by disabling caching of mergerfs but this has the side effect
  1117. that \f[I]all\f[] read and write calls will be passed to mergerfs which
  1118. may be slower than enabling caching, you lose shared \f[C]mmap\f[]
  1119. support which can affect apps such as rtorrent, and no read\-ahead will
  1120. take place.
  1121. The kernel will still cache the underlying filesystem data but that only
  1122. helps so much given mergerfs will still process all requests.
  1123. .PP
  1124. If you do enable file page caching,
  1125. \f[C]cache.files=partial|full|auto\-full\f[], you should also enable
  1126. \f[C]dropcacheonclose\f[] which will cause mergerfs to instruct the
  1127. kernel to flush the underlying file\[aq]s page cache when the file is
  1128. closed.
  1129. This behavior is the same as the rsync fadvise / drop cache patch and
  1130. Feh\[aq]s nocache project.
  1131. .PP
  1132. If most files are read once through and closed (like media) it is best
  1133. to enable \f[C]dropcacheonclose\f[] regardless of caching mode in order
  1134. to minimize buffer bloat.
  1135. .PP
  1136. It is difficult to balance memory usage, cache bloat & duplication, and
  1137. performance.
  1138. Ideally mergerfs would be able to disable caching for the files it
  1139. reads/writes but allow page caching for itself.
  1140. That would limit the FUSE overhead.
  1141. However, there isn\[aq]t a good way to achieve this.
  1142. It would need to open all files with O_DIRECT which places limitations
  1143. on the what underlying filesystems would be supported and complicates
  1144. the code.
  1145. .PP
  1146. kernel documenation:
  1147. https://www.kernel.org/doc/Documentation/filesystems/fuse\-io.txt
  1148. .SS entry & attribute caching
  1149. .PP
  1150. Given the relatively high cost of FUSE due to the kernel <\-> userspace
  1151. round trips there are kernel side caches for file entries and
  1152. attributes.
  1153. The entry cache limits the \f[C]lookup\f[] calls to mergerfs which ask
  1154. if a file exists.
  1155. The attribute cache limits the need to make \f[C]getattr\f[] calls to
  1156. mergerfs which provide file attributes (mode, size, type, etc.).
  1157. As with the page cache these should not be used if the underlying
  1158. filesystems are being manipulated at the same time as it could lead to
  1159. odd behavior or data corruption.
  1160. The options for setting these are \f[C]cache.entry\f[] and
  1161. \f[C]cache.negative_entry\f[] for the entry cache and
  1162. \f[C]cache.attr\f[] for the attributes cache.
  1163. \f[C]cache.negative_entry\f[] refers to the timeout for negative
  1164. responses to lookups (non\-existant files).
  1165. .SS policy caching
  1166. .PP
  1167. Policies are run every time a function (with a policy as mentioned
  1168. above) is called.
  1169. These policies can be expensive depending on mergerfs\[aq] setup and
  1170. client usage patterns.
  1171. Generally we wouldn\[aq]t want to cache policy results because it may
  1172. result in stale responses if the underlying drives are used directly.
  1173. .PP
  1174. The \f[C]open\f[] policy cache will cache the result of an \f[C]open\f[]
  1175. policy for a particular input for \f[C]cache.open\f[] seconds or until
  1176. the file is unlinked.
  1177. Each file close (release) will randomly chose to clean up the cache of
  1178. expired entries.
  1179. .PP
  1180. This cache is really only useful in cases where you have a large number
  1181. of branches and \f[C]open\f[] is called on the same files repeatedly
  1182. (like \f[B]Transmission\f[] which opens and closes a file on every
  1183. read/write presumably to keep file handle usage low).
  1184. .SS statfs caching
  1185. .PP
  1186. Of the syscalls used by mergerfs in policies the \f[C]statfs\f[] /
  1187. \f[C]statvfs\f[] call is perhaps the most expensive.
  1188. It\[aq]s used to find out the available space of a drive and whether it
  1189. is mounted read\-only.
  1190. Depending on the setup and usage pattern these queries can be relatively
  1191. costly.
  1192. When \f[C]cache.statfs\f[] is enabled all calls to \f[C]statfs\f[] by a
  1193. policy will be cached for the number of seconds its set to.
  1194. .PP
  1195. Example: If the create policy is \f[C]mfs\f[] and the timeout is 60 then
  1196. for that 60 seconds the same drive will be returned as the target for
  1197. creates because the available space won\[aq]t be updated for that time.
  1198. .SS symlink caching
  1199. .PP
  1200. As of version 4.20 Linux supports symlink caching.
  1201. Significant performance increases can be had in workloads which use a
  1202. lot of symlinks.
  1203. Setting \f[C]cache.symlinks=true\f[] will result in requesting symlink
  1204. caching from the kernel only if supported.
  1205. As a result its safe to enable it on systems prior to 4.20.
  1206. That said it is disabled by default for now.
  1207. You can see if caching is enabled by querying the xattr
  1208. \f[C]user.mergerfs.cache.symlinks\f[] but given it must be requested at
  1209. startup you can not change it at runtime.
  1210. .SS readdir caching
  1211. .PP
  1212. As of version 4.20 Linux supports readdir caching.
  1213. This can have a significant impact on directory traversal.
  1214. Especially when combined with entry (\f[C]cache.entry\f[]) and attribute
  1215. (\f[C]cache.attr\f[]) caching.
  1216. Setting \f[C]cache.readdir=true\f[] will result in requesting readdir
  1217. caching from the kernel on each \f[C]opendir\f[].
  1218. If the kernel doesn\[aq]t support readdir caching setting the option to
  1219. \f[C]true\f[] has no effect.
  1220. This option is configuarable at runtime via xattr
  1221. \f[C]user.mergerfs.cache.readdir\f[].
  1222. .SS writeback caching
  1223. .PP
  1224. writeback caching is a technique for improving write speeds by batching
  1225. writes at a faster device and then bulk writing to the slower device.
  1226. With FUSE the kernel will wait for a number of writes to be made and
  1227. then send it to the filesystem as one request.
  1228. mergerfs currently uses a modified and vendored libfuse 2.9.7 which does
  1229. not support writeback caching.
  1230. Adding said feature should not be difficult but benchmarking needs to be
  1231. done to see if what effect it will have.
  1232. .SS tiered caching
  1233. .PP
  1234. Some storage technologies support what some call "tiered" caching.
  1235. The placing of usually smaller, faster storage as a transparent cache to
  1236. larger, slower storage.
  1237. NVMe, SSD, Optane in front of traditional HDDs for instance.
  1238. .PP
  1239. MergerFS does not natively support any sort of tiered caching.
  1240. Most users have no use for such a feature and its inclusion would
  1241. complicate the code.
  1242. However, there are a few situations where a cache drive could help with
  1243. a typical mergerfs setup.
  1244. .IP "1." 3
  1245. Fast network, slow drives, many readers: You\[aq]ve a 10+Gbps network
  1246. with many readers and your regular drives can\[aq]t keep up.
  1247. .IP "2." 3
  1248. Fast network, slow drives, small\[aq]ish bursty writes: You have a
  1249. 10+Gbps network and wish to transfer amounts of data less than your
  1250. cache drive but wish to do so quickly.
  1251. .PP
  1252. With #1 its arguable if you should be using mergerfs at all.
  1253. RAID would probably be the better solution.
  1254. If you\[aq]re going to use mergerfs there are other tactics that may
  1255. help: spreading the data across drives (see the mergerfs.dup tool) and
  1256. setting \f[C]func.open=rand\f[], using \f[C]symlinkify\f[], or using
  1257. dm\-cache or a similar technology to add tiered cache to the underlying
  1258. device.
