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.

1936 lines
74 KiB

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
8 years ago
9 years ago
8 years ago
8 years ago
8 years ago
8 years ago
8 years ago
8 years ago
8 years ago
8 years ago
8 years ago
7 years ago
8 years ago
8 years ago
8 years ago
8 years ago
9 years ago
9 years ago
9 years ago
9 years ago
8 years ago
9 years ago
8 years ago
9 years ago
8 years ago
9 years ago
8 years ago
9 years ago
8 years ago
9 years ago
8 years ago
9 years ago
8 years ago
8 years ago
8 years ago
8 years ago
8 years ago
8 years ago
8 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
7 years ago
8 years ago
8 years ago
8 years ago
8 years ago
8 years ago
8 years ago
8 years ago
9 years ago
8 years ago
8 years ago
7 years ago
7 years ago
8 years ago
8 years ago
8 years ago
  1. .\"t
  2. .\" Automatically generated by Pandoc 1.19.2.4
  3. .\"
  4. .TH "mergerfs" "1" "2019\-01\-15" "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
  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. Turn read\-only files into symlinks to increase read performance
  42. .SH How it works
  43. .PP
  44. mergerfs logically merges multiple paths together.
  45. Think a union of sets.
  46. The file/s or directory/s acted on or presented through mergerfs are
  47. based on the policy chosen for that particular action.
  48. Read more about policies below.
  49. .IP
  50. .nf
  51. \f[C]
  52. A\ \ \ \ \ \ \ \ \ +\ \ \ \ \ \ B\ \ \ \ \ \ \ \ =\ \ \ \ \ \ \ C
  53. /disk1\ \ \ \ \ \ \ \ \ \ \ /disk2\ \ \ \ \ \ \ \ \ \ \ /merged
  54. |\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ |\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ |
  55. +\-\-\ /dir1\ \ \ \ \ \ \ \ +\-\-\ /dir1\ \ \ \ \ \ \ \ +\-\-\ /dir1
  56. |\ \ \ |\ \ \ \ \ \ \ \ \ \ \ \ |\ \ \ |\ \ \ \ \ \ \ \ \ \ \ \ |\ \ \ |
  57. |\ \ \ +\-\-\ file1\ \ \ \ |\ \ \ +\-\-\ file2\ \ \ \ |\ \ \ +\-\-\ file1
  58. |\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ |\ \ \ +\-\-\ file3\ \ \ \ |\ \ \ +\-\-\ file2
  59. +\-\-\ /dir2\ \ \ \ \ \ \ \ |\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ |\ \ \ +\-\-\ file3
  60. |\ \ \ |\ \ \ \ \ \ \ \ \ \ \ \ +\-\-\ /dir3\ \ \ \ \ \ \ \ |
  61. |\ \ \ +\-\-\ file4\ \ \ \ \ \ \ \ |\ \ \ \ \ \ \ \ \ \ \ \ +\-\-\ /dir2
  62. |\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ +\-\-\ file5\ \ \ |\ \ \ |
  63. +\-\-\ file6\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ |\ \ \ +\-\-\ file4
  64. \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ |
  65. \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ +\-\-\ /dir3
  66. \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ |\ \ \ |
  67. \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ |\ \ \ +\-\-\ file5
  68. \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ |
  69. \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ +\-\-\ file6
  70. \f[]
  71. .fi
  72. .PP
  73. mergerfs does \f[B]not\f[] support the copy\-on\-write (CoW) behavior
  74. found in \f[B]aufs\f[] and \f[B]overlayfs\f[].
  75. You can \f[B]not\f[] mount a read\-only filesystem and write to it.
  76. However, mergerfs will ignore read\-only drives when creating new files
  77. so you can mix rw and ro drives.
  78. .SH OPTIONS
  79. .SS mount options
  80. .IP \[bu] 2
  81. \f[B]defaults\f[]: a shortcut for FUSE\[aq]s \f[B]atomic_o_trunc\f[],
  82. \f[B]auto_cache\f[], \f[B]big_writes\f[], \f[B]default_permissions\f[],
  83. \f[B]splice_move\f[], \f[B]splice_read\f[], and \f[B]splice_write\f[].
  84. These options seem to provide the best performance.
  85. .IP \[bu] 2
  86. \f[B]allow_other\f[]: a libfuse option which allows users besides the
  87. one which ran mergerfs to see the filesystem.
  88. This is required for most use\-cases.
  89. .IP \[bu] 2
  90. \f[B]direct_io\f[]: causes FUSE to bypass caching which can increase
  91. write speeds at the detriment of reads.
  92. Note that not enabling \f[C]direct_io\f[] will cause double caching of
  93. files and therefore less memory for caching generally (enable
  94. \f[B]dropcacheonclose\f[] to help with this problem).
  95. However, \f[C]mmap\f[] does not work when \f[C]direct_io\f[] is enabled.
  96. .IP \[bu] 2
  97. \f[B]minfreespace=value\f[]: the minimum space value used for creation
  98. policies.
  99. Understands \[aq]K\[aq], \[aq]M\[aq], and \[aq]G\[aq] to represent
  100. kilobyte, megabyte, and gigabyte respectively.
  101. (default: 4G)
  102. .IP \[bu] 2
  103. \f[B]moveonenospc=true|false\f[]: when enabled (set to \f[B]true\f[]) if
  104. a \f[B]write\f[] fails with \f[B]ENOSPC\f[] or \f[B]EDQUOT\f[] a scan of
  105. all drives will be done looking for the drive with the most free space
  106. which is at least the size of the file plus the amount which failed to
  107. write.
  108. An attempt to move the file to that drive will occur (keeping all
  109. metadata possible) and if successful the original is unlinked and the
  110. write retried.
  111. (default: false)
  112. .IP \[bu] 2
  113. \f[B]use_ino\f[]: causes mergerfs to supply file/directory inodes rather
  114. than libfuse.
  115. While not a default it is generally recommended it be enabled so that
  116. hard linked files share the same inode value.
  117. .IP \[bu] 2
  118. \f[B]hard_remove\f[]: force libfuse to immedately remove files when
  119. unlinked.
  120. This will keep the \f[C]\&.fuse_hidden\f[] files from showing up but if
  121. software uses an opened but unlinked file in certain ways it could
  122. result in errors.
  123. .IP \[bu] 2
  124. \f[B]dropcacheonclose=true|false\f[]: when a file is requested to be
  125. closed call \f[C]posix_fadvise\f[] on it first to instruct the kernel
  126. that we no longer need the data and it can drop its cache.
  127. Recommended when \f[B]direct_io\f[] is not enabled to limit double
  128. caching.
  129. (default: false)
  130. .IP \[bu] 2
  131. \f[B]symlinkify=true|false\f[]: when enabled (set to \f[B]true\f[]) and
  132. a file is not writable and its mtime or ctime is older than
  133. \f[B]symlinkify_timeout\f[] files will be reported as symlinks to the
  134. original files.
  135. Please read more below before using.
  136. (default: false)
  137. .IP \[bu] 2
  138. \f[B]symlinkify_timeout=value\f[]: time to wait, in seconds, to activate
  139. the \f[B]symlinkify\f[] behavior.
  140. (default: 3600)
  141. .IP \[bu] 2
  142. \f[B]nullrw=true|false\f[]: turns reads and writes into no\-ops.
  143. The request will succeed but do nothing.
  144. Useful for benchmarking mergerfs.
  145. (default: false)
  146. .IP \[bu] 2
  147. \f[B]ignorepponrename=true|false\f[]: ignore path preserving on rename.
  148. Typically rename and link act differently depending on the policy of
  149. \f[C]create\f[] (read below).
  150. Enabling this will cause rename and link to always use the non\-path
  151. preserving behavior.
  152. This means files, when renamed or linked, will stay on the same drive.
  153. (default: false)
  154. .IP \[bu] 2
  155. \f[B]security_capability=true|false\f[]: If false return ENOATTR when
  156. xattr security.capability is queried.
  157. (default: true)
  158. .IP \[bu] 2
  159. \f[B]xattr=passthrough|noattr|nosys\f[]: Runtime control of xattrs.
  160. Default is to passthrough xattr requests.
  161. \[aq]noattr\[aq] will short circuit as if nothing exists.
  162. \[aq]nosys\[aq] will respond with ENOSYS as if xattrs are not supported
  163. or disabled.
  164. (default: passthrough)
  165. .IP \[bu] 2
  166. \f[B]link_cow=true|false\f[]: When enabled if a regular file is opened
  167. which has a link count > 1 it will copy the file to a temporary file and
  168. rename over the original.
  169. Breaking the link and providing a basic copy\-on\-write function similar
  170. to cow\-shell.
  171. (default: false)
  172. .IP \[bu] 2
  173. \f[B]statfs=base|full\f[]: Controls how statfs works.
  174. \[aq]base\[aq] means it will always use all branches in statfs
  175. calculations.
  176. \[aq]full\[aq] is in effect path preserving and only includes drives
  177. where the path exists.
  178. (default: base)
  179. .IP \[bu] 2
  180. \f[B]statfs_ignore=none|ro|nc\f[]: \[aq]ro\[aq] will cause statfs
  181. calculations to ignore available space for branches mounted or tagged as
  182. \[aq]read\-only\[aq] or \[aq]no create\[aq].
  183. \[aq]nc\[aq] will ignore available space for branches tagged as \[aq]no
  184. create\[aq].
  185. (default: none)
  186. .IP \[bu] 2
  187. \f[B]threads=num\f[]: number of threads to use in multithreaded mode.
  188. When set to zero (the default) it will attempt to discover and use the
  189. number of logical cores.
  190. If the lookup fails it will fall back to using 4.
  191. If the thread count is set negative it will look up the number of cores
  192. then divide by the absolute value.
  193. ie.
  194. threads=\-2 on an 8 core machine will result in 8 / 2 = 4 threads.
  195. There will always be at least 1 thread.
  196. NOTE: higher number of threads increases parallelism but usually
  197. decreases throughput.
  198. (default: number of cores) \f[I]NOTE2:\f[] the option is unavailable
  199. when built with system libfuse.
  200. .IP \[bu] 2
  201. \f[B]fsname=name\f[]: sets the name of the filesystem as seen in
  202. \f[B]mount\f[], \f[B]df\f[], etc.
  203. Defaults to a list of the source paths concatenated together with the
  204. longest common prefix removed.
  205. .IP \[bu] 2
  206. \f[B]func.<func>=<policy>\f[]: sets the specific FUSE function\[aq]s
  207. 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
  212. in the provided category.
  213. Example: \f[B]category.create=mfs\f[]
  214. .PP
  215. \f[B]NOTE:\f[] Options are evaluated in the order listed so if the
  216. options are \f[B]func.rmdir=rand,category.action=ff\f[] the
  217. \f[B]action\f[] category setting will override the \f[B]rmdir\f[]
  218. setting.
  219. .SS branches
  220. .PP
  221. The \[aq]branches\[aq] (formerly \[aq]srcmounts\[aq]) argument is a
  222. colon (\[aq]:\[aq]) delimited list of paths to be pooled together.
  223. It does not matter if the paths are on the same or different drives nor
  224. does it matter the filesystem.
  225. Used and available space will not be duplicated for paths on the same
  226. device and any features which aren\[aq]t supported by the underlying
  227. filesystem (such as file attributes or extended attributes) will return
  228. the appropriate errors.
  229. .PP
  230. To make it easier to include multiple branches mergerfs supports
  231. globbing (http://linux.die.net/man/7/glob).
  232. \f[B]The globbing tokens MUST be escaped when using via the shell else
  233. the shell itself will apply the glob itself.\f[]
  234. .PP
  235. Each branch can have a suffix of \f[C]=RW\f[] (read / write),
  236. \f[C]=RO\f[] (read\-only), or \f[C]=NC\f[] (no create).
  237. These suffixes work with globs as well and will apply to each path
  238. found.
  239. \f[C]RW\f[] is the default behavior and those paths will be eligible for
  240. all policy categories.
  241. \f[C]RO\f[] will exclude those paths from \f[C]create\f[] and
  242. \f[C]action\f[] policies (just as a filesystem being mounted \f[C]ro\f[]
  243. would).
