mergerfs/man/mergerfs.1

.\"t
.\" Automatically generated by Pandoc 2.9.2.1
.\"
.TH "mergerfs" "1" "" "mergerfs user manual" ""
.hy
.SH NAME
.PP
mergerfs - a featureful union filesystem
.SH SYNOPSIS
.PP
mergerfs -o<options> <branches> <mountpoint>
.SH DESCRIPTION
.PP
\f[B]mergerfs\f[R] is a union filesystem geared towards simplifying
storage and management of files across numerous commodity storage
devices.
It is similar to \f[B]mhddfs\f[R], \f[B]unionfs\f[R], and
\f[B]aufs\f[R].
.SH FEATURES
.IP \[bu] 2
Configurable behaviors / file placement
.IP \[bu] 2
Ability to add or remove filesystems at will
.IP \[bu] 2
Resistance to individual filesystem failure
.IP \[bu] 2
Support for extended attributes (xattrs)
.IP \[bu] 2
Support for file attributes (chattr)
.IP \[bu] 2
Runtime configurable (via xattrs)
.IP \[bu] 2
Works with heterogeneous filesystem types
.IP \[bu] 2
Moving of file when filesystem runs out of space while writing
.IP \[bu] 2
Ignore read-only filesystems when creating files
.IP \[bu] 2
Turn read-only files into symlinks to underlying file
.IP \[bu] 2
Hard link copy-on-write / CoW
.IP \[bu] 2
Support for POSIX ACLs
.IP \[bu] 2
Misc other things
.SH HOW IT WORKS
.PP
mergerfs logically merges multiple paths together.
Think a union of sets.
The file/s or directory/s acted on or presented through mergerfs are
based on the policy chosen for that particular action.
Read more about policies below.
.IP
.nf
\f[C]
A         +      B        =       C
/disk1           /disk2           /merged
|                |                |
+-- /dir1        +-- /dir1        +-- /dir1
|   |            |   |            |   |
|   +-- file1    |   +-- file2    |   +-- file1
|                |   +-- file3    |   +-- file2
+-- /dir2        |                |   +-- file3
|   |            +-- /dir3        |
|   +-- file4        |            +-- /dir2
|                     +-- file5   |   |
+-- file6                         |   +-- file4
                                  |
                                  +-- /dir3
                                  |   |
                                  |   +-- file5
                                  |
                                  +-- file6
\f[R]
.fi
.PP
mergerfs does \f[B]not\f[R] support the copy-on-write (CoW) or whiteout
behaviors found in \f[B]aufs\f[R] and \f[B]overlayfs\f[R].
You can \f[B]not\f[R] mount a read-only filesystem and write to it.
However, mergerfs will ignore read-only filesystems when creating new
files so you can mix read-write and read-only filesystems.
It also does \f[B]not\f[R] split data across filesystems.
It is not RAID0 / striping.
It is simply a union of other filesystems.
.SH TERMINOLOGY
.IP \[bu] 2
branch: A base path used in the pool.
.IP \[bu] 2
pool: The mergerfs mount.
The union of the branches.
.IP \[bu] 2
relative path: The path in the pool relative to the branch and mount.
.IP \[bu] 2
function: A filesystem call (open, unlink, create, getattr, rmdir, etc.)
.IP \[bu] 2
category: A collection of functions based on basic behavior (action,
create, search).
.IP \[bu] 2
policy: The algorithm used to select a file when performing a function.
.IP \[bu] 2
path preservation: Aspect of some policies which includes checking the
path for which a file would be created.
.SH BASIC SETUP
.PP
If you don\[cq]t already know that you have a special use case then just
start with one of the following option sets.
.SS You need \f[C]mmap\f[R] (used by rtorrent and many sqlite3 base software)
.PP
\f[C]cache.files=partial,dropcacheonclose=true,category.create=mfs\f[R]
.SS You don\[cq]t need \f[C]mmap\f[R]
.PP
\f[C]cache.files=off,dropcacheonclose=true,category.create=mfs\f[R]
.SS Command Line
.PP
\f[C]mergerfs -o cache.files=partial,dropcacheonclose=true,category.create=mfs /mnt/hdd0:/mnt/hdd1 /media\f[R]
.SS /etc/fstab
.PP
\f[C]/mnt/hdd0:/mnt/hdd1 /media fuse.mergerfs cache.files=partial,dropcacheonclose=true,category.create=mfs 0 0\f[R]
.SS systemd mount
.PP
https://github.com/trapexit/mergerfs/wiki/systemd
.IP
.nf
\f[C]
[Unit]
Description=mergerfs service

[Service]
Type=simple
KillMode=none
ExecStart=/usr/bin/mergerfs \[rs]
  -f \[rs]
  -o cache.files=partial \[rs]
  -o dropcacheonclose=true \[rs]
  -o category.create=mfs \[rs]
  /mnt/hdd0:/mnt/hdd1 \[rs]
  /media
ExecStop=/bin/fusermount -uz /media
Restart=on-failure

[Install]
WantedBy=default.target
\f[R]
.fi
.PP
See the mergerfs wiki for real world
deployments (https://github.com/trapexit/mergerfs/wiki/Real-World-Deployments)
for comparisons / ideas.