  1259. .PP
  1260. With #2 one could use dm\-cache as well but there is another solution
  1261. which requires only mergerfs and a cronjob.
  1262. .IP "1." 3
  1263. Create 2 mergerfs pools.
  1264. One which includes just the slow drives and one which has both the fast
  1265. drives (SSD,NVME,etc.) and slow drives.
  1266. .IP "2." 3
  1267. The \[aq]cache\[aq] pool should have the cache drives listed first.
  1268. .IP "3." 3
  1269. The best \f[C]create\f[] policies to use for the \[aq]cache\[aq] pool
  1270. would probably be \f[C]ff\f[], \f[C]epff\f[], \f[C]lfs\f[], or
  1271. \f[C]eplfs\f[].
  1272. The latter two under the assumption that the cache drive(s) are far
  1273. smaller than the backing drives.
  1274. If using path preserving policies remember that you\[aq]ll need to
  1275. manually create the core directories of those paths you wish to be
  1276. cached.
  1277. Be sure the permissions are in sync.
  1278. Use \f[C]mergerfs.fsck\f[] to check / correct them.
  1279. You could also tag the slow drives as \f[C]=NC\f[] though that\[aq]d
  1280. mean if the cache drives fill you\[aq]d get "out of space" errors.
  1281. .IP "4." 3
  1282. Enable \f[C]moveonenospc\f[] and set \f[C]minfreespace\f[]
  1283. appropriately.
  1284. Perhaps setting \f[C]minfreespace\f[] to the size of the largest cache
  1285. drive.
  1286. .IP "5." 3
  1287. Set your programs to use the cache pool.
  1288. .IP "6." 3
  1289. Save one of the below scripts or create you\[aq]re own.
  1290. .IP "7." 3
  1291. Use \f[C]cron\f[] (as root) to schedule the command at whatever
  1292. frequency is appropriate for your workflow.
  1293. .SS time based expiring
  1294. .PP
  1295. Move files from cache to backing pool based only on the last time the
  1296. file was accessed.
  1297. Replace \f[C]\-atime\f[] with \f[C]\-amin\f[] if you want minutes rather
  1298. than days.
  1299. May want to use the \f[C]fadvise\f[] / \f[C]\-\-drop\-cache\f[] version
  1300. of rsync or run rsync with the tool "nocache".
  1301. .IP
  1302. .nf
  1303. \f[C]
  1304. #!/bin/bash
  1305. if\ [\ $#\ !=\ 3\ ];\ then
  1306. \ \ echo\ "usage:\ $0\ <cache\-drive>\ <backing\-pool>\ <days\-old>"
  1307. \ \ exit\ 1
  1308. fi
  1309. CACHE="${1}"
  1310. BACKING="${2}"
  1311. N=${3}
  1312. find\ "${CACHE}"\ \-type\ f\ \-atime\ +${N}\ \-printf\ \[aq]%P\\n\[aq]\ |\ \\
  1313. \ \ rsync\ \-\-files\-from=\-\ \-axqHAXWES\ \-\-preallocate\ \-\-remove\-source\-files\ "${CACHE}/"\ "${BACKING}/"
  1314. \f[]
  1315. .fi
  1316. .SS percentage full expiring
  1317. .PP
  1318. Move the oldest file from the cache to the backing pool.
  1319. Continue till below percentage threshold.
  1320. .IP
  1321. .nf
  1322. \f[C]
  1323. #!/bin/bash
  1324. if\ [\ $#\ !=\ 3\ ];\ then
  1325. \ \ echo\ "usage:\ $0\ <cache\-drive>\ <backing\-pool>\ <percentage>"
  1326. \ \ exit\ 1
  1327. fi
  1328. CACHE="${1}"
  1329. BACKING="${2}"
  1330. PERCENTAGE=${3}
  1331. set\ \-o\ errexit
  1332. while\ [\ $(df\ \-\-output=pcent\ "${CACHE}"\ |\ grep\ \-v\ Use\ |\ cut\ \-d\[aq]%\[aq]\ \-f1)\ \-gt\ ${PERCENTAGE}\ ]
  1333. do
  1334. \ \ \ \ FILE=$(find\ "${CACHE}"\ \-type\ f\ \-printf\ \[aq]%A\@\ %P\\n\[aq]\ |\ \\
  1335. \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ sort\ |\ \\
  1336. \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ head\ \-n\ 1\ |\ \\
  1337. \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ cut\ \-d\[aq]\ \[aq]\ \-f2\-)
  1338. \ \ \ \ test\ \-n\ "${FILE}"
  1339. \ \ \ \ rsync\ \-axqHAXWES\ \-\-preallocate\ \-\-remove\-source\-files\ "${CACHE}/./${FILE}"\ "${BACKING}/"
  1340. done
  1341. \f[]
  1342. .fi
  1343. .SH TIPS / NOTES
  1344. .IP \[bu] 2
  1345. \f[B]use_ino\f[] will only work when used with mergerfs 2.18.0 and
  1346. above.
  1347. .IP \[bu] 2
  1348. Run mergerfs as \f[C]root\f[] (with \f[B]allow_other\f[]) unless
  1349. you\[aq]re merging paths which are owned by the same user otherwise
  1350. strange permission issues may arise.
  1351. .IP \[bu] 2
  1352. https://github.com/trapexit/backup\-and\-recovery\-howtos : A set of
  1353. guides / howtos on creating a data storage system, backing it up,
  1354. maintaining it, and recovering from failure.
  1355. .IP \[bu] 2
  1356. If you don\[aq]t see some directories and files you expect in a merged
  1357. point or policies seem to skip drives be sure the user has permission to
  1358. all the underlying directories.
  1359. Use \f[C]mergerfs.fsck\f[] to audit the drive for out of sync
  1360. permissions.
  1361. .IP \[bu] 2
  1362. Do \f[B]not\f[] use \f[C]cache.files=off\f[] or \f[C]direct_io\f[] if
  1363. you expect applications (such as rtorrent) to
  1364. mmap (http://linux.die.net/man/2/mmap) files.
  1365. Shared mmap is not currently supported in FUSE w/ \f[C]direct_io\f[]
  1366. enabled.
  1367. Enabling \f[C]dropcacheonclose\f[] is recommended when
  1368. \f[C]cache.files=partial|full|auto\-full\f[] or
  1369. \f[C]direct_io=false\f[].
  1370. .IP \[bu] 2
  1371. Since POSIX functions give only a singular error or success its
  1372. difficult to determine the proper behavior when applying the function to
  1373. multiple targets.
  1374. \f[B]mergerfs\f[] will return an error only if all attempts of an action
  1375. fail.
  1376. Any success will lead to a success returned.
  1377. This means however that some odd situations may arise.
  1378. .IP \[bu] 2
  1379. Kodi (http://kodi.tv), Plex (http://plex.tv),
  1380. Subsonic (http://subsonic.org), etc.
  1381. can use directory mtime (http://linux.die.net/man/2/stat) to more
  1382. efficiently determine whether to scan for new content rather than simply
  1383. performing a full scan.
  1384. If using the default \f[B]getattr\f[] policy of \f[B]ff\f[] its possible
  1385. those programs will miss an update on account of it returning the first
  1386. directory found\[aq]s \f[B]stat\f[] info and its a later directory on
  1387. another mount which had the \f[B]mtime\f[] recently updated.
  1388. To fix this you will want to set \f[B]func.getattr=newest\f[].
  1389. Remember though that this is just \f[B]stat\f[].
  1390. If the file is later \f[B]open\f[]\[aq]ed or \f[B]unlink\f[]\[aq]ed and
  1391. the policy is different for those then a completely different file or
  1392. directory could be acted on.