  244. \f[C]NC\f[] will exclude those paths from \f[C]create\f[] policies (you
  245. can\[aq]t create but you can change / delete).
  246. .IP
  247. .nf
  248. \f[C]
  249. $\ mergerfs\ \-o\ defaults,allow_other,use_ino\ /mnt/disk\\*:/mnt/cdrom\ /media/drives
  250. \f[]
  251. .fi
  252. .PP
  253. The above line will use all mount points in /mnt prefixed with
  254. \f[B]disk\f[] and the \f[B]cdrom\f[].
  255. .PP
  256. To have the pool mounted at boot or otherwise accessable from related
  257. tools use \f[B]/etc/fstab\f[].
  258. .IP
  259. .nf
  260. \f[C]
  261. #\ <file\ system>\ \ \ \ \ \ \ \ <mount\ point>\ \ <type>\ \ \ \ \ \ \ \ \ <options>\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ <dump>\ \ <pass>
  262. /mnt/disk*:/mnt/cdrom\ \ /media/drives\ \ fuse.mergerfs\ \ defaults,allow_other,use_ino\ \ \ 0\ \ \ \ \ \ \ 0
  263. \f[]
  264. .fi
  265. .PP
  266. \f[B]NOTE:\f[] the globbing is done at mount or xattr update time (see
  267. below).
  268. If a new directory is added matching the glob after the fact it will not
  269. be automatically included.
  270. .PP
  271. \f[B]NOTE:\f[] for mounting via \f[B]fstab\f[] to work you must have
  272. \f[B]mount.fuse\f[] installed.
  273. For Ubuntu/Debian it is included in the \f[B]fuse\f[] package.
  274. .SS symlinkify
  275. .PP
  276. Due to the levels of indirection introduced by mergerfs and the
  277. underlying technology FUSE there can be varying levels of performance
  278. degredation.
  279. This feature will turn non\-directories which are not writable into
  280. symlinks to the original file found by the \f[C]readlink\f[] policy
  281. after the mtime and ctime are older than the timeout.
  282. .PP
  283. \f[B]WARNING:\f[] The current implementation has a known issue in which
  284. if the file is open and being used when the file is converted to a
  285. symlink then the application which has that file open will receive an
  286. error when using it.
  287. This is unlikely to occur in practice but is something to keep in mind.
  288. .PP
  289. \f[B]WARNING:\f[] Some backup solutions, such as CrashPlan, do not
  290. backup the target of a symlink.
  291. If using this feature it will be necessary to point any backup software
  292. to the original drives or configure the software to follow symlinks if
  293. such an option is available.
  294. Alternatively create two mounts.
  295. One for backup and one for general consumption.
  296. .SS nullrw
  297. .PP
  298. Due to how FUSE works there is an overhead to all requests made to a
  299. FUSE filesystem.
  300. Meaning that even a simple passthrough will have some slowdown.
  301. However, generally the overhead is minimal in comparison to the cost of
  302. the underlying I/O.
  303. By disabling the underlying I/O we can test the theoretical performance
  304. boundries.
  305. .PP
  306. By enabling \f[C]nullrw\f[] mergerfs will work as it always does
  307. \f[B]except\f[] that all reads and writes will be no\-ops.
  308. A write will succeed (the size of the write will be returned as if it
  309. were successful) but mergerfs does nothing with the data it was given.
  310. Similarly a read will return the size requested but won\[aq]t touch the
  311. buffer.
  312. .PP
  313. Example:
  314. .IP
  315. .nf
  316. \f[C]
  317. $\ dd\ if=/dev/zero\ of=/path/to/mergerfs/mount/benchmark\ ibs=1M\ obs=512\ count=1024\ conv=fdatasync
  318. 1024+0\ records\ in
  319. 2097152+0\ records\ out
  320. 1073741824\ bytes\ (1.1\ GB,\ 1.0\ GiB)\ copied,\ 15.4067\ s,\ 69.7\ MB/s
  321. $\ dd\ if=/dev/zero\ of=/path/to/mergerfs/mount/benchmark\ ibs=1M\ obs=1M\ count=1024\ conv=fdatasync
  322. 1024+0\ records\ in
  323. 1024+0\ records\ out
  324. 1073741824\ bytes\ (1.1\ GB,\ 1.0\ GiB)\ copied,\ 0.219585\ s,\ 4.9\ GB/s
  325. $\ dd\ if=/path/to/mergerfs/mount/benchmark\ of=/dev/null\ bs=512\ count=102400\ conv=fdatasync
  326. 102400+0\ records\ in
  327. 102400+0\ records\ out
  328. 52428800\ bytes\ (52\ MB,\ 50\ MiB)\ copied,\ 0.757991\ s,\ 69.2\ MB/s
  329. $\ dd\ if=/path/to/mergerfs/mount/benchmark\ of=/dev/null\ bs=1M\ count=1024\ conv=fdatasync
  330. 1024+0\ records\ in
  331. 1024+0\ records\ out
  332. 1073741824\ bytes\ (1.1\ GB,\ 1.0\ GiB)\ copied,\ 0.18405\ s,\ 5.8\ GB/s
  333. \f[]
  334. .fi
  335. .PP
  336. It\[aq]s important to test with different \f[C]obs\f[] (output block
  337. size) values since the relative overhead is greater with smaller values.
  338. As you can see above the size of a read or write can massively impact
  339. theoretical performance.
  340. If an application performs much worse through mergerfs it could very
  341. well be that it doesn\[aq]t optimally size its read and write requests.
  342. In such cases contact the mergerfs author so it can be investigated.
  343. .SS xattr
  344. .PP
  345. Runtime extended attribute support can be managed via the \f[C]xattr\f[]
  346. option.
  347. By default it will passthrough any xattr calls.
  348. Given xattr support is rarely used and can have significant performance
  349. implications mergerfs allows it to be disabled at runtime.
  350. .PP
  351. \f[C]noattr\f[] will cause mergerfs to short circuit all xattr calls and
  352. return ENOATTR where appropriate.
  353. mergerfs still gets all the requests but they will not be forwarded on
  354. to the underlying filesystems.
  355. The runtime control will still function in this mode.
  356. .PP
  357. \f[C]nosys\f[] will cause mergerfs to return ENOSYS for any xattr call.
  358. The difference with \f[C]noattr\f[] is that the kernel will cache this
  359. fact and itself short circuit future calls.
  360. This will be more efficient than \f[C]noattr\f[] but will cause
  361. mergerfs\[aq] runtime control via the hidden file to stop working.
  362. .SH FUNCTIONS / POLICIES / CATEGORIES
  363. .PP
  364. The POSIX filesystem API is made up of a number of functions.
  365. \f[B]creat\f[], \f[B]stat\f[], \f[B]chown\f[], etc.
  366. In mergerfs most of the core functions are grouped into 3 categories:
  367. \f[B]action\f[], \f[B]create\f[], and \f[B]search\f[].
  368. These functions and categories can be assigned a policy which dictates
  369. what file or directory is chosen when performing that behavior.
  370. Any policy can be assigned to a function or category though some may not
  371. be very useful in practice.
  372. For instance: \f[B]rand\f[] (random) may be useful for file creation
  373. (create) but could lead to very odd behavior if used for \f[C]chmod\f[]
  374. if there were more than one copy of the file.
  375. .PP
  376. Some functions, listed in the category \f[C]N/A\f[] below, can not be
  377. assigned the normal policies.
  378. All functions which work on file handles use the handle which was
  379. acquired by \f[C]open\f[] or \f[C]create\f[].
  380. \f[C]readdir\f[] has no real need for a policy given the purpose is
  381. merely to return a list of entries in a directory.
  382. \f[C]statfs\f[]\[aq]s behavior can be modified via other options.
  383. .PP
  384. When using policies which are based on a branch\[aq]s available space
  385. the base path provided is used.
  386. Not the full path to the file in question.
  387. Meaning that sub mounts won\[aq]t be considered in the space
  388. calculations.
  389. The reason is that it doesn\[aq]t really work for non\-path preserving
  390. policies and can lead to non\-obvious behaviors.
  391. .SS Function / Category classifications
  392. .PP
  393. .TS
  394. tab(@);
  395. lw(7.9n) lw(62.1n).
  396. T{
  397. Category
  398. T}@T{
  399. FUSE Functions
  400. T}
  401. _
  402. T{
  403. action
  404. T}@T{
  405. chmod, chown, link, removexattr, rename, rmdir, setxattr, truncate,
  406. unlink, utimens
  407. T}
  408. T{
  409. create
  410. T}@T{
  411. create, mkdir, mknod, symlink
  412. T}
  413. T{
  414. search
  415. T}@T{
  416. access, getattr, getxattr, ioctl, listxattr, open, readlink
  417. T}
  418. T{
  419. N/A
  420. T}@T{
  421. fallocate, fgetattr, fsync, ftruncate, ioctl, read, readdir, release,
  422. statfs, write
  423. T}
  424. .TE
  425. .PP
  426. Due to FUSE limitations \f[B]ioctl\f[] behaves differently if its acting
  427. on a directory.
  428. It\[aq]ll use the \f[B]getattr\f[] policy to find and open the directory
  429. before issuing the \f[B]ioctl\f[].
  430. In other cases where something may be searched (to confirm a directory
  431. exists across all source mounts) \f[B]getattr\f[] will also be used.
  432. .SS Path Preservation
  433. .PP
  434. Policies, as described below, are of two basic types.
  435. \f[C]path\ preserving\f[] and \f[C]non\-path\ preserving\f[].
  436. .PP
  437. All policies which start with \f[C]ep\f[] (\f[B]epff\f[],
  438. \f[B]eplfs\f[], \f[B]eplus\f[], \f[B]epmfs\f[], \f[B]eprand\f[]) are
  439. \f[C]path\ preserving\f[].
  440. \f[C]ep\f[] stands for \f[C]existing\ path\f[].
  441. .PP
  442. A path preserving policy will only consider drives where the relative
  443. path being accessed already exists.
  444. .PP
  445. When using non\-path preserving policies paths will be cloned to target
  446. drives as necessary.
  447. .SS Filters
  448. .PP
  449. Policies basically search branches and create a list of files / paths
  450. for functions to work on.
  451. The policy is responsible for filtering and sorting.
  452. The policy type defines the sorting but filtering is mostly uniform as
  453. described below.
  454. .IP \[bu] 2
  455. No \f[B]search\f[] policies filter.
  456. .IP \[bu] 2
  457. All \f[B]action\f[] policies will filter out branches which are mounted
  458. \f[B]read\-only\f[] or tagged as \f[B]RO (read\-only)\f[].
  459. .IP \[bu] 2
  460. All \f[B]create\f[] policies will filter out branches which are mounted
  461. \f[B]read\-only\f[], tagged \f[B]RO (read\-only)\f[] or \f[B]NC (no
  462. create)\f[], or has available space less than \f[C]minfreespace\f[].
  463. .PP
  464. If all branches are filtered an error will be returned.
  465. Typically \f[B]EROFS\f[] or \f[B]ENOSPC\f[] depending on the reasons.
  466. .SS Policy descriptions
  467. .PP
  468. .TS
  469. tab(@);
  470. lw(16.6n) lw(53.4n).
  471. T{
  472. Policy
  473. T}@T{
  474. Description
  475. T}
  476. _
  477. T{
  478. all
  479. T}@T{
  480. Search category: same as \f[B]epall\f[].
  481. Action category: same as \f[B]epall\f[].
  482. Create category: for \f[B]mkdir\f[], \f[B]mknod\f[], and
  483. \f[B]symlink\f[] it will apply to all branches.
  484. \f[B]create\f[] works like \f[B]ff\f[].
  485. T}
  486. T{
  487. epall (existing path, all)
  488. T}@T{
  489. Search category: same as \f[B]epff\f[] (but more expensive because it
  490. doesn\[aq]t stop after finding a valid branch).
  491. Action category: apply to all found.
  492. Create category: for \f[B]mkdir\f[], \f[B]mknod\f[], and
  493. \f[B]symlink\f[] it will apply to all found.