.SH OPTIONS
.PP
These options are the same regardless of whether you use them with the
\f[C]mergerfs\f[R] commandline program, in fstab, or in a config file.
.SS mount options
.IP \[bu] 2
\f[B]config\f[R]: Path to a config file.
Same arguments as below in key=val / ini style format.
.IP \[bu] 2
\f[B]branches\f[R]: Colon delimited list of branches.
.IP \[bu] 2
\f[B]minfreespace=SIZE\f[R]: The minimum space value used for creation
policies.
Can be overridden by branch specific option.
Understands `K', `M', and `G' to represent kilobyte, megabyte, and
gigabyte respectively.
(default: 4G)
.IP \[bu] 2
\f[B]moveonenospc=BOOL|POLICY\f[R]: When enabled if a \f[B]write\f[R]
fails with \f[B]ENOSPC\f[R] (no space left on device) or
\f[B]EDQUOT\f[R] (disk quota exceeded) the policy selected will run to
find a new location for the file.
An attempt to move the file to that branch will occur (keeping all
metadata possible) and if successful the original is unlinked and the
write retried.
(default: false, true = mfs)
.IP \[bu] 2
\f[B]inodecalc=passthrough|path-hash|devino-hash|hybrid-hash\f[R]:
Selects the inode calculation algorithm.
(default: hybrid-hash)
.IP \[bu] 2
\f[B]dropcacheonclose=BOOL\f[R]: When a file is requested to be closed
call \f[C]posix_fadvise\f[R] on it first to instruct the kernel that we
no longer need the data and it can drop its cache.
Recommended when
\f[B]cache.files=partial|full|auto-full|per-process\f[R] to limit double
caching.
(default: false)
.IP \[bu] 2
\f[B]symlinkify=BOOL\f[R]: When enabled and a file is not writable and
its mtime or ctime is older than \f[B]symlinkify_timeout\f[R] files will
be reported as symlinks to the original files.
Please read more below before using.
(default: false)
.IP \[bu] 2
\f[B]symlinkify_timeout=UINT\f[R]: Time to wait, in seconds, to activate
the \f[B]symlinkify\f[R] behavior.
(default: 3600)
.IP \[bu] 2
\f[B]nullrw=BOOL\f[R]: Turns reads and writes into no-ops.
The request will succeed but do nothing.
Useful for benchmarking mergerfs.
(default: false)
.IP \[bu] 2
\f[B]lazy-umount-mountpoint=BOOL\f[R]: mergerfs will attempt to
\[lq]lazy umount\[rq] the mountpoint before mounting itself.
Useful when performing live upgrades of mergerfs.
(default: false)
.IP \[bu] 2
\f[B]ignorepponrename=BOOL\f[R]: Ignore path preserving on rename.
Typically rename and link act differently depending on the policy of
\f[C]create\f[R] (read below).
Enabling this will cause rename and link to always use the non-path
preserving behavior.
This means files, when renamed or linked, will stay on the same
filesystem.
(default: false)
.IP \[bu] 2
\f[B]security_capability=BOOL\f[R]: If false return ENOATTR when xattr
security.capability is queried.
(default: true)
.IP \[bu] 2
\f[B]xattr=passthrough|noattr|nosys\f[R]: Runtime control of xattrs.
Default is to passthrough xattr requests.
`noattr' will short circuit as if nothing exists.
`nosys' will respond with ENOSYS as if xattrs are not supported or
disabled.
(default: passthrough)
.IP \[bu] 2
\f[B]link_cow=BOOL\f[R]: When enabled if a regular file is opened which
has a link count > 1 it will copy the file to a temporary file and
rename over the original.
Breaking the link and providing a basic copy-on-write function similar
to cow-shell.
(default: false)
.IP \[bu] 2
\f[B]statfs=base|full\f[R]: Controls how statfs works.
`base' means it will always use all branches in statfs calculations.
`full' is in effect path preserving and only includes branches where the
path exists.
(default: base)
.IP \[bu] 2
\f[B]statfs_ignore=none|ro|nc\f[R]: `ro' will cause statfs calculations
to ignore available space for branches mounted or tagged as `read-only'
or `no create'.
`nc' will ignore available space for branches tagged as `no create'.
(default: none)
.IP \[bu] 2
\f[B]nfsopenhack=off|git|all\f[R]: A workaround for exporting mergerfs
over NFS where there are issues with creating files for write while
setting the mode to read-only.
(default: off)
.IP \[bu] 2
\f[B]branches-mount-timeout=UINT\f[R]: Number of seconds to wait at
startup for branches to be a mount other than the mountpoint\[cq]s
filesystem.