  1393. .IP \[bu] 2
  1394. Some policies mixed with some functions may result in strange behaviors.
  1395. Not that some of these behaviors and race conditions couldn\[aq]t happen
  1396. outside \f[B]mergerfs\f[] but that they are far more likely to occur on
  1397. account of the attempt to merge together multiple sources of data which
  1398. could be out of sync due to the different policies.
  1399. .IP \[bu] 2
  1400. For consistency its generally best to set \f[B]category\f[] wide
  1401. policies rather than individual \f[B]func\f[]\[aq]s.
  1402. This will help limit the confusion of tools such as
  1403. rsync (http://linux.die.net/man/1/rsync).
  1404. However, the flexibility is there if needed.
  1405. .SH KNOWN ISSUES / BUGS
  1406. .SS directory mtime is not being updated
  1407. .PP
  1408. Remember that the default policy for \f[C]getattr\f[] is \f[C]ff\f[].
  1409. The information for the first directory found will be returned.
  1410. If it wasn\[aq]t the directory which had been updated then it will
  1411. appear outdated.
  1412. .PP
  1413. The reason this is the default is because any other policy would be more
  1414. expensive and for many applications it is unnecessary.
  1415. To always return the directory with the most recent mtime or a faked
  1416. value based on all found would require a scan of all drives.
  1417. .PP
  1418. If you always want the directory information from the one with the most
  1419. recent mtime then use the \f[C]newest\f[] policy for \f[C]getattr\f[].
  1420. .SS \f[C]mv\ /mnt/pool/foo\ /mnt/disk1/foo\f[] removes \f[C]foo\f[]
  1421. .PP
  1422. This is not a bug.
  1423. .PP
  1424. Run in verbose mode to better undertand what\[aq]s happening:
  1425. .IP
  1426. .nf
  1427. \f[C]
  1428. $\ mv\ \-v\ /mnt/pool/foo\ /mnt/disk1/foo
  1429. copied\ \[aq]/mnt/pool/foo\[aq]\ \->\ \[aq]/mnt/disk1/foo\[aq]
  1430. removed\ \[aq]/mnt/pool/foo\[aq]
  1431. $\ ls\ /mnt/pool/foo
  1432. ls:\ cannot\ access\ \[aq]/mnt/pool/foo\[aq]:\ No\ such\ file\ or\ directory
  1433. \f[]
  1434. .fi
  1435. .PP
  1436. \f[C]mv\f[], when working across devices, is copying the source to
  1437. target and then removing the source.
  1438. Since the source \f[B]is\f[] the target in this case, depending on the
  1439. unlink policy, it will remove the just copied file and other files
  1440. across the branches.
  1441. .PP
  1442. If you want to move files to one drive just copy them there and use
  1443. mergerfs.dedup to clean up the old paths or manually remove them from
  1444. the branches directly.
  1445. .SS cached memory appears greater than it should be
  1446. .PP
  1447. Use \f[C]cache.files=off\f[] or \f[C]direct_io=true\f[].
  1448. See the section on page caching.
  1449. .SS NFS clients returning ESTALE / Stale file handle
  1450. .PP
  1451. Be sure to use \f[C]noforget\f[] and \f[C]use_ino\f[] arguments.
  1452. .SS NFS clients don\[aq]t work
  1453. .PP
  1454. Some NFS clients appear to fail when a mergerfs mount is exported.
  1455. Kodi in particular seems to have issues.
  1456. .PP
  1457. Try enabling the \f[C]use_ino\f[] option.
  1458. Some have reported that it fixes the issue.
  1459. .SS rtorrent fails with ENODEV (No such device)
  1460. .PP
  1461. Be sure to set \f[C]cache.files=partial|full|auto\-full\f[] or turn off
  1462. \f[C]direct_io\f[].
  1463. rtorrent and some other applications use
  1464. mmap (http://linux.die.net/man/2/mmap) to read and write to files and
  1465. offer no failback to traditional methods.
  1466. FUSE does not currently support mmap while using \f[C]direct_io\f[].
  1467. There may be a performance penalty on writes with \f[C]direct_io\f[] off
  1468. as well as the problem of double caching but it\[aq]s the only way to
  1469. get such applications to work.
  1470. If the performance loss is too high for other apps you can mount
  1471. mergerfs twice.
  1472. Once with \f[C]direct_io\f[] enabled and one without it.
  1473. Be sure to set \f[C]dropcacheonclose=true\f[] if not using
  1474. \f[C]direct_io\f[].
  1475. .SS rtorrent fails with files >= 4GiB
  1476. .PP
  1477. This is a kernel bug with mmap and FUSE on 32bit platforms.
  1478. A fix should become available for all LTS releases.
  1479. .PP
  1480. https://marc.info/?l=linux\-fsdevel&m=155550785230874&w=2
  1481. .SS Plex doesn\[aq]t work with mergerfs
  1482. .PP
  1483. It does.
  1484. If you\[aq]re trying to put Plex\[aq]s config / metadata on mergerfs you
  1485. have to leave \f[C]direct_io\f[] off because Plex is using sqlite which
  1486. apparently needs mmap.
  1487. mmap doesn\[aq]t work with \f[C]direct_io\f[].
  1488. To fix this place the data elsewhere or disable \f[C]direct_io\f[] (with
  1489. \f[C]dropcacheonclose=true\f[]).
  1490. .PP
  1491. If the issue is that scanning doesn\[aq]t seem to pick up media then be
  1492. sure to set \f[C]func.getattr=newest\f[] as mentioned above.
  1493. .SS mmap performance is really bad
  1494. .PP
  1495. There is a bug (https://lkml.org/lkml/2016/3/16/260) in caching which
  1496. affects overall performance of mmap through FUSE in Linux 4.x kernels.
  1497. It is fixed in 4.4.10 and 4.5.4 (https://lkml.org/lkml/2016/5/11/59).
  1498. .SS When a program tries to move or rename a file it fails
  1499. .PP
  1500. Please read the section above regarding rename & link (#rename--link).
  1501. .PP
  1502. The problem is that many applications do not properly handle
  1503. \f[C]EXDEV\f[] errors which \f[C]rename\f[] and \f[C]link\f[] may return
  1504. even though they are perfectly valid situations which do not indicate
  1505. actual drive or OS errors.
  1506. The error will only be returned by mergerfs if using a path preserving
  1507. policy as described in the policy section above.
  1508. If you do not care about path preservation simply change the mergerfs
  1509. policy to the non\-path preserving version.
  1510. For example: \f[C]\-o\ category.create=mfs\f[]
  1511. .PP
  1512. Ideally the offending software would be fixed and it is recommended that
  1513. if you run into this problem you contact the software\[aq]s author and
  1514. request proper handling of \f[C]EXDEV\f[] errors.
  1515. .SS Samba: Moving files / directories fails
  1516. .PP
  1517. Workaround: Copy the file/directory and then remove the original rather
  1518. than move.
  1519. .PP
  1520. This isn\[aq]t an issue with Samba but some SMB clients.
  1521. GVFS\-fuse v1.20.3 and prior (found in Ubuntu 14.04 among others) failed
  1522. to handle certain error codes correctly.
  1523. Particularly \f[B]STATUS_NOT_SAME_DEVICE\f[] which comes from the
  1524. \f[B]EXDEV\f[] which is returned by \f[B]rename\f[] when the call is
  1525. crossing mount points.