  494. \f[B]create\f[] works like \f[B]epff\f[] (but more expensive because it
  495. doesn\[aq]t stop after finding a valid branch).
  496. T}
  497. T{
  498. epff (existing path, first found)
  499. T}@T{
  500. Given the order of the branches, as defined at mount time or configured
  501. at runtime, act on the first one found where the relative path exists.
  502. T}
  503. T{
  504. eplfs (existing path, least free space)
  505. T}@T{
  506. Of all the branches on which the relative path exists choose the drive
  507. with the least free space.
  508. T}
  509. T{
  510. eplus (existing path, least used space)
  511. T}@T{
  512. Of all the branches on which the relative path exists choose the drive
  513. with the least used space.
  514. T}
  515. T{
  516. epmfs (existing path, most free space)
  517. T}@T{
  518. Of all the branches on which the relative path exists choose the drive
  519. with the most free space.
  520. T}
  521. T{
  522. eprand (existing path, random)
  523. T}@T{
  524. Calls \f[B]epall\f[] and then randomizes.
  525. T}
  526. T{
  527. erofs
  528. T}@T{
  529. Exclusively return \f[B]\-1\f[] with \f[B]errno\f[] set to
  530. \f[B]EROFS\f[] (read\-only filesystem).
  531. T}
  532. T{
  533. ff (first found)
  534. T}@T{
  535. Search category: same as \f[B]epff\f[].
  536. Action category: same as \f[B]epff\f[].
  537. Create category: Given the order of the drives, as defined at mount time
  538. or configured at runtime, act on the first one found.
  539. T}
  540. T{
  541. lfs (least free space)
  542. T}@T{
  543. Search category: same as \f[B]eplfs\f[].
  544. Action category: same as \f[B]eplfs\f[].
  545. Create category: Pick the drive with the least available free space.
  546. T}
  547. T{
  548. lus (least used space)
  549. T}@T{
  550. Search category: same as \f[B]eplus\f[].
  551. Action category: same as \f[B]eplus\f[].
  552. Create category: Pick the drive with the least used space.
  553. T}
  554. T{
  555. mfs (most free space)
  556. T}@T{
  557. Search category: same as \f[B]epmfs\f[].
  558. Action category: same as \f[B]epmfs\f[].
  559. Create category: Pick the drive with the most available free space.
  560. T}
  561. T{
  562. newest
  563. T}@T{
  564. Pick the file / directory with the largest mtime.
  565. T}
  566. T{
  567. rand (random)
  568. T}@T{
  569. Calls \f[B]all\f[] and then randomizes.
  570. T}
  571. .TE
  572. .SS Defaults
  573. .PP
  574. .TS
  575. tab(@);
  576. l l.
  577. T{
  578. Category
  579. T}@T{
  580. Policy
  581. T}
  582. _
  583. T{
  584. action
  585. T}@T{
  586. epall
  587. T}
  588. T{
  589. create
  590. T}@T{
  591. epmfs
  592. T}
  593. T{
  594. search
  595. T}@T{
  596. ff
  597. T}
  598. .TE
  599. .SS rename & link
  600. .PP
  601. \f[B]NOTE:\f[] If you\[aq]re receiving errors from software when files
  602. are moved / renamed / linked then you should consider changing the
  603. create policy to one which is \f[B]not\f[] path preserving, enabling
  604. \f[C]ignorepponrename\f[], or contacting the author of the offending
  605. software and requesting that \f[C]EXDEV\f[] be properly handled.
  606. .PP
  607. \f[C]rename\f[] and \f[C]link\f[] are tricky functions in a union
  608. filesystem.
  609. \f[C]rename\f[] only works within a single filesystem or device.
  610. If a rename can\[aq]t be done atomically due to the source and
  611. destination paths existing on different mount points it will return
  612. \f[B]\-1\f[] with \f[B]errno = EXDEV\f[] (cross device).
  613. So if a \f[C]rename\f[]\[aq]s source and target are on different drives
  614. within the pool it creates an issue.
  615. .PP
  616. Originally mergerfs would return EXDEV whenever a rename was requested
  617. which was cross directory in any way.
  618. This made the code simple and was technically complient with POSIX
  619. requirements.
  620. However, many applications fail to handle EXDEV at all and treat it as a
  621. normal error or otherwise handle it poorly.
  622. Such apps include: gvfsd\-fuse v1.20.3 and prior, Finder / CIFS/SMB
  623. client in Apple OSX 10.9+, NZBGet, Samba\[aq]s recycling bin feature.
  624. .PP
  625. As a result a compromise was made in order to get most software to work
  626. while still obeying mergerfs\[aq] policies.
  627. Below is the basic logic.
  628. .IP \[bu] 2
  629. If using a \f[B]create\f[] policy which tries to preserve directory
  630. paths (epff,eplfs,eplus,epmfs)
  631. .IP \[bu] 2
  632. Using the \f[B]rename\f[] policy get the list of files to rename
  633. .IP \[bu] 2
  634. For each file attempt rename:
  635. .RS 2
  636. .IP \[bu] 2
  637. If failure with ENOENT run \f[B]create\f[] policy
  638. .IP \[bu] 2
  639. If create policy returns the same drive as currently evaluating then
  640. clone the path
  641. .IP \[bu] 2
  642. Re\-attempt rename
  643. .RE
  644. .IP \[bu] 2
  645. If \f[B]any\f[] of the renames succeed the higher level rename is
  646. considered a success
  647. .IP \[bu] 2
  648. If \f[B]no\f[] renames succeed the first error encountered will be
  649. returned
  650. .IP \[bu] 2
  651. On success:
  652. .RS 2
  653. .IP \[bu] 2
  654. Remove the target from all drives with no source file
  655. .IP \[bu] 2
  656. Remove the source from all drives which failed to rename
  657. .RE
  658. .IP \[bu] 2
  659. If using a \f[B]create\f[] policy which does \f[B]not\f[] try to
  660. preserve directory paths
  661. .IP \[bu] 2
  662. Using the \f[B]rename\f[] policy get the list of files to rename
  663. .IP \[bu] 2
  664. Using the \f[B]getattr\f[] policy get the target path
  665. .IP \[bu] 2
  666. For each file attempt rename:
  667. .RS 2
  668. .IP \[bu] 2
  669. If the source drive != target drive:
  670. .IP \[bu] 2
  671. Clone target path from target drive to source drive
  672. .IP \[bu] 2
  673. Rename
  674. .RE
  675. .IP \[bu] 2
  676. If \f[B]any\f[] of the renames succeed the higher level rename is
  677. considered a success
  678. .IP \[bu] 2
  679. If \f[B]no\f[] renames succeed the first error encountered will be
  680. returned
  681. .IP \[bu] 2
  682. On success:
  683. .RS 2
  684. .IP \[bu] 2
  685. Remove the target from all drives with no source file
  686. .IP \[bu] 2
  687. Remove the source from all drives which failed to rename
  688. .RE
  689. .PP
  690. The the removals are subject to normal entitlement checks.
  691. .PP
  692. The above behavior will help minimize the likelihood of EXDEV being
  693. returned but it will still be possible.
  694. .PP
  695. \f[B]link\f[] uses the same strategy but without the removals.
  696. .SS readdir
  697. .PP
  698. readdir (http://linux.die.net/man/3/readdir) is different from all other
  699. filesystem functions.
  700. While it could have it\[aq]s own set of policies to tweak its behavior
  701. at this time it provides a simple union of files and directories found.
  702. Remember that any action or information queried about these files and
  703. directories come from the respective function.
  704. For instance: an \f[B]ls\f[] is a \f[B]readdir\f[] and for each
  705. file/directory returned \f[B]getattr\f[] is called.
  706. Meaning the policy of \f[B]getattr\f[] is responsible for choosing the
  707. file/directory which is the source of the metadata you see in an
  708. \f[B]ls\f[].
  709. .SS statfs / statvfs
  710. .PP
  711. statvfs (http://linux.die.net/man/2/statvfs) normalizes the source
  712. drives based on the fragment size and sums the number of adjusted blocks
  713. and inodes.
  714. This means you will see the combined space of all sources.
  715. Total, used, and free.
  716. The sources however are dedupped based on the drive so multiple sources
  717. on the same drive will not result in double counting it\[aq]s space.
  718. Filesystems mounted further down the tree of the branch will not be
  719. included when checking the mount\[aq]s stats.
  720. .PP
  721. The options \f[C]statfs\f[] and \f[C]statfs_ignore\f[] can be used to
  722. modify \f[C]statfs\f[] behavior.
  723. .SH BUILDING
  724. .PP
  725. \f[B]NOTE:\f[] Prebuilt packages can be found at:
  726. https://github.com/trapexit/mergerfs/releases
  727. .PP
  728. First get the code from github (https://github.com/trapexit/mergerfs).
  729. .IP
  730. .nf
  731. \f[C]
  732. $\ git\ clone\ https://github.com/trapexit/mergerfs.git
  733. $\ #\ or
  734. $\ wget\ https://github.com/trapexit/mergerfs/releases/download/<ver>/mergerfs\-<ver>.tar.gz
  735. \f[]
  736. .fi
  737. .SS Debian / Ubuntu
  738. .IP
  739. .nf
  740. \f[C]
  741. $\ cd\ mergerfs
  742. $\ sudo\ tools/install\-build\-pkgs
  743. $\ make\ deb
  744. $\ sudo\ dpkg\ \-i\ ../mergerfs_version_arch.deb
  745. \f[]
  746. .fi
  747. .SS Fedora
  748. .IP
  749. .nf
  750. \f[C]
  751. $\ su\ \-
  752. #\ cd\ mergerfs
  753. #\ tools/install\-build\-pkgs
  754. #\ make\ rpm
  755. #\ rpm\ \-i\ rpmbuild/RPMS/<arch>/mergerfs\-<verion>.<arch>.rpm
  756. \f[]
  757. .fi
  758. .SS Generically
  759. .PP
  760. Have git, g++, make, python, automake, libtool installed.
  761. .IP
  762. .nf
  763. \f[C]
  764. $\ cd\ mergerfs
  765. $\ make
  766. $\ sudo\ make\ install
  767. \f[]
  768. .fi
  769. .SS Generically with system libfuse
  770. .PP
  771. \f[B]NOTE:\f[] Configurable threading and thus \f[C]\-o\ threads=num\f[]
  772. option will be unavailable when built with system libfuse.
  773. .PP
  774. Have git, g++, make, python, pkg\-config installed.
  775. Also, install libfuse >= 2.9.7 (but not libfuse\-3.x) and matching
  776. libfuse\-dev (or libfuse\-devel).
  777. .IP
  778. .nf
  779. \f[C]
  780. $\ cd\ mergerfs
  781. $\ make\ INTERNAL_FUSE=0
  782. $\ sudo\ make\ INTERNAL_FUSE=0\ install
  783. \f[]
  784. .fi
  785. .SS Other build options
  786. .IP
  787. .nf
  788. \f[C]
  789. $\ make\ STATIC=1\ #\ builds\ a\ static\ binary
  790. $\ make\ LTO=1\ \ \ \ #\ perform\ link\ time\ optimization
  791. \f[]
  792. .fi
  793. .SH RUNTIME CONFIG
  794. .SS .mergerfs pseudo file
  795. .IP
  796. .nf
  797. \f[C]
  798. <mountpoint>/.mergerfs
  799. \f[]
  800. .fi
  801. .PP
  802. There is a pseudo file available at the mount point which allows for the
  803. runtime modification of certain \f[B]mergerfs\f[] options.
  804. The file will not show up in \f[B]readdir\f[] but can be
  805. \f[B]stat\f[]\[aq]ed and manipulated via
  806. {list,get,set}xattrs (http://linux.die.net/man/2/listxattr) calls.
  807. .PP
  808. Any changes made at runtime are \f[B]not\f[] persisted.
  809. If you wish for values to persist they must be included as options
  810. wherever you configure the mounting of mergerfs (/etc/fstab).
  811. .SS Keys
  812. .PP
  813. Use \f[C]xattr\ \-l\ /mountpoint/.mergerfs\f[] to see all supported
  814. keys.