(default: 0)
.IP \[bu] 2
\f[B]follow-symlinks=never|directory|regular|all\f[R]: Turns symlinks
into what they point to.
(default: never)
.IP \[bu] 2
\f[B]link-exdev=passthrough|rel-symlink|abs-base-symlink|abs-pool-symlink\f[R]:
When a link fails with EXDEV optionally create a symlink to the file
instead.
.IP \[bu] 2
\f[B]rename-exdev=passthrough|rel-symlink|abs-symlink\f[R]: When a
rename fails with EXDEV optionally move the file to a special directory
and symlink to it.
.IP \[bu] 2
\f[B]readahead=UINT\f[R]: Set readahead (in kilobytes) for mergerfs and
branches if greater than 0.
(default: 0)
.IP \[bu] 2
\f[B]posix_acl=BOOL\f[R]: Enable POSIX ACL support (if supported by
kernel and underlying filesystem).
(default: false)
.IP \[bu] 2
\f[B]async_read=BOOL\f[R]: Perform reads asynchronously.
If disabled or unavailable the kernel will ensure there is at most one
pending read request per file handle and will attempt to order requests
by offset.
(default: true)
.IP \[bu] 2
\f[B]fuse_msg_size=UINT\f[R]: Set the max number of pages per FUSE
message.
Only available on Linux >= 4.20 and ignored otherwise.
(min: 1; max: 256; default: 256)
.IP \[bu] 2
\f[B]threads=INT\f[R]: Number of threads to use.
When used alone (\f[C]process-thread-count=-1\f[R]) it sets the number
of threads reading and processing FUSE messages.
When used together it sets the number of threads reading from FUSE.
When set to zero it will attempt to discover and use the number of
logical cores.
If the thread count is set negative it will look up the number of cores
then divide by the absolute value.
ie.
threads=-2 on an 8 core machine will result in 8 / 2 = 4 threads.
There will always be at least 1 thread.
If set to -1 in combination with \f[C]process-thread-count\f[R] then it
will try to pick reasonable values based on CPU thread count.
NOTE: higher number of threads increases parallelism but usually
decreases throughput.
(default: 0)
.IP \[bu] 2
\f[B]read-thread-count=INT\f[R]: Alias for \f[C]threads\f[R].
.IP \[bu] 2
\f[B]process-thread-count=INT\f[R]: Enables separate thread pool to
asynchronously process FUSE requests.
In this mode \f[C]read-thread-count\f[R] refers to the number of threads
reading FUSE messages which are dispatched to process threads.
-1 means disabled otherwise acts like \f[C]read-thread-count\f[R].
(default: -1)
.IP \[bu] 2
\f[B]process-thread-queue-depth=UINT\f[R]: Sets the number of requests
any single process thread can have queued up at one time.
Meaning the total memory usage of the queues is queue depth multiplied
by the number of process threads plus read thread count.
0 sets the depth to the same as the process thread count.
(default: 0)
.IP \[bu] 2
\f[B]pin-threads=STR\f[R]: Selects a strategy to pin threads to CPUs
(default: unset)
.IP \[bu] 2
\f[B]flush-on-close=never|always|opened-for-write\f[R]: Flush data cache
on file close.
Mostly for when writeback is enabled or merging network filesystems.
(default: opened-for-write)
.IP \[bu] 2
\f[B]scheduling-priority=INT\f[R]: Set mergerfs\[cq] scheduling
priority.
Valid values range from -20 to 19.
See \f[C]setpriority\f[R] man page for more details.
(default: -10)
.IP \[bu] 2
\f[B]fsname=STR\f[R]: Sets the name of the filesystem as seen in
\f[B]mount\f[R], \f[B]df\f[R], etc.
Defaults to a list of the source paths concatenated together with the
longest common prefix removed.
.IP \[bu] 2
\f[B]func.FUNC=POLICY\f[R]: Sets the specific FUSE function\[cq]s
policy.
See below for the list of value types.
Example: \f[B]func.getattr=newest\f[R]
.IP \[bu] 2
\f[B]func.readdir=seq|cosr|cor|cosr:INT|cor:INT\f[R]: Sets
\f[C]readdir\f[R] policy.
INT value sets the number of threads to use for concurrency.
(default: seq)
.IP \[bu] 2
\f[B]category.action=POLICY\f[R]: Sets policy of all FUSE functions in
the action category.
(default: epall)
.IP \[bu] 2
\f[B]category.create=POLICY\f[R]: Sets policy of all FUSE functions in
the create category.
(default: epmfs)
.IP \[bu] 2
\f[B]category.search=POLICY\f[R]: Sets policy of all FUSE functions in
the search category.
(default: ff)
.IP \[bu] 2
\f[B]cache.open=UINT\f[R]: `open' policy cache timeout in seconds.
(default: 0)
.IP \[bu] 2
\f[B]cache.statfs=UINT\f[R]: `statfs' cache timeout in seconds.