  1526. When a program gets an \f[B]EXDEV\f[] it needs to explicitly take an
  1527. alternate action to accomplish its goal.
  1528. In the case of \f[B]mv\f[] or similar it tries \f[B]rename\f[] and on
  1529. \f[B]EXDEV\f[] falls back to a manual copying of data between the two
  1530. locations and unlinking the source.
  1531. In these older versions of GVFS\-fuse if it received \f[B]EXDEV\f[] it
  1532. would translate that into \f[B]EIO\f[].
  1533. This would cause \f[B]mv\f[] or most any application attempting to move
  1534. files around on that SMB share to fail with a IO error.
  1535. .PP
  1536. GVFS\-fuse v1.22.0 (https://bugzilla.gnome.org/show_bug.cgi?id=734568)
  1537. and above fixed this issue but a large number of systems use the older
  1538. release.
  1539. On Ubuntu the version can be checked by issuing
  1540. \f[C]apt\-cache\ showpkg\ gvfs\-fuse\f[].
  1541. Most distros released in 2015 seem to have the updated release and will
  1542. work fine but older systems may not.
  1543. Upgrading gvfs\-fuse or the distro in general will address the problem.
  1544. .PP
  1545. In Apple\[aq]s MacOSX 10.9 they replaced Samba (client and server) with
  1546. their own product.
  1547. It appears their new client does not handle \f[B]EXDEV\f[] either and
  1548. responds similar to older release of gvfs on Linux.
  1549. .SS Trashing files occasionally fails
  1550. .PP
  1551. This is the same issue as with Samba.
  1552. \f[C]rename\f[] returns \f[C]EXDEV\f[] (in our case that will really
  1553. only happen with path preserving policies like \f[C]epmfs\f[]) and the
  1554. software doesn\[aq]t handle the situtation well.
  1555. This is unfortunately a common failure of software which moves files
  1556. around.
  1557. The standard indicates that an implementation \f[C]MAY\f[] choose to
  1558. support non\-user home directory trashing of files (which is a
  1559. \f[C]MUST\f[]).
  1560. The implementation \f[C]MAY\f[] also support "top directory trashes"
  1561. which many probably do.
  1562. .PP
  1563. To create a \f[C]$topdir/.Trash\f[] directory as defined in the standard
  1564. use the mergerfs\-tools (https://github.com/trapexit/mergerfs-tools)
  1565. tool \f[C]mergerfs.mktrash\f[].
  1566. .SS tar: Directory renamed before its status could be extracted
  1567. .PP
  1568. Make sure to use the \f[C]use_ino\f[] option.
  1569. .SS Supplemental user groups
  1570. .PP
  1571. Due to the overhead of
  1572. getgroups/setgroups (http://linux.die.net/man/2/setgroups) mergerfs
  1573. utilizes a cache.
  1574. This cache is opportunistic and per thread.
  1575. Each thread will query the supplemental groups for a user when that
  1576. particular thread needs to change credentials and will keep that data
  1577. for the lifetime of the thread.
  1578. This means that if a user is added to a group it may not be picked up
  1579. without the restart of mergerfs.
  1580. However, since the high level FUSE API\[aq]s (at least the standard
  1581. version) thread pool dynamically grows and shrinks it\[aq]s possible
  1582. that over time a thread will be killed and later a new thread with no
  1583. cache will start and query the new data.
  1584. .PP
  1585. The gid cache uses fixed storage to simplify the design and be
  1586. compatible with older systems which may not have C++11 compilers.
  1587. There is enough storage for 256 users\[aq] supplemental groups.
  1588. Each user is allowed upto 32 supplemental groups.
  1589. Linux >= 2.6.3 allows upto 65535 groups per user but most other *nixs
  1590. allow far less.
  1591. NFS allowing only 16.
  1592. The system does handle overflow gracefully.
  1593. If the user has more than 32 supplemental groups only the first 32 will
  1594. be used.
  1595. If more than 256 users are using the system when an uncached user is
  1596. found it will evict an existing user\[aq]s cache at random.
  1597. So long as there aren\[aq]t more than 256 active users this should be
  1598. fine.
  1599. If either value is too low for your needs you will have to modify
  1600. \f[C]gidcache.hpp\f[] to increase the values.
  1601. Note that doing so will increase the memory needed by each thread.
  1602. .SS mergerfs or libfuse crashing
  1603. .PP
  1604. \f[B]NOTE:\f[] as of mergerfs 2.22.0 it includes the most recent version
  1605. of libfuse (or requires libfuse\-2.9.7) so any crash should be reported.
  1606. For older releases continue reading...
  1607. .PP
  1608. If suddenly the mergerfs mount point disappears and
  1609. \f[C]Transport\ endpoint\ is\ not\ connected\f[] is returned when
  1610. attempting to perform actions within the mount directory \f[B]and\f[]
  1611. the version of libfuse (use \f[C]mergerfs\ \-v\f[] to find the version)
  1612. is older than \f[C]2.9.4\f[] its likely due to a bug in libfuse.
  1613. Affected versions of libfuse can be found in Debian Wheezy, Ubuntu
  1614. Precise and others.
  1615. .PP
  1616. In order to fix this please install newer versions of libfuse.
  1617. If using a Debian based distro (Debian,Ubuntu,Mint) you can likely just
  1618. install newer versions of
  1619. libfuse (https://packages.debian.org/unstable/libfuse2) and
  1620. fuse (https://packages.debian.org/unstable/fuse) from the repo of a
  1621. newer release.
  1622. .SS mergerfs appears to be crashing or exiting
  1623. .PP
  1624. There seems to be an issue with Linux version \f[C]4.9.0\f[] and above
  1625. in which an invalid message appears to be transmitted to libfuse (used
  1626. by mergerfs) causing it to exit.
  1627. No messages will be printed in any logs as its not a proper crash.
  1628. Debugging of the issue is still ongoing and can be followed via the
  1629. fuse\-devel
  1630. thread (https://sourceforge.net/p/fuse/mailman/message/35662577).
  1631. .SS mergerfs under heavy load and memory preasure leads to kernel panic
  1632. .PP
  1633. https://lkml.org/lkml/2016/9/14/527
  1634. .IP
  1635. .nf
  1636. \f[C]
  1637. [25192.515454]\ kernel\ BUG\ at\ /build/linux\-a2WvEb/linux\-4.4.0/mm/workingset.c:346!