  815. Some are informational and therefore read\-only.
  816. .SS user.mergerfs.branches
  817. .PP
  818. \f[B]NOTE:\f[] formerly \f[C]user.mergerfs.srcmounts\f[] but said key is
  819. still supported.
  820. .PP
  821. Used to query or modify the list of branches.
  822. When modifying there are several shortcuts to easy manipulation of the
  823. list.
  824. .PP
  825. .TS
  826. tab(@);
  827. l l.
  828. T{
  829. Value
  830. T}@T{
  831. Description
  832. T}
  833. _
  834. T{
  835. [list]
  836. T}@T{
  837. set
  838. T}
  839. T{
  840. +<[list]
  841. T}@T{
  842. prepend
  843. T}
  844. T{
  845. +>[list]
  846. T}@T{
  847. append
  848. T}
  849. T{
  850. \-[list]
  851. T}@T{
  852. remove all values provided
  853. T}
  854. T{
  855. \-<
  856. T}@T{
  857. remove first in list
  858. T}
  859. T{
  860. \->
  861. T}@T{
  862. remove last in list
  863. T}
  864. .TE
  865. .PP
  866. \f[C]xattr\ \-w\ user.mergerfs.branches\ +</mnt/drive3\ /mnt/pool/.mergerfs\f[]
  867. .PP
  868. The \f[C]=NC\f[], \f[C]=RO\f[], \f[C]=RW\f[] syntax works just as on the
  869. command line.
  870. .SS minfreespace
  871. .PP
  872. Input: interger with an optional multiplier suffix.
  873. \f[B]K\f[], \f[B]M\f[], or \f[B]G\f[].
  874. .PP
  875. Output: value in bytes
  876. .SS moveonenospc
  877. .PP
  878. Input: \f[B]true\f[] and \f[B]false\f[]
  879. .PP
  880. Ouput: \f[B]true\f[] or \f[B]false\f[]
  881. .SS categories / funcs
  882. .PP
  883. Input: short policy string as described elsewhere in this document
  884. .PP
  885. Output: the policy string except for categories where its funcs have
  886. multiple types.
  887. In that case it will be a comma separated list
  888. .SS Example
  889. .IP
  890. .nf
  891. \f[C]
  892. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-l\ .mergerfs
  893. user.mergerfs.branches:\ /mnt/a:/mnt/b
  894. user.mergerfs.minfreespace:\ 4294967295
  895. user.mergerfs.moveonenospc:\ false
  896. \&...
  897. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-p\ user.mergerfs.category.search\ .mergerfs
  898. ff
  899. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-w\ user.mergerfs.category.search\ newest\ .mergerfs
  900. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-p\ user.mergerfs.category.search\ .mergerfs
  901. newest
  902. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-w\ user.mergerfs.branches\ +/mnt/c\ .mergerfs
  903. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-p\ user.mergerfs.branches\ .mergerfs
  904. /mnt/a:/mnt/b:/mnt/c
  905. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-w\ user.mergerfs.branches\ =/mnt/c\ .mergerfs
  906. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-p\ user.mergerfs.branches\ .mergerfs
  907. /mnt/c
  908. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-w\ user.mergerfs.branches\ \[aq]+</mnt/a:/mnt/b\[aq]\ .mergerfs
  909. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-p\ user.mergerfs.branches\ .mergerfs
  910. /mnt/a:/mnt/b:/mnt/c
  911. \f[]
  912. .fi
  913. .SS file / directory xattrs
  914. .PP
  915. While they won\[aq]t show up when using
  916. listxattr (http://linux.die.net/man/2/listxattr) \f[B]mergerfs\f[]
  917. offers a number of special xattrs to query information about the files
  918. served.
  919. To access the values you will need to issue a
  920. getxattr (http://linux.die.net/man/2/getxattr) for one of the following:
  921. .IP \[bu] 2
  922. \f[B]user.mergerfs.basepath:\f[] the base mount point for the file given
  923. the current getattr policy
  924. .IP \[bu] 2
  925. \f[B]user.mergerfs.relpath:\f[] the relative path of the file from the
  926. perspective of the mount point
  927. .IP \[bu] 2
  928. \f[B]user.mergerfs.fullpath:\f[] the full path of the original file
  929. given the getattr policy
  930. .IP \[bu] 2
  931. \f[B]user.mergerfs.allpaths:\f[] a NUL (\[aq]\[aq]) separated list of
  932. full paths to all files found
  933. .IP
  934. .nf
  935. \f[C]
  936. [trapexit:/mnt/mergerfs]\ $\ ls
  937. A\ B\ C
  938. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-p\ user.mergerfs.fullpath\ A
  939. /mnt/a/full/path/to/A
  940. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-p\ user.mergerfs.basepath\ A
  941. /mnt/a
  942. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-p\ user.mergerfs.relpath\ A
  943. /full/path/to/A
  944. [trapexit:/mnt/mergerfs]\ $\ xattr\ \-p\ user.mergerfs.allpaths\ A\ |\ tr\ \[aq]\\0\[aq]\ \[aq]\\n\[aq]
  945. /mnt/a/full/path/to/A
  946. /mnt/b/full/path/to/A
  947. \f[]
  948. .fi
  949. .SH TOOLING
  950. .IP \[bu] 2
  951. https://github.com/trapexit/mergerfs\-tools
  952. .IP \[bu] 2
  953. mergerfs.ctl: A tool to make it easier to query and configure mergerfs
  954. at runtime
  955. .IP \[bu] 2
  956. mergerfs.fsck: Provides permissions and ownership auditing and the
  957. ability to fix them
  958. .IP \[bu] 2
  959. mergerfs.dedup: Will help identify and optionally remove duplicate files
  960. .IP \[bu] 2
  961. mergerfs.dup: Ensure there are at least N copies of a file across the
  962. pool
  963. .IP \[bu] 2
  964. mergerfs.balance: Rebalance files across drives by moving them from the
  965. most filled to the least filled
  966. .IP \[bu] 2
  967. mergerfs.mktrash: Creates FreeDesktop.org Trash specification compatible
  968. directories on a mergerfs mount
  969. .IP \[bu] 2
  970. https://github.com/trapexit/scorch
  971. .IP \[bu] 2
  972. scorch: A tool to help discover silent corruption of files and keep
  973. track of files
  974. .IP \[bu] 2
  975. https://github.com/trapexit/bbf
  976. .IP \[bu] 2
  977. bbf (bad block finder): a tool to scan for and \[aq]fix\[aq] hard drive
  978. bad blocks and find the files using those blocks
  979. .SH CACHING
  980. .SS page caching
  981. .PP
  982. The kernel performs caching of data pages on all files not opened with
  983. \f[C]O_DIRECT\f[].
  984. Due to mergerfs using FUSE and therefore being a userland process the
  985. kernel can double cache the content being read through mergerfs.
  986. Once from the underlying filesystem and once for mergerfs.
  987. Using \f[C]direct_io\f[] and/or \f[C]dropcacheonclose\f[] help minimize
  988. the double caching.
  989. \f[C]direct_io\f[] will instruct the kernel to bypass the page cache for
  990. files opened through mergerfs.
  991. \f[C]dropcacheonclose\f[] will cause mergerfs to instruct the kernel to
  992. flush a file\[aq]s page cache for which it had opened when closed.
  993. If most data is read once its probably best to enable both (read above
  994. for details and limitations).
  995. .PP
  996. If a cache is desired for mergerfs do not enable \f[C]direct_io\f[] and
  997. instead possibly use \f[C]auto_cache\f[] or \f[C]kernel_cache\f[].
  998. By default FUSE will invalidate cached pages when a file is opened.
  999. By using \f[C]auto_cache\f[] it will instead use \f[C]getattr\f[] to
  1000. check if a file has changed when the file is opened and if so will flush
  1001. the cache.
  1002. \f[C]ac_attr_timeout\f[] is the timeout for keeping said cache.
  1003. Alternatively \f[C]kernel_cache\f[] will keep the cache across opens
  1004. unless invalidated through other means.
  1005. You should only uses these if you do not plan to write/modify the same
  1006. files through mergerfs and the underlying filesystem at the same time.
  1007. It could lead to corruption.
  1008. Then again doing so without caching can also cause issues.
  1009. .PP
  1010. It\[aq]s a difficult balance between memory usage, cache bloat &
  1011. duplication, and performance.
  1012. Ideally mergerfs would be able to disable caching for the files it
  1013. reads/writes but allow page caching for itself.
  1014. That would limit the FUSE overhead.
  1015. However, there isn\[aq]t good way to achieve this.
  1016. .SS entry & attribute caching
  1017. .PP
  1018. Given the relatively high cost of FUSE due to the kernel <\-> userspace
  1019. round trips there are kernel side caches for file entries and
  1020. attributes.
  1021. The entry cache limits the \f[C]lookup\f[] calls to mergerfs which ask
  1022. if a file exists.
  1023. The attribute cache limits the need to make \f[C]getattr\f[] calls to
  1024. mergerfs which provide file attributes (mode, size, type, etc.).
  1025. As with the page cache these should not be used if the underlying
  1026. filesystems are being manipulated at the same time as it could lead to
  1027. odd behavior or data corruption.
  1028. The options for setting these are \f[C]entry_timeout\f[] and
  1029. \f[C]negative_timeout\f[] for the entry cache and \f[C]attr_timeout\f[]
  1030. for the attributes cache.
  1031. \f[C]negative_timeout\f[] refers to the timeout for negative responses
  1032. to lookups (non\-existant files).
  1033. .SS writeback caching
  1034. .PP
  1035. writeback caching is a technique for improving write speeds by batching
  1036. writes at a faster device and then bulk writing to the slower device.
  1037. With FUSE the kernel will wait for a number of writes to be made and
  1038. then send it to the filesystem as one request.
  1039. mergerfs currently uses a slightly modified and vendored libfuse 2.9.7
  1040. which does not support writeback caching.
  1041. However, a prototype port to libfuse 3.x has been made and the writeback
  1042. cache appears to work as expected (though performance improvements
  1043. greatly depend on the way the client app writes data).
  1044. Once the port is complete and thoroughly tested writeback caching will
  1045. be available.
  1046. .SS tiered caching
  1047. .PP
  1048. Some storage technologies support what some call "tiered" caching.
  1049. The placing of usually smaller, faster storage as a transparent cache to
  1050. larger, slower storage.
  1051. NVMe, SSD, Optane in front of traditional HDDs for instance.
  1052. .PP
  1053. MergerFS does not natively support any sort of tiered caching.
  1054. Most users have no use for such a feature and its inclusion would
  1055. complicate the code.
  1056. However, there are a few situations where a cache drive could help with
  1057. a typical mergerfs setup.
  1058. .IP "1." 3
  1059. Fast network, slow drives, many readers: You\[aq]ve a 10+Gbps network
  1060. with many readers and your regular drives can\[aq]t keep up.
  1061. .IP "2." 3
  1062. Fast network, slow drives, small\[aq]ish bursty writes: You have a
  1063. 10+Gbps network and wish to transfer amounts of data less than your
  1064. cache drive but wish to do so quickly.
  1065. .PP
  1066. With #1 its arguable if you should be using mergerfs at all.
  1067. RAID would probably be the better solution.
  1068. If you\[aq]re going to use mergerfs there are other tactics that may
  1069. help: spreading the data across drives (see the mergerfs.dup tool) and
  1070. setting \f[C]func.open=rand\f[], using \f[C]symlinkify\f[], or using
  1071. dm\-cache or a similar technology to add tiered cache to the underlying
  1072. device.
  1073. .PP
  1074. With #2 one could use dm\-cache as well but there is another solution
  1075. which requires only mergerfs and a cronjob.
  1076. .IP "1." 3
  1077. Create 2 mergerfs pools.
  1078. One which includes just the slow drives and one which has both the fast
  1079. drives (SSD,NVME,etc.) and slow drives.
  1080. .IP "2." 3
  1081. The \[aq]cache\[aq] pool should have the cache drives listed first.
  1082. .IP "3." 3
  1083. The best \f[C]create\f[] policies to use for the \[aq]cache\[aq] pool
  1084. would probably be \f[C]ff\f[], \f[C]epff\f[], \f[C]lfs\f[], or
  1085. \f[C]eplfs\f[].