(default:
.RS 2
.RE
.IP \[bu] 2
\f[B]cache.attr=UINT\f[R]: File attribute cache timeout in seconds.
(default: 1)
.IP \[bu] 2
\f[B]cache.entry=UINT\f[R]: File name lookup cache timeout in seconds.
(default: 1)
.IP \[bu] 2
\f[B]cache.negative_entry=UINT\f[R]: Negative file name lookup cache
timeout in seconds.
(default: 0)
.IP \[bu] 2
\f[B]cache.files=libfuse|off|partial|full|auto-full|per-process\f[R]:
File page caching mode (default: libfuse)
.IP \[bu] 2
\f[B]cache.files.process-names=LIST\f[R]: A pipe | delimited list of
process comm (https://man7.org/linux/man-pages/man5/proc.5.html) names
to enable page caching for when \f[C]cache.files=per-process\f[R].
(default: \[lq]rtorrent|qbittorrent-nox\[rq])
.IP \[bu] 2
\f[B]cache.writeback=BOOL\f[R]: Enable kernel writeback caching
(default: false)
.IP \[bu] 2
\f[B]cache.symlinks=BOOL\f[R]: Cache symlinks (if supported by kernel)
(default: false)
.IP \[bu] 2
\f[B]cache.readdir=BOOL\f[R]: Cache readdir (if supported by kernel)
(default: false)
.IP \[bu] 2
\f[B]parallel-direct-writes=BOOL\f[R]: Allow the kernel to dispatch
multiple, parallel (non-extending) write requests for files opened with
\f[C]cache.files=per-process\f[R] (if the process is not in
\f[C]process-names\f[R]) or \f[C]cache.files=off\f[R].
(This requires kernel support, and was added in v6.2)
.IP \[bu] 2
\f[B]direct_io\f[R]: deprecated - Bypass page cache.
Use \f[C]cache.files=off\f[R] instead.
(default: false)
.IP \[bu] 2
\f[B]kernel_cache\f[R]: deprecated - Do not invalidate data cache on
file open.
Use \f[C]cache.files=full\f[R] instead.
(default: false)
.IP \[bu] 2
\f[B]auto_cache\f[R]: deprecated - Invalidate data cache if file mtime
or size change.
Use \f[C]cache.files=auto-full\f[R] instead.
(default: false)
.IP \[bu] 2
\f[B]async_read\f[R]: deprecated - Perform reads asynchronously.
Use \f[C]async_read=true\f[R] instead.
.IP \[bu] 2
\f[B]sync_read\f[R]: deprecated - Perform reads synchronously.
Use \f[C]async_read=false\f[R] instead.
.IP \[bu] 2
\f[B]splice_read\f[R]: deprecated - Does nothing.
.IP \[bu] 2
\f[B]splice_write\f[R]: deprecated - Does nothing.
.IP \[bu] 2
\f[B]splice_move\f[R]: deprecated - Does nothing.
.IP \[bu] 2
\f[B]allow_other\f[R]: deprecated - mergerfs v2.35.0 and newer sets this
FUSE option automatically if running as root.
.IP \[bu] 2
\f[B]use_ino\f[R]: deprecated - mergerfs should always control inode
calculation so this is enabled all the time.
.PP
\f[B]NOTE:\f[R] Options are evaluated in the order listed so if the
options are \f[B]func.rmdir=rand,category.action=ff\f[R] the
\f[B]action\f[R] category setting will override the \f[B]rmdir\f[R]
setting.
.PP
\f[B]NOTE:\f[R] Always look at the documentation for the version of
mergerfs you\[cq]re using.
Not all features are available in older releases.
Use \f[C]man mergerfs\f[R] or find the docs as linked in the release.
.SS Value Types
.IP \[bu] 2
BOOL = `true' | `false'
.IP \[bu] 2
INT = [MIN_INT,MAX_INT]
.IP \[bu] 2
UINT = [0,MAX_INT]
.IP \[bu] 2
SIZE = `NNM'; NN = INT, M = `K' | `M' | `G' | `T'
.IP \[bu] 2
STR = string (may refer to an enumerated value, see details of argument)
.IP \[bu] 2
FUNC = filesystem function
.IP \[bu] 2
CATEGORY = function category
.IP \[bu] 2
POLICY = mergerfs function policy
.SS branches
.PP
The `branches' argument is a colon (`:') delimited list of paths to be
pooled together.
It does not matter if the paths are on the same or different filesystems
nor does it matter the filesystem type (within reason).
Used and available space will not be duplicated for paths on the same
filesystem and any features which aren\[cq]t supported by the underlying
filesystem (such as file attributes or extended attributes) will return
the appropriate errors.
.PP
Branches currently have two options which can be set.
A type which impacts whether or not the branch is included in a policy
calculation and a individual minfreespace value.
The values are set by prepending an \f[C]=\f[R] at the end of a branch
designation and using commas as delimiters.
Example: \f[C]/mnt/drive=RW,1234\f[R]
.SS branch mode
.IP \[bu] 2
RW: (read/write) - Default behavior.