  1638. [25192.517521]\ invalid\ opcode:\ 0000\ [#1]\ SMP
  1639. [25192.519602]\ Modules\ linked\ in:\ netconsole\ ip6t_REJECT\ nf_reject_ipv6\ ipt_REJECT\ nf_reject_ipv4\ configfs\ binfmt_misc\ veth\ bridge\ stp\ llc\ nf_conntrack_ipv6\ nf_defrag_ipv6\ xt_conntrack\ ip6table_filter\ ip6_tables\ xt_multiport\ iptable_filter\ ipt_MASQUERADE\ nf_nat_masquerade_ipv4\ xt_comment\ xt_nat\ iptable_nat\ nf_conntrack_ipv4\ nf_defrag_ipv4\ nf_nat_ipv4\ nf_nat\ nf_conntrack\ xt_CHECKSUM\ xt_tcpudp\ iptable_mangle\ ip_tables\ x_tables\ intel_rapl\ x86_pkg_temp_thermal\ intel_powerclamp\ eeepc_wmi\ asus_wmi\ coretemp\ sparse_keymap\ kvm_intel\ ppdev\ kvm\ irqbypass\ mei_me\ 8250_fintek\ input_leds\ serio_raw\ parport_pc\ tpm_infineon\ mei\ shpchp\ mac_hid\ parport\ lpc_ich\ autofs4\ drbg\ ansi_cprng\ dm_crypt\ algif_skcipher\ af_alg\ btrfs\ raid456\ async_raid6_recov\ async_memcpy\ async_pq\ async_xor\ async_tx\ xor\ raid6_pq\ libcrc32c\ raid0\ multipath\ linear\ raid10\ raid1\ i915\ crct10dif_pclmul\ crc32_pclmul\ aesni_intel\ i2c_algo_bit\ aes_x86_64\ drm_kms_helper\ lrw\ gf128mul\ glue_helper\ ablk_helper\ syscopyarea\ cryptd\ sysfillrect\ sysimgblt\ fb_sys_fops\ drm\ ahci\ r8169\ libahci\ mii\ wmi\ fjes\ video\ [last\ unloaded:\ netconsole]
  1640. [25192.540910]\ CPU:\ 2\ PID:\ 63\ Comm:\ kswapd0\ Not\ tainted\ 4.4.0\-36\-generic\ #55\-Ubuntu
  1641. [25192.543411]\ Hardware\ name:\ System\ manufacturer\ System\ Product\ Name/P8H67\-M\ PRO,\ BIOS\ 3904\ 04/27/2013
  1642. [25192.545840]\ task:\ ffff88040cae6040\ ti:\ ffff880407488000\ task.ti:\ ffff880407488000
  1643. [25192.548277]\ RIP:\ 0010:[<ffffffff811ba501>]\ \ [<ffffffff811ba501>]\ shadow_lru_isolate+0x181/0x190
  1644. [25192.550706]\ RSP:\ 0018:ffff88040748bbe0\ \ EFLAGS:\ 00010002
  1645. [25192.553127]\ RAX:\ 0000000000001c81\ RBX:\ ffff8802f91ee928\ RCX:\ ffff8802f91eeb38
  1646. [25192.555544]\ RDX:\ ffff8802f91ee938\ RSI:\ ffff8802f91ee928\ RDI:\ ffff8804099ba2c0
  1647. [25192.557914]\ RBP:\ ffff88040748bc08\ R08:\ 000000000001a7b6\ R09:\ 000000000000003f
  1648. [25192.560237]\ R10:\ 000000000001a750\ R11:\ 0000000000000000\ R12:\ ffff8804099ba2c0
  1649. [25192.562512]\ R13:\ ffff8803157e9680\ R14:\ ffff8803157e9668\ R15:\ ffff8804099ba2c8
  1650. [25192.564724]\ FS:\ \ 0000000000000000(0000)\ GS:ffff88041f280000(0000)\ knlGS:0000000000000000
  1651. [25192.566990]\ CS:\ \ 0010\ DS:\ 0000\ ES:\ 0000\ CR0:\ 0000000080050033
  1652. [25192.569201]\ CR2:\ 00007ffabb690000\ CR3:\ 0000000001e0a000\ CR4:\ 00000000000406e0
  1653. [25192.571419]\ Stack:
  1654. [25192.573550]\ \ ffff8804099ba2c0\ ffff88039e4f86f0\ ffff8802f91ee928\ ffff8804099ba2c8
  1655. [25192.575695]\ \ ffff88040748bd08\ ffff88040748bc58\ ffffffff811b99bf\ 0000000000000052
  1656. [25192.577814]\ \ 0000000000000000\ ffffffff811ba380\ 000000000000008a\ 0000000000000080
  1657. [25192.579947]\ Call\ Trace:
  1658. [25192.582022]\ \ [<ffffffff811b99bf>]\ __list_lru_walk_one.isra.3+0x8f/0x130
  1659. [25192.584137]\ \ [<ffffffff811ba380>]\ ?\ memcg_drain_all_list_lrus+0x190/0x190
  1660. [25192.586165]\ \ [<ffffffff811b9a83>]\ list_lru_walk_one+0x23/0x30
  1661. [25192.588145]\ \ [<ffffffff811ba544>]\ scan_shadow_nodes+0x34/0x50
  1662. [25192.590074]\ \ [<ffffffff811a0e9d>]\ shrink_slab.part.40+0x1ed/0x3d0
  1663. [25192.591985]\ \ [<ffffffff811a53da>]\ shrink_zone+0x2ca/0x2e0
  1664. [25192.593863]\ \ [<ffffffff811a64ce>]\ kswapd+0x51e/0x990
  1665. [25192.595737]\ \ [<ffffffff811a5fb0>]\ ?\ mem_cgroup_shrink_node_zone+0x1c0/0x1c0
  1666. [25192.597613]\ \ [<ffffffff810a0808>]\ kthread+0xd8/0xf0
  1667. [25192.599495]\ \ [<ffffffff810a0730>]\ ?\ kthread_create_on_node+0x1e0/0x1e0
  1668. [25192.601335]\ \ [<ffffffff8182e34f>]\ ret_from_fork+0x3f/0x70
  1669. [25192.603193]\ \ [<ffffffff810a0730>]\ ?\ kthread_create_on_node+0x1e0/0x1e0
  1670. \f[]
  1671. .fi
  1672. .PP
  1673. There is a bug in the kernel.
  1674. A work around appears to be turning off \f[C]splice\f[].
  1675. Don\[aq]t add the \f[C]splice_*\f[] arguments or add
  1676. \f[C]no_splice_write,no_splice_move,no_splice_read\f[].
  1677. This, however, is not guaranteed to work.
  1678. .SS rm: fts_read failed: No such file or directory
  1679. .PP
  1680. NOTE: This is only relevant to mergerfs versions at or below v2.25.x and
  1681. should not occur in more recent versions.
  1682. See the notes on \f[C]unlink\f[].
  1683. .PP
  1684. Not \f[I]really\f[] a bug.
  1685. The FUSE library will move files when asked to delete them as a way to
  1686. deal with certain edge cases and then later delete that file when its
  1687. clear the file is no longer needed.
  1688. This however can lead to two issues.
  1689. One is that these hidden files are noticed by \f[C]rm\ \-rf\f[] or
  1690. \f[C]find\f[] when scanning directories and they may try to remove them
  1691. and they might have disappeared already.
  1692. There is nothing \f[I]wrong\f[] about this happening but it can be
  1693. annoying.
  1694. The second issue is that a directory might not be able to removed on
  1695. account of the hidden file being still there.
  1696. .PP
  1697. Using the \f[B]hard_remove\f[] option will make it so these temporary
  1698. files are not used and files are deleted immedately.
  1699. That has a side effect however.
  1700. Files which are unlinked and then they are still used (in certain forms)
  1701. will result in an error (ENOENT).
  1702. .SH FAQ
  1703. .SS How well does mergerfs scale? Is it "production ready?"
  1704. .PP
  1705. Users have reported running mergerfs on everything from a Raspberry Pi
  1706. to dual socket Xeon systems with >20 cores.
  1707. I\[aq]m aware of at least a few companies which use mergerfs in
  1708. production.
  1709. Open Media Vault (https://www.openmediavault.org) includes mergerfs as
  1710. its sole solution for pooling drives.
  1711. .SS Can mergerfs be used with drives which already have data / are in
  1712. use?
  1713. .PP
  1714. Yes.
  1715. MergerFS is a proxy and does \f[B]NOT\f[] interfere with the normal form
  1716. or function of the drives / mounts / paths it manages.
  1717. .PP
  1718. MergerFS is \f[B]not\f[] a traditional filesystem.
  1719. MergerFS is \f[B]not\f[] RAID.
  1720. It does \f[B]not\f[] manipulate the data that passes through it.
  1721. It does \f[B]not\f[] shard data across drives.
  1722. It merely shards some \f[B]behavior\f[] and aggregates others.
  1723. .SS Can mergerfs be removed without affecting the data?