  1086. The latter two under the assumption that the cache drive(s) are far
  1087. smaller than the backing drives.
  1088. If using path preserving policies remember that you\[aq]ll need to
  1089. manually create the core directories of those paths you wish to be
  1090. cached.
  1091. Be sure the permissions are in sync.
  1092. Use \f[C]mergerfs.fsck\f[] to check / correct them.
  1093. You could also tag the slow drives as \f[C]=NC\f[] though that\[aq]d
  1094. mean if the cache drives fill you\[aq]d get "out of space" errors.
  1095. .IP "4." 3
  1096. Enable \f[C]moveonenospc\f[] and set \f[C]minfreespace\f[]
  1097. appropriately.
  1098. Perhaps setting \f[C]minfreespace\f[] to the size of the largest cache
  1099. drive.
  1100. .IP "5." 3
  1101. Set your programs to use the cache pool.
  1102. .IP "6." 3
  1103. Save one of the below scripts or create you\[aq]re own.
  1104. .IP "7." 3
  1105. Use \f[C]cron\f[] (as root) to schedule the command at whatever
  1106. frequency is appropriate for your workflow.
  1107. .SS time based expiring
  1108. .PP
  1109. Move files from cache to backing pool based only on the last time the
  1110. file was accessed.
  1111. Replace \f[C]\-atime\f[] with \f[C]\-amin\f[] if you want minutes rather
  1112. than days.
  1113. May want to use the \f[C]fadvise\f[] / \f[C]\-\-drop\-cache\f[] version
  1114. of rsync or run rsync with the tool "nocache".
  1115. .IP
  1116. .nf
  1117. \f[C]
  1118. #!/bin/bash
  1119. if\ [\ $#\ !=\ 3\ ];\ then
  1120. \ \ echo\ "usage:\ $0\ <cache\-drive>\ <backing\-pool>\ <days\-old>"
  1121. \ \ exit\ 1
  1122. fi
  1123. CACHE="${1}"
  1124. BACKING="${2}"
  1125. N=${3}
  1126. find\ "${CACHE}"\ \-type\ f\ \-atime\ +${N}\ \-printf\ \[aq]%P\\n\[aq]\ |\ \\
  1127. \ \ rsync\ \-\-files\-from=\-\ \-axqHAXWES\ \-\-preallocate\ \-\-remove\-source\-files\ "${CACHE}/"\ "${BACKING}/"
  1128. \f[]
  1129. .fi
  1130. .SS percentage full expiring
  1131. .PP
  1132. Move the oldest file from the cache to the backing pool.
  1133. Continue till below percentage threshold.
  1134. .IP
  1135. .nf
  1136. \f[C]
  1137. #!/bin/bash
  1138. if\ [\ $#\ !=\ 3\ ];\ then
  1139. \ \ echo\ "usage:\ $0\ <cache\-drive>\ <backing\-pool>\ <percentage>"
  1140. \ \ exit\ 1
  1141. fi
  1142. CACHE="${1}"
  1143. BACKING="${2}"
  1144. PERCENTAGE=${3}
  1145. set\ \-o\ errexit
  1146. while\ [\ $(df\ \-\-output=pcent\ "${CACHE}"\ |\ grep\ \-v\ Use\ |\ cut\ \-d\[aq]%\[aq]\ \-f1)\ \-gt\ ${PERCENTAGE}\ ]
  1147. do
  1148. \ \ \ \ FILE=$(find\ "${CACHE}"\ \-type\ f\ \-printf\ \[aq]%A\@\ %P\\n\[aq]\ |\ \\
  1149. \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ sort\ |\ \\
  1150. \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ head\ \-n\ 1\ |\ \\
  1151. \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ cut\ \-d\[aq]\ \[aq]\ \-f2\-)
  1152. \ \ \ \ test\ \-n\ "${FILE}"
  1153. \ \ \ \ rsync\ \-axqHAXWES\ \-\-preallocate\ \-\-remove\-source\-files\ "${CACHE}/./${FILE}"\ "${BACKING}/"
  1154. done
  1155. \f[]
  1156. .fi
  1157. .SH TIPS / NOTES
  1158. .IP \[bu] 2
  1159. The recommended base options are
  1160. \f[B]defaults,allow_other,direct_io,use_ino\f[].
  1161. (\f[B]use_ino\f[] will only work when used with mergerfs 2.18.0 and
  1162. above.)
  1163. .IP \[bu] 2
  1164. Run mergerfs as \f[C]root\f[] unless you\[aq]re merging paths which are
  1165. owned by the same user otherwise strange permission issues may arise.
  1166. .IP \[bu] 2
  1167. https://github.com/trapexit/backup\-and\-recovery\-howtos : A set of
  1168. guides / howtos on creating a data storage system, backing it up,
  1169. maintaining it, and recovering from failure.
  1170. .IP \[bu] 2
  1171. If you don\[aq]t see some directories and files you expect in a merged
  1172. point or policies seem to skip drives be sure the user has permission to
  1173. all the underlying directories.
  1174. Use \f[C]mergerfs.fsck\f[] to audit the drive for out of sync
  1175. permissions.
  1176. .IP \[bu] 2
  1177. Do \f[B]not\f[] use \f[C]direct_io\f[] if you expect applications (such
  1178. as rtorrent) to mmap (http://linux.die.net/man/2/mmap) files.
  1179. It is not currently supported in FUSE w/ \f[C]direct_io\f[] enabled.
  1180. Enabling \f[C]dropcacheonclose\f[] is recommended when
  1181. \f[C]direct_io\f[] is disabled.
  1182. .IP \[bu] 2
  1183. Since POSIX gives you only error or success on calls its difficult to
  1184. determine the proper behavior when applying the behavior to multiple
  1185. targets.
  1186. \f[B]mergerfs\f[] will return an error only if all attempts of an action
  1187. fail.
  1188. Any success will lead to a success returned.
  1189. This means however that some odd situations may arise.
  1190. .IP \[bu] 2
  1191. Kodi (http://kodi.tv), Plex (http://plex.tv),
  1192. Subsonic (http://subsonic.org), etc.
  1193. can use directory mtime (http://linux.die.net/man/2/stat) to more
  1194. efficiently determine whether to scan for new content rather than simply
  1195. performing a full scan.
  1196. If using the default \f[B]getattr\f[] policy of \f[B]ff\f[] its possible
  1197. those programs will miss an update on account of it returning the first
  1198. directory found\[aq]s \f[B]stat\f[] info and its a later directory on
  1199. another mount which had the \f[B]mtime\f[] recently updated.
  1200. To fix this you will want to set \f[B]func.getattr=newest\f[].
  1201. Remember though that this is just \f[B]stat\f[].
  1202. If the file is later \f[B]open\f[]\[aq]ed or \f[B]unlink\f[]\[aq]ed and
  1203. the policy is different for those then a completely different file or
  1204. directory could be acted on.
  1205. .IP \[bu] 2
  1206. Some policies mixed with some functions may result in strange behaviors.
  1207. Not that some of these behaviors and race conditions couldn\[aq]t happen
  1208. outside \f[B]mergerfs\f[] but that they are far more likely to occur on
  1209. account of the attempt to merge together multiple sources of data which
  1210. could be out of sync due to the different policies.
  1211. .IP \[bu] 2
  1212. For consistency its generally best to set \f[B]category\f[] wide
  1213. policies rather than individual \f[B]func\f[]\[aq]s.
  1214. This will help limit the confusion of tools such as
  1215. rsync (http://linux.die.net/man/1/rsync).
  1216. However, the flexibility is there if needed.
  1217. .SH KNOWN ISSUES / BUGS
  1218. .SS directory mtime is not being updated
  1219. .PP
  1220. Remember that the default policy for \f[C]getattr\f[] is \f[C]ff\f[].
  1221. The information for the first directory found will be returned.
  1222. If it wasn\[aq]t the directory which had been updated then it will
  1223. appear outdated.
  1224. .PP
  1225. The reason this is the default is because any other policy would be far
  1226. more expensive and for many applications it is unnecessary.
  1227. To always return the directory with the most recent mtime or a faked
  1228. value based on all found would require a scan of all drives.
  1229. That alone is far more expensive than \f[C]ff\f[] but would also
  1230. possibly spin up sleeping drives.
  1231. .PP
  1232. If you always want the directory information from the one with the most
  1233. recent mtime then use the \f[C]newest\f[] policy for \f[C]getattr\f[].
  1234. .SS \f[C]mv\ /mnt/pool/foo\ /mnt/disk1/foo\f[] removes \f[C]foo\f[]
  1235. .PP
  1236. This is not a bug.
  1237. .PP
  1238. Run in verbose mode to better undertand what\[aq]s happening:
  1239. .IP
  1240. .nf
  1241. \f[C]
  1242. $\ mv\ \-v\ /mnt/pool/foo\ /mnt/disk1/foo
  1243. copied\ \[aq]/mnt/pool/foo\[aq]\ \->\ \[aq]/mnt/disk1/foo\[aq]
  1244. removed\ \[aq]/mnt/pool/foo\[aq]
  1245. $\ ls\ /mnt/pool/foo
  1246. ls:\ cannot\ access\ \[aq]/mnt/pool/foo\[aq]:\ No\ such\ file\ or\ directory
  1247. \f[]
  1248. .fi
  1249. .PP
  1250. \f[C]mv\f[], when working across devices, is copying the source to
  1251. target and then removing the source.
  1252. Since the source \f[B]is\f[] the target in this case, depending on the
  1253. unlink policy, it will remove the just copied file and other files
  1254. across the branches.
  1255. .PP
  1256. If you want to move files to one drive just copy them there and use
  1257. mergerfs.dedup to clean up the old paths or manually remove them from
  1258. the branches directly.
  1259. .SS cached memory appears greater than it should be
  1260. .PP
  1261. Use the \f[C]direct_io\f[] option as described above.
  1262. Due to what mergerfs is doing there ends up being two caches of a file
  1263. under normal usage.
  1264. One from the underlying filesystem and one from mergerfs.
  1265. Enabling \f[C]direct_io\f[] removes the mergerfs cache.
  1266. This saves on memory but means the kernel needs to communicate with
  1267. mergerfs more often and can therefore result in slower speeds.
  1268. .PP
  1269. Since enabling \f[C]direct_io\f[] disables \f[C]mmap\f[] this is not an
  1270. ideal situation however write speeds should be increased.
  1271. .PP
  1272. If \f[C]direct_io\f[] is disabled it is probably a good idea to enable
  1273. \f[C]dropcacheonclose\f[] to minimize double caching.
  1274. .SS NFS clients don\[aq]t work
  1275. .PP
  1276. Some NFS clients appear to fail when a mergerfs mount is exported.
  1277. Kodi in particular seems to have issues.
  1278. .PP
  1279. Try enabling the \f[C]use_ino\f[] option.
  1280. Some have reported that it fixes the issue.
  1281. .SS rtorrent fails with ENODEV (No such device)
  1282. .PP
  1283. Be sure to turn off \f[C]direct_io\f[].
  1284. rtorrent and some other applications use
  1285. mmap (http://linux.die.net/man/2/mmap) to read and write to files and
  1286. offer no failback to traditional methods.
  1287. FUSE does not currently support mmap while using \f[C]direct_io\f[].
  1288. There may be a performance penalty on writes with \f[C]direct_io\f[] off
  1289. as well as the problem of double caching but it\[aq]s the only way to
  1290. get such applications to work.
  1291. If the performance loss is too high for other apps you can mount
  1292. mergerfs twice.
  1293. Once with \f[C]direct_io\f[] enabled and one without it.
  1294. Be sure to set \f[C]dropcacheonclose=true\f[] if not using
  1295. \f[C]direct_io\f[].
  1296. .SS Plex doesn\[aq]t work with mergerfs
  1297. .PP
  1298. It does.
  1299. If you\[aq]re trying to put Plex\[aq]s config / metadata on mergerfs you
  1300. have to leave \f[C]direct_io\f[] off because Plex is using sqlite which
  1301. apparently needs mmap.
  1302. mmap doesn\[aq]t work with \f[C]direct_io\f[].