Will be eligible in all policy categories.
.IP \[bu] 2
RO: (read-only) - Will be excluded from \f[C]create\f[R] and
\f[C]action\f[R] policies.
Same as a read-only mounted filesystem would be (though faster to
process).
.IP \[bu] 2
NC: (no-create) - Will be excluded from \f[C]create\f[R] policies.
You can\[cq]t create on that branch but you can change or delete.
.SS minfreespace
.PP
Same purpose and syntax as the global option but specific to the branch.
If not set the global value is used.
.SS globbing
.PP
To make it easier to include multiple branches mergerfs supports
globbing (http://linux.die.net/man/7/glob).
\f[B]The globbing tokens MUST be escaped when using via the shell else
the shell itself will apply the glob itself.\f[R]
.IP
.nf
\f[C]
# mergerfs /mnt/hdd\[rs]*:/mnt/ssd /media
\f[R]
.fi
.PP
The above line will use all mount points in /mnt prefixed with
\f[B]hdd\f[R] and \f[B]ssd\f[R].
.PP
To have the pool mounted at boot or otherwise accessible from related
tools use \f[B]/etc/fstab\f[R].
.IP
.nf
\f[C]
# <file system>        <mount point>  <type>         <options>             <dump>  <pass>
/mnt/hdd*:/mnt/ssd    /media          fuse.mergerfs  minfreespace=16G      0       0
\f[R]
.fi
.PP
\f[B]NOTE:\f[R] the globbing is done at mount or when updated using the
runtime API.
If a new directory is added matching the glob after the fact it will not
be automatically included.
.PP
\f[B]NOTE:\f[R] for mounting via \f[B]fstab\f[R] to work you must have
\f[B]mount.fuse\f[R] installed.
For Ubuntu/Debian it is included in the \f[B]fuse\f[R] package.
.SS inodecalc
.PP
Inodes (st_ino) are unique identifiers within a filesystem.
Each mounted filesystem has device ID (st_dev) as well and together they
can uniquely identify a file on the whole of the system.
Entries on the same device with the same inode are in fact references to
the same underlying file.
It is a many to one relationship between names and an inode.
Directories, however, do not have multiple links on most systems due to
the complexity they add.
.PP
FUSE allows the server (mergerfs) to set inode values but not device
IDs.
Creating an inode value is somewhat complex in mergerfs\[cq] case as
files aren\[cq]t really in its control.
If a policy changes what directory or file is to be selected or
something changes out of band it becomes unclear what value should be
used.
Most software does not to care what the values are but those that do
often break if a value changes unexpectedly.
The tool \f[C]find\f[R] will abort a directory walk if it sees a
directory inode change.
NFS will return stale handle errors if the inode changes out of band.
File dedup tools will usually leverage device ids and inodes as a
shortcut in searching for duplicate files and would resort to full file
comparisons should it find different inode values.
.PP
mergerfs offers multiple ways to calculate the inode in hopes of
covering different usecases.
.IP \[bu] 2
passthrough: Passes through the underlying inode value.
Mostly intended for testing as using this does not address any of the
problems mentioned above and could confuse file deduplication software
as inodes from different filesystems can be the same.
.IP \[bu] 2
path-hash: Hashes the relative path of the entry in question.
The underlying file\[cq]s values are completely ignored.
This means the inode value will always be the same for that file path.
This is useful when using NFS and you make changes out of band such as
copy data between branches.
This also means that entries that do point to the same file will not be
recognizable via inodes.
That \f[B]does not\f[R] mean hard links don\[cq]t work.
They will.
.IP \[bu] 2
path-hash32: 32bit version of path-hash.
.IP \[bu] 2
devino-hash: Hashes the device id and inode of the underlying entry.
This won\[cq]t prevent issues with NFS should the policy pick a
different file or files move out of band but will present the same inode
for underlying files that do too.
.IP \[bu] 2
devino-hash32: 32bit version of devino-hash.
.IP \[bu] 2
hybrid-hash: Performs \f[C]path-hash\f[R] on directories and
\f[C]devino-hash\f[R] on other file types.
Since directories can\[cq]t have hard links the static value won\[cq]t
make a difference and the files will get values useful for finding
duplicates.
Probably the best to use if not using NFS.
As such it is the default.
.IP \[bu] 2
hybrid-hash32: 32bit version of hybrid-hash.
.PP
32bit versions are provided as there is some software which does not
handle 64bit inodes well.
.PP
While there is a risk of hash collision in tests of a couple million
entries there were zero collisions.
Unlike a typical filesystem FUSE filesystems can reuse inodes and not
refer to the same entry.
The internal identifier used to reference a file in FUSE is different
from the inode value presented.
The former is the \f[C]nodeid\f[R] and is actually a tuple of 2 64bit
values: \f[C]nodeid\f[R] and \f[C]generation\f[R].
This tuple is not client facing.
The inode that is presented to the client is passed through the kernel
uninterpreted.