  1724. .PP
  1725. See the previous question\[aq]s answer.
  1726. .SS Do hard links work?
  1727. .PP
  1728. Yes.
  1729. You need to use \f[C]use_ino\f[] to support proper reporting of inodes.
  1730. .PP
  1731. What mergerfs does not do is fake hard links across branches.
  1732. Read the section "rename & link" for how it works.
  1733. .SS Does mergerfs support CoW / copy\-on\-write?
  1734. .PP
  1735. Not in the sense of a filesystem like BTRFS or ZFS nor in the overlayfs
  1736. or aufs sense.
  1737. It does offer a
  1738. cow\-shell (http://manpages.ubuntu.com/manpages/bionic/man1/cow-shell.1.html)
  1739. like hard link breaking (copy to temp file then rename over original)
  1740. which can be useful when wanting to save space by hardlinking duplicate
  1741. files but wish to treat each name as if it were a unique and separate
  1742. file.
  1743. .SS Why can\[aq]t I see my files / directories?
  1744. .PP
  1745. It\[aq]s almost always a permissions issue.
  1746. Unlike mhddfs, which runs as root and attempts to access content as
  1747. such, mergerfs always changes its credentials to that of the caller.
  1748. This means that if the user does not have access to a file or directory
  1749. than neither will mergerfs.
  1750. However, because mergerfs is creating a union of paths it may be able to
  1751. read some files and directories on one drive but not another resulting
  1752. in an incomplete set.
  1753. .PP
  1754. Whenever you run into a split permission issue (seeing some but not all
  1755. files) try using
  1756. mergerfs.fsck (https://github.com/trapexit/mergerfs-tools) tool to check
  1757. for and fix the mismatch.
  1758. If you aren\[aq]t seeing anything at all be sure that the basic
  1759. permissions are correct.
  1760. The user and group values are correct and that directories have their
  1761. executable bit set.
  1762. A common mistake by users new to Linux is to \f[C]chmod\ \-R\ 644\f[]
  1763. when they should have \f[C]chmod\ \-R\ u=rwX,go=rX\f[].
  1764. .PP
  1765. If using a network filesystem such as NFS, SMB, CIFS (Samba) be sure to
  1766. pay close attention to anything regarding permissioning and users.
  1767. Root squashing and user translation for instance has bitten a few
  1768. mergerfs users.
  1769. Some of these also affect the use of mergerfs from container platforms
  1770. such as Docker.
  1771. .SS Why is only one drive being used?
  1772. .PP
  1773. Are you using a path preserving policy?
  1774. The default policy for file creation is \f[C]epmfs\f[].
  1775. That means only the drives with the path preexisting will be considered
  1776. when creating a file.
  1777. If you don\[aq]t care about where files and directories are created you
  1778. likely shouldn\[aq]t be using a path preserving policy and instead
  1779. something like \f[C]mfs\f[].
  1780. .PP
  1781. This can be especially apparent when filling an empty pool from an
  1782. external source.
  1783. If you do want path preservation you\[aq]ll need to perform the manual
  1784. act of creating paths on the drives you want the data to land on before
  1785. transfering your data.
  1786. Setting \f[C]func.mkdir=epall\f[] can simplify managing path
  1787. perservation for \f[C]create\f[].
  1788. .SS Why was libfuse embedded into mergerfs?
  1789. .IP "1." 3
  1790. A significant number of users use mergerfs on distros with old versions
  1791. of libfuse which have serious bugs.
  1792. Requiring updated versions of libfuse on those distros isn\[aq]t
  1793. pratical (no package offered, user inexperience, etc.).
  1794. The only practical way to provide a stable runtime on those systems was
  1795. to "vendor" / embed the library into the project.
  1796. .IP "2." 3
  1797. mergerfs was written to use the high level API.
  1798. There are a number of limitations in the HLAPI that make certain
  1799. features difficult or impossible to implement.
  1800. While some of these features could be patched into newer versions of
  1801. libfuse without breaking the public API some of them would require hacky
  1802. code to provide backwards compatibility.
  1803. While it may still be worth working with upstream to address these
  1804. issues in future versions, since the library needs to be vendored for
  1805. stability and compatibility reasons it is preferable / easier to modify
  1806. the API.
  1807. Longer term the plan is to rewrite mergerfs to use the low level API.
  1808. .SS Why did support for system libfuse get removed?
  1809. .PP
  1810. See above first.
  1811. .PP
  1812. If/when mergerfs is rewritten to use the low\-level API then it\[aq]ll
  1813. be plausible to support system libfuse but till then its simply too much
  1814. work to manage the differences across the versions.
  1815. .SS Why use mergerfs over mhddfs?
  1816. .PP
  1817. mhddfs is no longer maintained and has some known stability and security
  1818. issues (see below).
  1819. MergerFS provides a superset of mhddfs\[aq] features and should offer
  1820. the same or maybe better performance.
  1821. .PP
  1822. Below is an example of mhddfs and mergerfs setup to work similarly.
  1823. .PP
  1824. \f[C]mhddfs\ \-o\ mlimit=4G,allow_other\ /mnt/drive1,/mnt/drive2\ /mnt/pool\f[]
  1825. .PP
  1826. \f[C]mergerfs\ \-o\ minfreespace=4G,allow_other,category.create=ff\ /mnt/drive1:/mnt/drive2\ /mnt/pool\f[]
  1827. .SS Why use mergerfs over aufs?
  1828. .PP
  1829. aufs is mostly abandoned and no longer available in many distros.
  1830. .PP
  1831. While aufs can offer better peak performance mergerfs provides more
  1832. configurability and is generally easier to use.
  1833. mergerfs however does not offer the overlay / copy\-on\-write (CoW)
  1834. features which aufs and overlayfs have.
  1835. .SS Why use mergerfs over unionfs?
  1836. .PP
  1837. UnionFS is more like aufs then mergerfs in that it offers overlay / CoW
  1838. features.
  1839. If you\[aq]re just looking to create a union of drives and want
  1840. flexibility in file/directory placement then mergerfs offers that
  1841. whereas unionfs is more for overlaying RW filesystems over RO ones.
  1842. .SS Why use mergerfs over LVM/ZFS/BTRFS/RAID0 drive concatenation /
  1843. striping?
  1844. .PP
  1845. With simple JBOD / drive concatenation / stripping / RAID0 a single
  1846. drive failure will result in full pool failure.
  1847. mergerfs performs a similar behavior without the possibility of
  1848. catastrophic failure and the difficulties in recovery.
  1849. Drives may fail however all other data will continue to be accessable.
  1850. .PP
  1851. When combined with something like SnapRaid (http://www.snapraid.it)
  1852. and/or an offsite backup solution you can have the flexibilty of JBOD
  1853. without the single point of failure.
  1854. .SS Why use mergerfs over ZFS?
  1855. .PP
  1856. MergerFS is not intended to be a replacement for ZFS.
  1857. MergerFS is intended to provide flexible pooling of arbitrary drives
  1858. (local or remote), of arbitrary sizes, and arbitrary filesystems.
  1859. For \f[C]write\ once,\ read\ many\f[] usecases such as bulk media
  1860. storage.
  1861. Where data integrity and backup is managed in other ways.
  1862. In that situation ZFS can introduce major maintance and cost burdens as
  1863. described
  1864. here (http://louwrentius.com/the-hidden-cost-of-using-zfs-for-your-home-nas.html).
  1865. .SS Can drives be written to directly? Outside of mergerfs while pooled?