  1303. To fix this place the data elsewhere or disable \f[C]direct_io\f[] (with
  1304. \f[C]dropcacheonclose=true\f[]).
  1305. .PP
  1306. If the issue is that scanning doesn\[aq]t seem to pick up media then be
  1307. sure to set \f[C]func.getattr=newest\f[] as mentioned above.
  1308. .SS mmap performance is really bad
  1309. .PP
  1310. There is a bug (https://lkml.org/lkml/2016/3/16/260) in caching which
  1311. affects overall performance of mmap through FUSE in Linux 4.x kernels.
  1312. It is fixed in 4.4.10 and 4.5.4 (https://lkml.org/lkml/2016/5/11/59).
  1313. .SS When a program tries to move or rename a file it fails
  1314. .PP
  1315. Please read the section above regarding rename & link (#rename--link).
  1316. .PP
  1317. The problem is that many applications do not properly handle
  1318. \f[C]EXDEV\f[] errors which \f[C]rename\f[] and \f[C]link\f[] may return
  1319. even though they are perfectly valid situations which do not indicate
  1320. actual drive or OS errors.
  1321. The error will only be returned by mergerfs if using a path preserving
  1322. policy as described in the policy section above.
  1323. If you do not care about path preservation simply change the mergerfs
  1324. policy to the non\-path preserving version.
  1325. For example: \f[C]\-o\ category.create=mfs\f[]
  1326. .PP
  1327. Ideally the offending software would be fixed and it is recommended that
  1328. if you run into this problem you contact the software\[aq]s author and
  1329. request proper handling of \f[C]EXDEV\f[] errors.
  1330. .SS Samba: Moving files / directories fails
  1331. .PP
  1332. Workaround: Copy the file/directory and then remove the original rather
  1333. than move.
  1334. .PP
  1335. This isn\[aq]t an issue with Samba but some SMB clients.
  1336. GVFS\-fuse v1.20.3 and prior (found in Ubuntu 14.04 among others) failed
  1337. to handle certain error codes correctly.
  1338. Particularly \f[B]STATUS_NOT_SAME_DEVICE\f[] which comes from the
  1339. \f[B]EXDEV\f[] which is returned by \f[B]rename\f[] when the call is
  1340. crossing mount points.
  1341. When a program gets an \f[B]EXDEV\f[] it needs to explicitly take an
  1342. alternate action to accomplish it\[aq]s goal.
  1343. In the case of \f[B]mv\f[] or similar it tries \f[B]rename\f[] and on
  1344. \f[B]EXDEV\f[] falls back to a manual copying of data between the two
  1345. locations and unlinking the source.
  1346. In these older versions of GVFS\-fuse if it received \f[B]EXDEV\f[] it
  1347. would translate that into \f[B]EIO\f[].
  1348. This would cause \f[B]mv\f[] or most any application attempting to move
  1349. files around on that SMB share to fail with a IO error.
  1350. .PP
  1351. GVFS\-fuse v1.22.0 (https://bugzilla.gnome.org/show_bug.cgi?id=734568)
  1352. and above fixed this issue but a large number of systems use the older
  1353. release.
  1354. On Ubuntu the version can be checked by issuing
  1355. \f[C]apt\-cache\ showpkg\ gvfs\-fuse\f[].
  1356. Most distros released in 2015 seem to have the updated release and will
  1357. work fine but older systems may not.
  1358. Upgrading gvfs\-fuse or the distro in general will address the problem.
  1359. .PP
  1360. In Apple\[aq]s MacOSX 10.9 they replaced Samba (client and server) with
  1361. their own product.
  1362. It appears their new client does not handle \f[B]EXDEV\f[] either and
  1363. responds similar to older release of gvfs on Linux.
  1364. .SS Trashing files occasionally fails
  1365. .PP
  1366. This is the same issue as with Samba.
  1367. \f[C]rename\f[] returns \f[C]EXDEV\f[] (in our case that will really
  1368. only happen with path preserving policies like \f[C]epmfs\f[]) and the
  1369. software doesn\[aq]t handle the situtation well.
  1370. This is unfortunately a common failure of software which moves files
  1371. around.
  1372. The standard indicates that an implementation \f[C]MAY\f[] choose to
  1373. support non\-user home directory trashing of files (which is a
  1374. \f[C]MUST\f[]).
  1375. The implementation \f[C]MAY\f[] also support "top directory trashes"
  1376. which many probably do.
  1377. .PP
  1378. To create a \f[C]$topdir/.Trash\f[] directory as defined in the standard
  1379. use the mergerfs\-tools (https://github.com/trapexit/mergerfs-tools)
  1380. tool \f[C]mergerfs.mktrash\f[].
  1381. .SS tar: Directory renamed before its status could be extracted
  1382. .PP
  1383. Make sure to use the \f[C]use_ino\f[] option.
  1384. .SS Supplemental user groups
  1385. .PP
  1386. Due to the overhead of
  1387. getgroups/setgroups (http://linux.die.net/man/2/setgroups) mergerfs
  1388. utilizes a cache.
  1389. This cache is opportunistic and per thread.
  1390. Each thread will query the supplemental groups for a user when that
  1391. particular thread needs to change credentials and will keep that data
  1392. for the lifetime of the thread.
  1393. This means that if a user is added to a group it may not be picked up
  1394. without the restart of mergerfs.
  1395. However, since the high level FUSE API\[aq]s (at least the standard
  1396. version) thread pool dynamically grows and shrinks it\[aq]s possible
  1397. that over time a thread will be killed and later a new thread with no
  1398. cache will start and query the new data.
  1399. .PP
  1400. The gid cache uses fixed storage to simplify the design and be
  1401. compatible with older systems which may not have C++11 compilers.
  1402. There is enough storage for 256 users\[aq] supplemental groups.
  1403. Each user is allowed upto 32 supplemental groups.
  1404. Linux >= 2.6.3 allows upto 65535 groups per user but most other *nixs
  1405. allow far less.
  1406. NFS allowing only 16.
  1407. The system does handle overflow gracefully.
  1408. If the user has more than 32 supplemental groups only the first 32 will
  1409. be used.
  1410. If more than 256 users are using the system when an uncached user is
  1411. found it will evict an existing user\[aq]s cache at random.
  1412. So long as there aren\[aq]t more than 256 active users this should be
  1413. fine.
  1414. If either value is too low for your needs you will have to modify
  1415. \f[C]gidcache.hpp\f[] to increase the values.
  1416. Note that doing so will increase the memory needed by each thread.
  1417. .SS mergerfs or libfuse crashing
  1418. .PP
  1419. \f[B]NOTE:\f[] as of mergerfs 2.22.0 it includes the most recent version
  1420. of libfuse (or requires libfuse\-2.9.7) so any crash should be reported.
  1421. For older releases continue reading...
  1422. .PP
  1423. If suddenly the mergerfs mount point disappears and
  1424. \f[C]Transport\ endpoint\ is\ not\ connected\f[] is returned when
  1425. attempting to perform actions within the mount directory \f[B]and\f[]
  1426. the version of libfuse (use \f[C]mergerfs\ \-v\f[] to find the version)
  1427. is older than \f[C]2.9.4\f[] its likely due to a bug in libfuse.
  1428. Affected versions of libfuse can be found in Debian Wheezy, Ubuntu
  1429. Precise and others.
  1430. .PP
  1431. In order to fix this please install newer versions of libfuse.
  1432. If using a Debian based distro (Debian,Ubuntu,Mint) you can likely just
  1433. install newer versions of
  1434. libfuse (https://packages.debian.org/unstable/libfuse2) and
  1435. fuse (https://packages.debian.org/unstable/fuse) from the repo of a
  1436. newer release.
  1437. .SS mergerfs appears to be crashing or exiting
  1438. .PP
  1439. There seems to be an issue with Linux version \f[C]4.9.0\f[] and above
  1440. in which an invalid message appears to be transmitted to libfuse (used
  1441. by mergerfs) causing it to exit.
  1442. No messages will be printed in any logs as its not a proper crash.
  1443. Debugging of the issue is still ongoing and can be followed via the
  1444. fuse\-devel
  1445. thread (https://sourceforge.net/p/fuse/mailman/message/35662577).
  1446. .SS mergerfs under heavy load and memory preasure leads to kernel panic
  1447. .PP
  1448. https://lkml.org/lkml/2016/9/14/527
  1449. .IP
  1450. .nf
  1451. \f[C]
  1452. [25192.515454]\ kernel\ BUG\ at\ /build/linux\-a2WvEb/linux\-4.4.0/mm/workingset.c:346!
  1453. [25192.517521]\ invalid\ opcode:\ 0000\ [#1]\ SMP
  1454. [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]
  1455. [25192.540910]\ CPU:\ 2\ PID:\ 63\ Comm:\ kswapd0\ Not\ tainted\ 4.4.0\-36\-generic\ #55\-Ubuntu
  1456. [25192.543411]\ Hardware\ name:\ System\ manufacturer\ System\ Product\ Name/P8H67\-M\ PRO,\ BIOS\ 3904\ 04/27/2013
  1457. [25192.545840]\ task:\ ffff88040cae6040\ ti:\ ffff880407488000\ task.ti:\ ffff880407488000
  1458. [25192.548277]\ RIP:\ 0010:[<ffffffff811ba501>]\ \ [<ffffffff811ba501>]\ shadow_lru_isolate+0x181/0x190
  1459. [25192.550706]\ RSP:\ 0018:ffff88040748bbe0\ \ EFLAGS:\ 00010002
  1460. [25192.553127]\ RAX:\ 0000000000001c81\ RBX:\ ffff8802f91ee928\ RCX:\ ffff8802f91eeb38
  1461. [25192.555544]\ RDX:\ ffff8802f91ee938\ RSI:\ ffff8802f91ee928\ RDI:\ ffff8804099ba2c0
  1462. [25192.557914]\ RBP:\ ffff88040748bc08\ R08:\ 000000000001a7b6\ R09:\ 000000000000003f
  1463. [25192.560237]\ R10:\ 000000000001a750\ R11:\ 0000000000000000\ R12:\ ffff8804099ba2c0
  1464. [25192.562512]\ R13:\ ffff8803157e9680\ R14:\ ffff8803157e9668\ R15:\ ffff8804099ba2c8
  1465. [25192.564724]\ FS:\ \ 0000000000000000(0000)\ GS:ffff88041f280000(0000)\ knlGS:0000000000000000
  1466. [25192.566990]\ CS:\ \ 0010\ DS:\ 0000\ ES:\ 0000\ CR0:\ 0000000080050033
  1467. [25192.569201]\ CR2:\ 00007ffabb690000\ CR3:\ 0000000001e0a000\ CR4:\ 00000000000406e0
  1468. [25192.571419]\ Stack:
  1469. [25192.573550]\ \ ffff8804099ba2c0\ ffff88039e4f86f0\ ffff8802f91ee928\ ffff8804099ba2c8
  1470. [25192.575695]\ \ ffff88040748bd08\ ffff88040748bc58\ ffffffff811b99bf\ 0000000000000052
  1471. [25192.577814]\ \ 0000000000000000\ ffffffff811ba380\ 000000000000008a\ 0000000000000080
  1472. [25192.579947]\ Call\ Trace:
  1473. [25192.582022]\ \ [<ffffffff811b99bf>]\ __list_lru_walk_one.isra.3+0x8f/0x130
  1474. [25192.584137]\ \ [<ffffffff811ba380>]\ ?\ memcg_drain_all_list_lrus+0x190/0x190
  1475. [25192.586165]\ \ [<ffffffff811b9a83>]\ list_lru_walk_one+0x23/0x30
  1476. [25192.588145]\ \ [<ffffffff811ba544>]\ scan_shadow_nodes+0x34/0x50
  1477. [25192.590074]\ \ [<ffffffff811a0e9d>]\ shrink_slab.part.40+0x1ed/0x3d0
  1478. [25192.591985]\ \ [<ffffffff811a53da>]\ shrink_zone+0x2ca/0x2e0
  1479. [25192.593863]\ \ [<ffffffff811a64ce>]\ kswapd+0x51e/0x990
  1480. [25192.595737]\ \ [<ffffffff811a5fb0>]\ ?\ mem_cgroup_shrink_node_zone+0x1c0/0x1c0
  1481. [25192.597613]\ \ [<ffffffff810a0808>]\ kthread+0xd8/0xf0
  1482. [25192.599495]\ \ [<ffffffff810a0730>]\ ?\ kthread_create_on_node+0x1e0/0x1e0
  1483. [25192.601335]\ \ [<ffffffff8182e34f>]\ ret_from_fork+0x3f/0x70
  1484. [25192.603193]\ \ [<ffffffff810a0730>]\ ?\ kthread_create_on_node+0x1e0/0x1e0
  1485. \f[]
  1486. .fi
  1487. .PP
  1488. There is a bug in the kernel.
  1489. A work around appears to be turning off \f[C]splice\f[].
  1490. Add \f[C]no_splice_write,no_splice_move,no_splice_read\f[] to
  1491. mergerfs\[aq] options.