.PP
From FUSE docs for \f[C]use_ino\f[R]:
.IP
.nf
\f[C]
Honor the st_ino field in the functions getattr() and
fill_dir(). This value is used to fill in the st_ino field
in the stat(2), lstat(2), fstat(2) functions and the d_ino
field in the readdir(2) function. The filesystem does not
have to guarantee uniqueness, however some applications
rely on this value being unique for the whole filesystem.
Note that this does *not* affect the inode that libfuse
and the kernel use internally (also called the \[dq]nodeid\[dq]).
\f[R]
.fi
.PP
As of version 2.35.0 the \f[C]use_ino\f[R] option has been removed.
mergerfs should always be managing inode values.
.SS pin-threads
.PP
Simple strategies for pinning read and/or process threads.
If process threads are not enabled than the strategy simply works on the
read threads.
Invalid values are ignored.
.IP \[bu] 2
R1L: All read threads pinned to a single logical CPU.
.IP \[bu] 2
R1P: All read threads pinned to a single physical CPU.
.IP \[bu] 2
RP1L: All read and process threads pinned to a single logical CPU.
.IP \[bu] 2
RP1P: All read and process threads pinned to a single physical CPU.
.IP \[bu] 2
R1LP1L: All read threads pinned to a single logical CPU, all process
threads pinned to a (if possible) different logical CPU.
.IP \[bu] 2
R1PP1P: All read threads pinned to a single physical CPU, all process
threads pinned to a (if possible) different logical CPU.
.IP \[bu] 2
RPSL: All read and process threads are spread across all logical CPUs.
.IP \[bu] 2
RPSP: All read and process threads are spread across all physical CPUs.
.IP \[bu] 2
R1PPSP: All read threads are pinned to a single physical CPU while
process threads are spread across all other phsycial CPUs.
.SS fuse_msg_size
.PP
FUSE applications communicate with the kernel over a special character
device: \f[C]/dev/fuse\f[R].
A large portion of the overhead associated with FUSE is the cost of
going back and forth from user space and kernel space over that device.
Generally speaking the fewer trips needed the better the performance
will be.
Reducing the number of trips can be done a number of ways.
Kernel level caching and increasing message sizes being two significant
ones.
When it comes to reads and writes if the message size is doubled the
number of trips are approximately halved.
.PP
In Linux 4.20 a new feature was added allowing the negotiation of the
max message size.
Since the size is in multiples of
pages (https://en.wikipedia.org/wiki/Page_(computer_memory)) the feature
is called \f[C]max_pages\f[R].
There is a maximum \f[C]max_pages\f[R] value of 256 (1MiB) and minimum
of 1 (4KiB).
The default used by Linux >=4.20, and hardcoded value used before 4.20,
is 32 (128KiB).
In mergerfs its referred to as \f[C]fuse_msg_size\f[R] to make it clear
what it impacts and provide some abstraction.
.PP
Since there should be no downsides to increasing \f[C]fuse_msg_size\f[R]
/ \f[C]max_pages\f[R], outside a minor bump in RAM usage due to larger
message buffers, mergerfs defaults the value to 256.
On kernels before 4.20 the value has no effect.
The reason the value is configurable is to enable experimentation and
benchmarking.
See the BENCHMARKING section for examples.
.SS follow-symlinks
.PP
This feature, when enabled, will cause symlinks to be interpreted by
mergerfs as their target (depending on the mode).
.PP
When there is a getattr/stat request for a file mergerfs will check if
the file is a symlink and depending on the \f[C]follow-symlinks\f[R]
setting will replace the information about the symlink with that of that
which it points to.
.PP
When unlink\[cq]ing or rmdir\[cq]ing the followed symlink it will remove
the symlink itself and not that which it points to.
.IP \[bu] 2
never: Behave as normal.
Symlinks are treated as such.
.IP \[bu] 2
directory: Resolve symlinks only which point to directories.
.IP \[bu] 2
regular: Resolve symlinks only which point to regular files.
.IP \[bu] 2
all: Resolve all symlinks to that which they point to.
.PP
Symlinks which do not point to anything are left as is.
.PP
WARNING: This feature works but there might be edge cases yet found.
If you find any odd behaviors please file a ticket on
github (https://github.com/trapexit/mergerfs/issues).
.SS link-exdev
.PP
If using path preservation and a \f[C]link\f[R] fails with EXDEV make a
call to \f[C]symlink\f[R] where the \f[C]target\f[R] is the
\f[C]oldlink\f[R] and the \f[C]linkpath\f[R] is the \f[C]newpath\f[R].
The \f[C]target\f[R] value is determined by the value of
\f[C]link-exdev\f[R].
.IP \[bu] 2
passthrough: Return EXDEV as normal.
.IP \[bu] 2
rel-symlink: A relative path from the \f[C]newpath\f[R].
.IP \[bu] 2
abs-base-symlink: A absolute value using the underlying branch.