  1866. .PP
  1867. Yes, however its not recommended to use the same file from within the
  1868. pool and from without at the same time.
  1869. Especially if using caching of any kind (cache.files, cache.entry,
  1870. cache.attr, cache.negative_entry, cache.symlinks, cache.readdir, etc.).
  1871. .SS Why do I get an "out of space" / "no space left on device" / ENOSPC
  1872. error even though there appears to be lots of space available?
  1873. .PP
  1874. First make sure you\[aq]ve read the sections above about policies, path
  1875. preservation, branch filtering, and the options \f[B]minfreespace\f[],
  1876. \f[B]moveonenospc\f[], \f[B]statfs\f[], and \f[B]statfs_ignore\f[].
  1877. .PP
  1878. mergerfs is simply presenting a union of the content within multiple
  1879. branches.
  1880. The reported free space is an aggregate of space available within the
  1881. pool (behavior modified by \f[B]statfs\f[] and \f[B]statfs_ignore\f[]).
  1882. It does not represent a contiguous space.
  1883. In the same way that read\-only filesystems, those with quotas, or
  1884. reserved space report the full theoretical space available.
  1885. .PP
  1886. Due to path preservation, branch tagging, read\-only status, and
  1887. \f[B]minfreespace\f[] settings it is perfectly valid that
  1888. \f[C]ENOSPC\f[] / "out of space" / "no space left on device" be
  1889. returned.
  1890. It is doing what was asked of it: filtering possible branches due to
  1891. those settings.
  1892. Only one error can be returned and if one of the reasons for filtering a
  1893. branch was \f[B]minfreespace\f[] then it will be returned as such.
  1894. \f[B]moveonenospc\f[] is only relevant to writing a file which is too
  1895. large for the drive its currently on.
  1896. .PP
  1897. It is also possible that the filesystem selected has run out of inodes.
  1898. Use \f[C]df\ \-i\f[] to list the total and available inodes per
  1899. filesystem.
  1900. .PP
  1901. If you don\[aq]t care about path preservation then simply change the
  1902. \f[C]create\f[] policy to one which isn\[aq]t.
  1903. \f[C]mfs\f[] is probably what most are looking for.
  1904. The reason its not default is because it was originally set to
  1905. \f[C]epmfs\f[] and changing it now would change people\[aq]s setup.
  1906. Such a setting change will likely occur in mergerfs 3.
  1907. .SS Can mergerfs mounts be exported over NFS?
  1908. .PP
  1909. Yes.
  1910. Due to current usage of libfuse by mergerfs and how NFS interacts with
  1911. it it is necessary to add \f[C]noforget\f[] to mergerfs options to keep
  1912. from getting "stale file handle" errors.
  1913. .PP
  1914. Some clients (Kodi) have issues in which the contents of the NFS mount
  1915. will not be presented but users have found that enabling the
  1916. \f[C]use_ino\f[] option often fixes that problem.
  1917. .SS Can mergerfs mounts be exported over Samba / SMB?
  1918. .PP
  1919. Yes.
  1920. While some users have reported problems it appears to always be related
  1921. to how Samba is setup in relation to permissions.
  1922. .SS How are inodes calculated?
  1923. .PP
  1924. mergerfs\-inode = (original\-inode | (device\-id << 32))
  1925. .PP
  1926. While \f[C]ino_t\f[] is 64 bits only a few filesystems use more than 32.
  1927. Similarly, while \f[C]dev_t\f[] is also 64 bits it was traditionally 16
  1928. bits.
  1929. Bitwise or\[aq]ing them together should work most of the time.
  1930. While totally unique inodes are preferred the overhead which would be
  1931. needed does not seem to outweighted by the benefits.
  1932. .PP
  1933. While atypical, yes, inodes can be reused and not refer to the same
  1934. file.
  1935. The internal id used to reference a file in FUSE is different from the
  1936. inode value presented.
  1937. The former is the \f[C]nodeid\f[] and is actually a tuple of
  1938. (nodeid,generation).
  1939. That tuple is not user facing.
  1940. The inode is merely metadata passed through the kernel and found using
  1941. the \f[C]stat\f[] family of calls or \f[C]readdir\f[].
  1942. .PP
  1943. From FUSE docs regarding \f[C]use_ino\f[]:
  1944. .IP
  1945. .nf
  1946. \f[C]
  1947. Honor\ the\ st_ino\ field\ in\ the\ functions\ getattr()\ and
  1948. fill_dir().\ This\ value\ is\ used\ to\ fill\ in\ the\ st_ino\ field
  1949. in\ the\ stat(2),\ lstat(2),\ fstat(2)\ functions\ and\ the\ d_ino
  1950. field\ in\ the\ readdir(2)\ function.\ The\ filesystem\ does\ not
  1951. have\ to\ guarantee\ uniqueness,\ however\ some\ applications
  1952. rely\ on\ this\ value\ being\ unique\ for\ the\ whole\ filesystem.
  1953. Note\ that\ this\ does\ *not*\ affect\ the\ inode\ that\ libfuse
  1954. and\ the\ kernel\ use\ internally\ (also\ called\ the\ "nodeid").
  1955. \f[]
  1956. .fi
  1957. .SS I notice massive slowdowns of writes over NFS
  1958. .PP
  1959. Due to how NFS works and interacts with FUSE when not using
  1960. \f[C]cache.files=off\f[] or \f[C]direct_io\f[] its possible that a
  1961. getxattr for \f[C]security.capability\f[] will be issued prior to any
  1962. write.
  1963. This will usually result in a massive slowdown for writes.
  1964. Using \f[C]cache.files=off\f[] or \f[C]direct_io\f[] will keep this from
  1965. happening (and generally good to enable unless you need the features it
  1966. disables) but the \f[C]security_capability\f[] option can also help by
  1967. short circuiting the call and returning \f[C]ENOATTR\f[].
  1968. .PP
  1969. You could also set \f[C]xattr\f[] to \f[C]noattr\f[] or \f[C]nosys\f[]
  1970. to short circuit or stop all xattr requests.
  1971. .SS What are these .fuse_hidden files?
  1972. .PP
  1973. NOTE: mergerfs >= 2.26.0 will not have these temporary files.
  1974. See the notes on \f[C]unlink\f[].
  1975. .PP
  1976. When not using \f[B]hard_remove\f[] libfuse will create
  1977. \&.fuse_hiddenXXXXXXXX files when an opened file is unlinked.
  1978. This is to simplify "use after unlink" usecases.
  1979. There is a possibility these files end up being picked up by software
  1980. scanning directories and not ignoring hidden files.
  1981. This is rarely a problem but a solution is in the works.
  1982. .PP
  1983. The files are cleaned up once the file is finally closed.
  1984. Only if mergerfs crashes or is killed would they be left around.
  1985. They are safe to remove as they are already unlinked files.
  1986. .SS It\[aq]s mentioned that there are some security issues with mhddfs.
  1987. What are they? How does mergerfs address them?
  1988. .PP
  1989. mhddfs (https://github.com/trapexit/mhddfs) manages running as
  1990. \f[B]root\f[] by calling
  1991. getuid() (https://github.com/trapexit/mhddfs/blob/cae96e6251dd91e2bdc24800b4a18a74044f6672/src/main.c#L319)
  1992. and if it returns \f[B]0\f[] then it will
  1993. chown (http://linux.die.net/man/1/chown) the file.
  1994. Not only is that a race condition but it doesn\[aq]t handle other
  1995. situations.