  1492. Should be placed after \f[C]defaults\f[] if it is used since it will
  1493. turn them on.
  1494. This however is not guaranteed to work.
  1495. .SS rm: fts_read failed: No such file or directory
  1496. .PP
  1497. Not \f[I]really\f[] a bug.
  1498. The FUSE library will move files when asked to delete them as a way to
  1499. deal with certain edge cases and then later delete that file when its
  1500. clear the file is no longer needed.
  1501. This however can lead to two issues.
  1502. One is that these hidden files are noticed by \f[C]rm\ \-rf\f[] or
  1503. \f[C]find\f[] when scanning directories and they may try to remove them
  1504. and they might have disappeared already.
  1505. There is nothing \f[I]wrong\f[] about this happening but it can be
  1506. annoying.
  1507. The second issue is that a directory might not be able to removed on
  1508. account of the hidden file being still there.
  1509. .PP
  1510. Using the \f[B]hard_remove\f[] option will make it so these temporary
  1511. files are not used and files are deleted immedately.
  1512. That has a side effect however.
  1513. Files which are unlinked and then they are still used (in certain forms)
  1514. will result in an error.
  1515. .PP
  1516. A fix is in the works for this.
  1517. .SH FAQ
  1518. .SS How well does mergerfs scale? Is it "production ready?"
  1519. .PP
  1520. Users have reported running mergerfs on everything from a Raspberry Pi
  1521. to dual socket Xeon systems with >20 cores.
  1522. I\[aq]m aware of at least a few companies which use mergerfs in
  1523. production.
  1524. Open Media Vault (https://www.openmediavault.org) includes mergerfs is
  1525. it\[aq]s sole solution for pooling drives.
  1526. .SS Can mergerfs be used with drives which already have data / are in
  1527. use?
  1528. .PP
  1529. Yes.
  1530. MergerFS is a proxy and does \f[B]NOT\f[] interfere with the normal form
  1531. or function of the drives / mounts / paths it manages.
  1532. .PP
  1533. MergerFS is \f[B]not\f[] a traditional filesystem.
  1534. MergerFS is \f[B]not\f[] RAID.
  1535. It does \f[B]not\f[] manipulate the data that passes through it.
  1536. It does \f[B]not\f[] shard data across drives.
  1537. It merely shards some \f[B]behavior\f[] and aggregates others.
  1538. .SS Can mergerfs be removed without affecting the data?
  1539. .PP
  1540. See the previous question\[aq]s answer.
  1541. .SS Do hard links work?
  1542. .PP
  1543. Yes.
  1544. You need to use \f[C]use_ino\f[] to support proper reporting of inodes.
  1545. .PP
  1546. What mergerfs does not do is fake hard links across branches.
  1547. Read the section "rename & link" for how it.
  1548. .SS Does mergerfs support CoW / copy\-on\-write?
  1549. .PP
  1550. Not in the sense of a filesystem like BTRFS or ZFS nor in the overlayfs
  1551. or aufs sense.
  1552. It does offer a
  1553. cow\-shell (http://manpages.ubuntu.com/manpages/bionic/man1/cow-shell.1.html)
  1554. like hard link breaking (copy to temp file then rename over original)
  1555. which can be useful when wanting to save space by hardlinking duplicate
  1556. files but wish to treat each name as if it were a unique and separate
  1557. file.
  1558. .SS Why can\[aq]t I see my files / directories?
  1559. .PP
  1560. It\[aq]s almost always a permissions issue.
  1561. Unlike mhddfs, which runs as root and attempts to access content as
  1562. such, mergerfs always changes it\[aq]s credentials to that of the
  1563. caller.
  1564. This means that if the user does not have access to a file or directory
  1565. than neither will mergerfs.
  1566. However, because mergerfs is creating a union of paths it may be able to
  1567. read some files and directories on one drive but not another resulting
  1568. in an incomplete set.
  1569. .PP
  1570. Whenever you run into a split permission issue (seeing some but not all
  1571. files) try using
  1572. mergerfs.fsck (https://github.com/trapexit/mergerfs-tools) tool to check
  1573. for and fix the mismatch.
  1574. If you aren\[aq]t seeing anything at all be sure that the basic
  1575. permissions are correct.
  1576. The user and group values are correct and that directories have their
  1577. executable bit set.
  1578. A common mistake by users new to Linux is to \f[C]chmod\ \-R\ 644\f[]
  1579. when they should have \f[C]chmod\ \-R\ u=rwX,go=rX\f[].
  1580. .PP
  1581. If using a network filesystem such as NFS, SMB, CIFS (Samba) be sure to
  1582. pay close attention to anything regarding permissioning and users.
  1583. Root squashing and user translation for instance has bitten a few
  1584. mergerfs users.
  1585. Some of these also affect the use of mergerfs from container platforms
  1586. such as Docker.
  1587. .SS Why is only one drive being used?
  1588. .PP
  1589. Are you using a path preserving policy?
  1590. The default policy for file creation is \f[C]epmfs\f[].
  1591. That means only the drives with the path preexisting will be considered
  1592. when creating a file.
  1593. If you don\[aq]t care about where files and directories are created you
  1594. likely shouldn\[aq]t be using a path preserving policy and instead
  1595. something like \f[C]mfs\f[].
  1596. .PP
  1597. This can be especially apparent when filling an empty pool from an
  1598. external source.
  1599. If you do want path preservation you\[aq]ll need to perform the manual
  1600. act of creating paths on the drives you want the data to land on before
  1601. transfering your data.
  1602. Setting \f[C]func.mkdir=epall\f[] can simplify managing path
  1603. perservation for \f[C]create\f[].
  1604. .SS Why was libfuse embedded into mergerfs?
  1605. .PP
  1606. A significant number of users use mergerfs on distros with old versions
  1607. of libfuse which have serious bugs.
  1608. Requiring updated versions of libfuse on those distros isn\[aq]t
  1609. pratical (no package offered, user inexperience, etc.).
  1610. The only practical way to provide a stable runtime on those systems was
  1611. to "vendor" the library into the project.
  1612. .SS Why use mergerfs over mhddfs?
  1613. .PP
  1614. mhddfs is no longer maintained and has some known stability and security
  1615. issues (see below).
  1616. MergerFS provides a superset of mhddfs\[aq] features and should offer
  1617. the same or maybe better performance.
  1618. .PP
  1619. Below is an example of mhddfs and mergerfs setup to work similarly.
  1620. .PP
  1621. \f[C]mhddfs\ \-o\ mlimit=4G,allow_other\ /mnt/drive1,/mnt/drive2\ /mnt/pool\f[]
  1622. .PP
  1623. \f[C]mergerfs\ \-o\ minfreespace=4G,defaults,allow_other,category.create=ff\ /mnt/drive1:/mnt/drive2\ /mnt/pool\f[]
  1624. .SS Why use mergerfs over aufs?
  1625. .PP
  1626. aufs is mostly abandoned and no longer available in many distros.
  1627. .PP
  1628. While aufs can offer better peak performance mergerfs provides more
  1629. configurability and is generally easier to use.
  1630. mergerfs however does not offer the overlay / copy\-on\-write (CoW)
  1631. features which aufs and overlayfs have.
  1632. .SS Why use mergerfs over unionfs?
  1633. .PP
  1634. UnionFS is more like aufs then mergerfs in that it offers overlay / CoW
  1635. features.
  1636. If you\[aq]re just looking to create a union of drives and want
  1637. flexibility in file/directory placement then mergerfs offers that
  1638. whereas unionfs is more for overlaying RW filesystems over RO ones.
  1639. .SS Why use mergerfs over LVM/ZFS/BTRFS/RAID0 drive concatenation /
  1640. striping?
  1641. .PP
  1642. With simple JBOD / drive concatenation / stripping / RAID0 a single
  1643. drive failure will result in full pool failure.
  1644. mergerfs performs a similar behavior without the possibility of
  1645. catastrophic failure and the difficulties in recovery.
  1646. Drives may fail however all other data will continue to be accessable.
  1647. .PP
  1648. When combined with something like SnapRaid (http://www.snapraid.it)
  1649. and/or an offsite backup solution you can have the flexibilty of JBOD
  1650. without the single point of failure.
  1651. .SS Why use mergerfs over ZFS?
  1652. .PP
  1653. MergerFS is not intended to be a replacement for ZFS.
  1654. MergerFS is intended to provide flexible pooling of arbitrary drives
  1655. (local or remote), of arbitrary sizes, and arbitrary filesystems.
  1656. For \f[C]write\ once,\ read\ many\f[] usecases such as bulk media
  1657. storage.
  1658. Where data integrity and backup is managed in other ways.
  1659. In that situation ZFS can introduce major maintance and cost burdens as
  1660. described
  1661. here (http://louwrentius.com/the-hidden-cost-of-using-zfs-for-your-home-nas.html).
  1662. .SS Can drives be written to directly? Outside of mergerfs while pooled?
  1663. .PP
  1664. Yes, however its not recommended to use the same file from within the
  1665. pool and from without at the same time.
  1666. Especially if using caching of any kind (entry_timeout, attr_timeout,
  1667. ac_attr_timeout, negative_timeout, auto_cache, kernel_cache).
  1668. .SS Why do I get an "out of space" / "no space left on device" / ENOSPC
  1669. error even though there appears to be lots of space available?
  1670. .PP
  1671. First make sure you\[aq]ve read the sections above about policies, path
  1672. preservation, branch filtering, and the options \f[B]minfreespace\f[],
  1673. \f[B]moveonenospc\f[], \f[B]statfs\f[], and \f[B]statfs_ignore\f[].
  1674. .PP
  1675. mergerfs is simply presenting a union of the content within multiple
  1676. branches.
  1677. The reported free space is an aggregate of space available within the
  1678. pool (behavior modified by \f[B]statfs\f[] and \f[B]statfs_ignore\f[]).
  1679. It does not represent a contiguous space.
  1680. In the same way that read\-only filesystems, those with quotas, or
  1681. reserved space report the full theoretical space available.
  1682. .PP
  1683. Due to path preservation, branch tagging, read\-only status, and
  1684. \f[B]minfreespace\f[] settings it is perfectly valid that
  1685. \f[C]ENOSPC\f[] / "out of space" / "no space left on device" be
  1686. returned.
  1687. It is doing what was asked of it: filtering possible branches due to
  1688. those settings.
  1689. Only one error can be returned and if one of the reasons for filtering a
  1690. branch was \f[B]minfreespace\f[] then it will be returned as such.
  1691. \f[B]moveonenospc\f[] is only relevant to writing a file which is too
  1692. large for the drive its currently on.
  1693. .PP
  1694. It is also possible that the filesystem selected has run out of inodes.
  1695. Use \f[C]df\ \-i\f[] to list the total and available inodes per
  1696. filesystem.
  1697. .PP
  1698. If you don\[aq]t care about path preservation then simply change the
  1699. \f[C]create\f[] policy to one which isn\[aq]t.
  1700. \f[C]mfs\f[] is probably what most are looking for.
  1701. The reason its not default is because it was originally set to
  1702. \f[C]epmfs\f[] and changing it now would change people\[aq]s setup.
  1703. Such a setting change will likely occur in mergerfs 3.
  1704. .SS Can mergerfs mounts be exported over NFS?