.IP \[bu] 2
abs-pool-symlink: A absolute value using the mergerfs mount point.
.PP
NOTE: It is possible that some applications check the file they link.
In those cases it is possible it will error or complain.
.SS rename-exdev
.PP
If using path preservation and a \f[C]rename\f[R] fails with EXDEV:
.IP "1." 3
Move file from \f[B]/branch/a/b/c\f[R] to
\f[B]/branch/.mergerfs_rename_exdev/a/b/c\f[R].
.IP "2." 3
symlink the rename\[cq]s \f[C]newpath\f[R] to the moved file.
.PP
The \f[C]target\f[R] value is determined by the value of
\f[C]rename-exdev\f[R].
.IP \[bu] 2
passthrough: Return EXDEV as normal.
.IP \[bu] 2
rel-symlink: A relative path from the \f[C]newpath\f[R].
.IP \[bu] 2
abs-symlink: A absolute value using the mergerfs mount point.
.PP
NOTE: It is possible that some applications check the file they rename.
In those cases it is possible it will error or complain.
.PP
NOTE: The reason \f[C]abs-symlink\f[R] is not split into two like
\f[C]link-exdev\f[R] is due to the complexities in managing absolute
base symlinks when multiple \f[C]oldpaths\f[R] exist.
.SS symlinkify
.PP
Due to the levels of indirection introduced by mergerfs and the
underlying technology FUSE there can be varying levels of performance
degradation.
This feature will turn non-directories which are not writable into
symlinks to the original file found by the \f[C]readlink\f[R] policy
after the mtime and ctime are older than the timeout.
.PP
\f[B]WARNING:\f[R] The current implementation has a known issue in which
if the file is open and being used when the file is converted to a
symlink then the application which has that file open will receive an
error when using it.
This is unlikely to occur in practice but is something to keep in mind.
.PP
\f[B]WARNING:\f[R] Some backup solutions, such as CrashPlan, do not
backup the target of a symlink.
If using this feature it will be necessary to point any backup software
to the original filesystems or configure the software to follow symlinks
if such an option is available.
Alternatively create two mounts.
One for backup and one for general consumption.
.SS nullrw
.PP
Due to how FUSE works there is an overhead to all requests made to a
FUSE filesystem that wouldn\[cq]t exist for an in kernel one.
Meaning that even a simple passthrough will have some slowdown.
However, generally the overhead is minimal in comparison to the cost of
the underlying I/O.
By disabling the underlying I/O we can test the theoretical performance
boundaries.
.PP
By enabling \f[C]nullrw\f[R] mergerfs will work as it always does
\f[B]except\f[R] that all reads and writes will be no-ops.
A write will succeed (the size of the write will be returned as if it
were successful) but mergerfs does nothing with the data it was given.
Similarly a read will return the size requested but won\[cq]t touch the
buffer.
.PP
See the BENCHMARKING section for suggestions on how to test.
.SS xattr
.PP
Runtime extended attribute support can be managed via the
\f[C]xattr\f[R] option.
By default it will passthrough any xattr calls.
Given xattr support is rarely used and can have significant performance
implications mergerfs allows it to be disabled at runtime.
The performance problems mostly comes when file caching is enabled.
The kernel will send a \f[C]getxattr\f[R] for
\f[C]security.capability\f[R] \f[I]before every single write\f[R].
It doesn\[cq]t cache the responses to any \f[C]getxattr\f[R].
This might be addressed in the future but for now mergerfs can really
only offer the following workarounds.
.PP
\f[C]noattr\f[R] will cause mergerfs to short circuit all xattr calls
and return ENOATTR where appropriate.
mergerfs still gets all the requests but they will not be forwarded on
to the underlying filesystems.
The runtime control will still function in this mode.
.PP
\f[C]nosys\f[R] will cause mergerfs to return ENOSYS for any xattr call.
The difference with \f[C]noattr\f[R] is that the kernel will cache this
fact and itself short circuit future calls.
This is more efficient than \f[C]noattr\f[R] but will cause
mergerfs\[cq] runtime control via the hidden file to stop working.
.SS nfsopenhack
.PP
NFS is not fully POSIX compliant and historically certain behaviors,
such as opening files with O_EXCL, are not or not well supported.
When mergerfs (or any FUSE filesystem) is exported over NFS some of
these issues come up due to how NFS and FUSE interact.
.PP
This hack addresses the issue where the creation of a file with a
read-only mode but with a read/write or write only flag.
Normally this is perfectly valid but NFS chops the one open call into
multiple calls.
Exactly how it is translated depends on the configuration and versions
of the NFS server and clients but it results in a permission error
because a normal user is not allowed to open a read-only file as
writable.
.PP
Even though it\[cq]s a more niche situation this hack breaks normal
security and behavior and as such is \f[C]off\f[R] by default.
If set to \f[C]git\f[R] it will only perform the hack when the path in
question includes \f[C]/.git/\f[R].
\f[C]all\f[R] will result it applying anytime a read-only file which is
empty is opened for writing.