  1996. Rather than attempting to simulate POSIX ACL behavior the proper way to
  1997. manage this is to use seteuid (http://linux.die.net/man/2/seteuid) and
  1998. setegid (http://linux.die.net/man/2/setegid), in effect becoming the
  1999. user making the original call, and perform the action as them.
  2000. This is what mergerfs does and why mergerfs should always run as root.
  2001. .PP
  2002. In Linux setreuid syscalls apply only to the thread.
  2003. GLIBC hides this away by using realtime signals to inform all threads to
  2004. change credentials.
  2005. Taking after \f[B]Samba\f[], mergerfs uses
  2006. \f[B]syscall(SYS_setreuid,...)\f[] to set the callers credentials for
  2007. that thread only.
  2008. Jumping back to \f[B]root\f[] as necessary should escalated privileges
  2009. be needed (for instance: to clone paths between drives).
  2010. .PP
  2011. For non\-Linux systems mergerfs uses a read\-write lock and changes
  2012. credentials only when necessary.
  2013. If multiple threads are to be user X then only the first one will need
  2014. to change the processes credentials.
  2015. So long as the other threads need to be user X they will take a readlock
  2016. allowing multiple threads to share the credentials.
  2017. Once a request comes in to run as user Y that thread will attempt a
  2018. write lock and change to Y\[aq]s credentials when it can.
  2019. If the ability to give writers priority is supported then that flag will
  2020. be used so threads trying to change credentials don\[aq]t starve.
  2021. This isn\[aq]t the best solution but should work reasonably well
  2022. assuming there are few users.
  2023. .SH PERFORMANCE TWEAKING
  2024. .PP
  2025. NOTE: be sure to read about these features before changing them
  2026. .IP \[bu] 2
  2027. enable (or disable) \f[C]splice_move\f[], \f[C]splice_read\f[], and
  2028. \f[C]splice_write\f[]
  2029. .IP \[bu] 2
  2030. increase cache timeouts \f[C]cache.attr\f[], \f[C]cache.entry\f[],
  2031. \f[C]cache.negative_entry\f[]
  2032. .IP \[bu] 2
  2033. enable (or disable) page caching (\f[C]cache.files\f[])
  2034. .IP \[bu] 2
  2035. enable \f[C]cache.open\f[]
  2036. .IP \[bu] 2
  2037. enable \f[C]cache.statfs\f[]
  2038. .IP \[bu] 2
  2039. enable \f[C]cache.symlinks\f[]
  2040. .IP \[bu] 2
  2041. enable \f[C]cache.readdir\f[]
  2042. .IP \[bu] 2
  2043. change the number of worker threads
  2044. .IP \[bu] 2
  2045. disable \f[C]security_capability\f[] and/or \f[C]xattr\f[]
  2046. .IP \[bu] 2
  2047. disable \f[C]posix_acl\f[]
  2048. .IP \[bu] 2
  2049. disable \f[C]async_read\f[]
  2050. .IP \[bu] 2
  2051. test theoretical performance using \f[C]nullrw\f[] or mounting a ram
  2052. disk
  2053. .IP \[bu] 2
  2054. use \f[C]symlinkify\f[] if your data is largely static
  2055. .IP \[bu] 2
  2056. use tiered cache drives
  2057. .IP \[bu] 2
  2058. use lvm and lvm cache to place a SSD in front of your HDDs (howto
  2059. coming)
  2060. .SH SUPPORT
  2061. .PP
  2062. Filesystems are very complex and difficult to debug.
  2063. mergerfs, while being just a proxy of sorts, is also very difficult to
  2064. debug given the large number of possible settings it can have itself and
  2065. the massive number of environments it can run in.
  2066. When reporting on a suspected issue \f[B]please, please\f[] include as
  2067. much of the below information as possible otherwise it will be difficult
  2068. or impossible to diagnose.
  2069. Also please make sure to read all of the above documentation as it
  2070. includes nearly every known system or user issue previously encountered.
  2071. .SS Information to include in bug reports
  2072. .IP \[bu] 2
  2073. Version of mergerfs: \f[C]mergerfs\ \-V\f[]
  2074. .IP \[bu] 2
  2075. mergerfs settings: from \f[C]/etc/fstab\f[] or command line execution
  2076. .IP \[bu] 2
  2077. Version of Linux: \f[C]uname\ \-a\f[]
  2078. .IP \[bu] 2
  2079. Versions of any additional software being used
  2080. .IP \[bu] 2
  2081. List of drives, their filesystems, and sizes (before and after issue):
  2082. \f[C]df\ \-h\f[]
  2083. .IP \[bu] 2
  2084. A \f[C]strace\f[] of the app having problems:
  2085. .IP \[bu] 2
  2086. \f[C]strace\ \-f\ \-o\ /tmp/app.strace.txt\ <cmd>\f[]
  2087. .IP \[bu] 2
  2088. A \f[C]strace\f[] of mergerfs while the program is trying to do whatever
  2089. it\[aq]s failing to do:
  2090. .IP \[bu] 2
  2091. \f[C]strace\ \-f\ \-p\ <mergerfsPID>\ \-o\ /tmp/mergerfs.strace.txt\f[]
  2092. .IP \[bu] 2
  2093. \f[B]Precise\f[] directions on replicating the issue.
  2094. Do not leave \f[B]anything\f[] out.
  2095. .IP \[bu] 2
  2096. Try to recreate the problem in the simplist way using standard programs.
  2097. .SS Contact / Issue submission
  2098. .IP \[bu] 2
  2099. github.com: https://github.com/trapexit/mergerfs/issues
  2100. .IP \[bu] 2
  2101. email: trapexit\@spawn.link
  2102. .IP \[bu] 2
  2103. twitter: https://twitter.com/_trapexit
  2104. .IP \[bu] 2
  2105. reddit: https://www.reddit.com/user/trapexit
  2106. .IP \[bu] 2
  2107. discord: https://discord.gg/MpAr69V
  2108. .SS Support development
  2109. .PP
  2110. This software is free to use and released under a very liberal license.
  2111. That said if you like this software and would like to support its
  2112. development donations are welcome.
  2113. .IP \[bu] 2
  2114. PayPal: https://paypal.me/trapexit
  2115. .IP \[bu] 2
  2116. Patreon: https://www.patreon.com/trapexit
  2117. .IP \[bu] 2
  2118. SubscribeStar: https://www.subscribestar.com/trapexit
  2119. .IP \[bu] 2
  2120. Bitcoin (BTC): 12CdMhEPQVmjz3SSynkAEuD5q9JmhTDCZA
  2121. .IP \[bu] 2
  2122. Bitcoin Cash (BCH): 1AjPqZZhu7GVEs6JFPjHmtsvmDL4euzMzp
  2123. .IP \[bu] 2
  2124. Ethereum (ETH): 0x09A166B11fCC127324C7fc5f1B572255b3046E94
  2125. .IP \[bu] 2
  2126. Litecoin (LTC): LXAsq6yc6zYU3EbcqyWtHBrH1Ypx4GjUjm
  2127. .SH LINKS
  2128. .IP \[bu] 2
  2129. https://spawn.link
  2130. .IP \[bu] 2
  2131. https://github.com/trapexit/mergerfs
  2132. .IP \[bu] 2
  2133. https://github.com/trapexit/mergerfs\-tools
  2134. .IP \[bu] 2
  2135. https://github.com/trapexit/scorch
  2136. .IP \[bu] 2
  2137. https://github.com/trapexit/bbf
  2138. .IP \[bu] 2
  2139. https://github.com/trapexit/backup\-and\-recovery\-howtos
  2140. .SH AUTHORS
  2141. Antonio SJ Musumeci <trapexit@spawn.link>.