  1705. .PP
  1706. Yes.
  1707. Some clients (Kodi) have issues in which the contents of the NFS mount
  1708. will not be presented but users have found that enabling the
  1709. \f[C]use_ino\f[] option often fixes that problem.
  1710. .SS Can mergerfs mounts be exported over Samba / SMB?
  1711. .PP
  1712. Yes.
  1713. While some users have reported problems it appears to always be related
  1714. to how Samba is setup in relation to permissions.
  1715. .SS How are inodes calculated?
  1716. .PP
  1717. mergerfs\-inode = (original\-inode | (device\-id << 32))
  1718. .PP
  1719. While \f[C]ino_t\f[] is 64 bits only a few filesystems use more than 32.
  1720. Similarly, while \f[C]dev_t\f[] is also 64 bits it was traditionally 16
  1721. bits.
  1722. Bitwise or\[aq]ing them together should work most of the time.
  1723. While totally unique inodes are preferred the overhead which would be
  1724. needed does not seem to outweighted by the benefits.
  1725. .PP
  1726. While atypical, yes, inodes can be reused and not refer to the same
  1727. file.
  1728. The internal id used to reference a file in FUSE is different from the
  1729. inode value presented.
  1730. The former is the \f[C]nodeid\f[] and is actually a tuple of
  1731. (nodeid,generation).
  1732. That tuple is not user facing.
  1733. The inode is merely metadata passed through the kernel and found using
  1734. the \f[C]stat\f[] family of calls or \f[C]readdir\f[].
  1735. .PP
  1736. From FUSE docs regarding \f[C]use_ino\f[]:
  1737. .IP
  1738. .nf
  1739. \f[C]
  1740. Honor\ the\ st_ino\ field\ in\ the\ functions\ getattr()\ and
  1741. fill_dir().\ This\ value\ is\ used\ to\ fill\ in\ the\ st_ino\ field
  1742. in\ the\ stat(2),\ lstat(2),\ fstat(2)\ functions\ and\ the\ d_ino
  1743. field\ in\ the\ readdir(2)\ function.\ The\ filesystem\ does\ not
  1744. have\ to\ guarantee\ uniqueness,\ however\ some\ applications
  1745. rely\ on\ this\ value\ being\ unique\ for\ the\ whole\ filesystem.
  1746. Note\ that\ this\ does\ *not*\ affect\ the\ inode\ that\ libfuse
  1747. and\ the\ kernel\ use\ internally\ (also\ called\ the\ "nodeid").
  1748. \f[]
  1749. .fi
  1750. .SS I notice massive slowdowns of writes over NFS
  1751. .PP
  1752. Due to how NFS works and interacts with FUSE when not using
  1753. \f[C]direct_io\f[] its possible that a getxattr for
  1754. \f[C]security.capability\f[] will be issued prior to any write.
  1755. This will usually result in a massive slowdown for writes.
  1756. Using \f[C]direct_io\f[] will keep this from happening (and generally
  1757. good to enable unless you need the features it disables) but the
  1758. \f[C]security_capability\f[] option can also help by short circuiting
  1759. the call and returning \f[C]ENOATTR\f[].
  1760. .PP
  1761. You could also set \f[C]xattr\f[] to \f[C]noattr\f[] or \f[C]nosys\f[]
  1762. to short circuit or stop all xattr requests.
  1763. .SS What are these .fuse_hidden files?
  1764. .PP
  1765. When not using \f[C]hard_remove\f[] libfuse will create
  1766. \&.fuse_hiddenXXXXXXXX files when an opened file is unlinked.
  1767. This is to simplify "use after unlink" usecases.
  1768. There is a possibility these files end up being picked up by software
  1769. scanning directories and not ignoring hidden files.
  1770. This is rarely a problem but a solution is in the works.
  1771. .PP
  1772. The files are cleaned up once the file is finally closed.
  1773. Only if mergerfs crashes or is killed would they be left around.
  1774. They are safe to remove as they are already unlinked files.
  1775. .SS It\[aq]s mentioned that there are some security issues with mhddfs.
  1776. What are they? How does mergerfs address them?
  1777. .PP
  1778. mhddfs (https://github.com/trapexit/mhddfs) manages running as
  1779. \f[B]root\f[] by calling
  1780. getuid() (https://github.com/trapexit/mhddfs/blob/cae96e6251dd91e2bdc24800b4a18a74044f6672/src/main.c#L319)
  1781. and if it returns \f[B]0\f[] then it will
  1782. chown (http://linux.die.net/man/1/chown) the file.
  1783. Not only is that a race condition but it doesn\[aq]t handle other
  1784. situations.
  1785. Rather than attempting to simulate POSIX ACL behavior the proper way to
  1786. manage this is to use seteuid (http://linux.die.net/man/2/seteuid) and
  1787. setegid (http://linux.die.net/man/2/setegid), in effect becoming the
  1788. user making the original call, and perform the action as them.
  1789. This is what mergerfs does and why mergerfs should always run as root.
  1790. .PP
  1791. In Linux setreuid syscalls apply only to the thread.
  1792. GLIBC hides this away by using realtime signals to inform all threads to
  1793. change credentials.
  1794. Taking after \f[B]Samba\f[], mergerfs uses
  1795. \f[B]syscall(SYS_setreuid,...)\f[] to set the callers credentials for
  1796. that thread only.
  1797. Jumping back to \f[B]root\f[] as necessary should escalated privileges
  1798. be needed (for instance: to clone paths between drives).
  1799. .PP
  1800. For non\-Linux systems mergerfs uses a read\-write lock and changes
  1801. credentials only when necessary.
  1802. If multiple threads are to be user X then only the first one will need
  1803. to change the processes credentials.
  1804. So long as the other threads need to be user X they will take a readlock
  1805. allowing multiple threads to share the credentials.
  1806. Once a request comes in to run as user Y that thread will attempt a
  1807. write lock and change to Y\[aq]s credentials when it can.
  1808. If the ability to give writers priority is supported then that flag will
  1809. be used so threads trying to change credentials don\[aq]t starve.
  1810. This isn\[aq]t the best solution but should work reasonably well
  1811. assuming there are few users.
  1812. .SH PERFORMANCE TWEAKING
  1813. .IP \[bu] 2
  1814. try adding (or removing) \f[C]direct_io\f[]
  1815. .IP \[bu] 2
  1816. try adding (or removing) \f[C]auto_cache\f[] / \f[C]noauto_cache\f[]
  1817. (included in \f[C]defaults\f[])
  1818. .IP \[bu] 2
  1819. try adding (or removing) \f[C]kernel_cache\f[] (don\[aq]t use the
  1820. underlying filesystems directly if enabling \f[C]kernel_cache\f[])
  1821. .IP \[bu] 2
  1822. try adding (or removing) \f[C]splice_move\f[], \f[C]splice_read\f[], and
  1823. \f[C]splice_write\f[] (all three included in \f[C]defaults\f[])
  1824. .IP \[bu] 2
  1825. try increasing cache timeouts \f[C]attr_timeout\f[],
  1826. \f[C]entry_timeout\f[], \f[C]ac_attr_timeout\f[],
  1827. \f[C]negative_timeout\f[]
  1828. .IP \[bu] 2
  1829. try changing the number of worker threads
  1830. .IP \[bu] 2
  1831. try disabling \f[C]security_capability\f[] or \f[C]xattr\f[]
  1832. .IP \[bu] 2
  1833. test theoretical performance using \f[C]nullrw\f[] or mounting a ram
  1834. disk
  1835. .IP \[bu] 2
  1836. use \f[C]symlinkify\f[] if your data is largely static and you need
  1837. native speed reads
  1838. .IP \[bu] 2
  1839. use lvm and lvm cache to place a SSD in front of your HDDs (howto
  1840. coming)
  1841. .SH SUPPORT
  1842. .PP
  1843. Filesystems are very complex and difficult to debug.
  1844. mergerfs, while being just a proxy of sorts, is also very difficult to
  1845. debug given the large number of possible settings it can have itself and
  1846. the massive number of environments it can run in.
  1847. When reporting on a suspected issue \f[B]please, please\f[] include as
  1848. much of the below information as possible otherwise it will be difficult
  1849. or impossible to diagnose.
  1850. Also please make sure to read all of the above documentation as it
  1851. includes nearly every known system or user issue previously encountered.
  1852. .SS Information to include in bug reports
  1853. .IP \[bu] 2
  1854. Version of mergerfs: \f[C]mergerfs\ \-V\f[]
  1855. .IP \[bu] 2
  1856. mergerfs settings: from \f[C]/etc/fstab\f[] or command line execution
  1857. .IP \[bu] 2
  1858. Version of Linux: \f[C]uname\ \-a\f[]
  1859. .IP \[bu] 2
  1860. Versions of any additional software being used
  1861. .IP \[bu] 2
  1862. List of drives, their filesystems, and sizes (before and after issue):
  1863. \f[C]df\ \-h\f[]
  1864. .IP \[bu] 2
  1865. A \f[C]strace\f[] of the app having problems:
  1866. .IP \[bu] 2
  1867. \f[C]strace\ \-f\ \-o\ /tmp/app.strace.txt\ <cmd>\f[]
  1868. .IP \[bu] 2
  1869. A \f[C]strace\f[] of mergerfs while the program is trying to do whatever
  1870. it\[aq]s failing to do:
  1871. .IP \[bu] 2
  1872. \f[C]strace\ \-f\ \-p\ <mergerfsPID>\ \-o\ /tmp/mergerfs.strace.txt\f[]
  1873. .IP \[bu] 2
  1874. \f[B]Precise\f[] directions on replicating the issue.
  1875. Do not leave \f[B]anything\f[] out.
  1876. .IP \[bu] 2
  1877. Try to recreate the problem in the simplist way using standard programs.
  1878. .SS Contact / Issue submission
  1879. .IP \[bu] 2
  1880. github.com: https://github.com/trapexit/mergerfs/issues
  1881. .IP \[bu] 2
  1882. email: trapexit\@spawn.link
  1883. .IP \[bu] 2
  1884. twitter: https://twitter.com/_trapexit
  1885. .IP \[bu] 2
  1886. reddit: https://www.reddit.com/user/trapexit
  1887. .SS Support development
  1888. .PP
  1889. This software is free to use and released under a very liberal license.
  1890. That said if you like this software and would like to support its
  1891. development donations are welcome.
  1892. .IP \[bu] 2
  1893. PayPal: trapexit\@spawn.link
  1894. .IP \[bu] 2
  1895. Patreon: https://www.patreon.com/trapexit
  1896. .IP \[bu] 2
  1897. SubscribeStar: https://www.subscribestar.com/trapexit
  1898. .IP \[bu] 2
  1899. Bitcoin (BTC): 12CdMhEPQVmjz3SSynkAEuD5q9JmhTDCZA
  1900. .IP \[bu] 2
  1901. Bitcoin Cash (BCH): 1AjPqZZhu7GVEs6JFPjHmtsvmDL4euzMzp
  1902. .IP \[bu] 2
  1903. Ethereum (ETH): 0x09A166B11fCC127324C7fc5f1B572255b3046E94
  1904. .IP \[bu] 2
  1905. Litecoin (LTC): LXAsq6yc6zYU3EbcqyWtHBrH1Ypx4GjUjm
  1906. .IP \[bu] 2
  1907. Ripple (XRP): rNACR2hqGjpbHuCKwmJ4pDpd2zRfuRATcE
  1908. .SH LINKS
  1909. .IP \[bu] 2
  1910. https://spawn.link
  1911. .IP \[bu] 2
  1912. https://github.com/trapexit/mergerfs
  1913. .IP \[bu] 2
  1914. https://github.com/trapexit/mergerfs\-tools
  1915. .IP \[bu] 2
  1916. https://github.com/trapexit/scorch
  1917. .IP \[bu] 2
  1918. https://github.com/trapexit/bbf
  1919. .IP \[bu] 2
  1920. https://github.com/trapexit/backup\-and\-recovery\-howtos
  1921. .SH AUTHORS
  1922. Antonio SJ Musumeci <trapexit@spawn.link>.