.SH FUNCTIONS, CATEGORIES and POLICIES
.PP
The POSIX filesystem API is made up of a number of functions.
\f[B]creat\f[R], \f[B]stat\f[R], \f[B]chown\f[R], etc.
For ease of configuration in mergerfs most of the core functions are
grouped into 3 categories: \f[B]action\f[R], \f[B]create\f[R], and
\f[B]search\f[R].
These functions and categories can be assigned a policy which dictates
which branch is chosen when performing that function.
.PP
Some functions, listed in the category \f[C]N/A\f[R] below, can not be
assigned the normal policies.
These functions work with file handles, rather than file paths, which
were created by \f[C]open\f[R] or \f[C]create\f[R].
That said many times the current FUSE kernel driver will not always
provide the file handle when a client calls \f[C]fgetattr\f[R],
\f[C]fchown\f[R], \f[C]fchmod\f[R], \f[C]futimens\f[R],
\f[C]ftruncate\f[R], etc.
This means it will call the regular, path based, versions.
\f[C]statfs\f[R]\[cq]s behavior can be modified via other options.
.PP
When using policies which are based on a branch\[cq]s available space
the base path provided is used.
Not the full path to the file in question.
Meaning that mounts in the branch won\[cq]t be considered in the space
calculations.
The reason is that it doesn\[cq]t really work for non-path preserving
policies and can lead to non-obvious behaviors.
.PP
NOTE: While any policy can be assigned to a function or category, some
may not be very useful in practice.
For instance: \f[B]rand\f[R] (random) may be useful for file creation
(create) but could lead to very odd behavior if used for \f[C]chmod\f[R]
if there were more than one copy of the file.
.SS Functions and their Category classifications
.PP
.TS
tab(@);
lw(7.4n) lw(62.6n).
T{
Category
T}@T{
FUSE Functions
T}
_
T{
action
T}@T{
chmod, chown, link, removexattr, rename, rmdir, setxattr, truncate,
unlink, utimens
T}
T{
create
T}@T{
create, mkdir, mknod, symlink
T}
T{
search
T}@T{
access, getattr, getxattr, ioctl (directories), listxattr, open,
readlink
T}
T{
N/A
T}@T{
fchmod, fchown, futimens, ftruncate, fallocate, fgetattr, fsync, ioctl
(files), read, readdir, release, statfs, write, copy_file_range
T}
.TE
.PP
In cases where something may be searched for (such as a path to clone)
\f[B]getattr\f[R] will usually be used.
.SS Policies
.PP
A policy is the algorithm used to choose a branch or branches for a
function to work on or generally how the function behaves.
.PP
Any function in the \f[C]create\f[R] category will clone the relative
path if needed.
Some other functions (\f[C]rename\f[R],\f[C]link\f[R],\f[C]ioctl\f[R])
have special requirements or behaviors which you can read more about
below.
.SS Filtering
.PP
Most policies basically search branches and create a list of files /
paths for functions to work on.
The policy is responsible for filtering and sorting the branches.
Filters include \f[B]minfreespace\f[R], whether or not a branch is
mounted read-only, and the branch tagging (RO,NC,RW).
These filters are applied across most policies.
.IP \[bu] 2
No \f[B]search\f[R] function policies filter.
.IP \[bu] 2
All \f[B]action\f[R] function policies filter out branches which are
mounted \f[B]read-only\f[R] or tagged as \f[B]RO (read-only)\f[R].
.IP \[bu] 2
All \f[B]create\f[R] function policies filter out branches which are
mounted \f[B]read-only\f[R], tagged \f[B]RO (read-only)\f[R] or \f[B]NC
(no create)\f[R], or has available space less than
\f[C]minfreespace\f[R].
.PP
Policies may have their own additional filtering such as those that
require existing paths to be present.
.PP
If all branches are filtered an error will be returned.
Typically \f[B]EROFS\f[R] (read-only filesystem) or \f[B]ENOSPC\f[R] (no
space left on device) depending on the most recent reason for filtering
a branch.
\f[B]ENOENT\f[R] will be returned if no eligible branch is found.
.PP
If \f[B]create\f[R], \f[B]mkdir\f[R], \f[B]mknod\f[R], or
\f[B]symlink\f[R] fail with \f[C]EROFS\f[R] or other fundimental errors
then mergerfs will mark any branch found to be read-only as such (IE
will set the mode \f[C]RO\f[R]) and will rerun the policy and try again.
This is mostly for \f[C]ext4\f[R] filesystems that can suddenly become
read-only when it encounters an error.
.SS Path Preservation
.PP
Policies, as described below, are of two basic classifications.
\f[C]path preserving\f[R] and \f[C]non-path preserving\f[R].
.PP
All policies which start with \f[C]ep\f[R] (\f[B]epff\f[R],
\f[B]eplfs\f[R], \f[B]eplus\f[R], \f[B]epmfs\f[R], \f[B]eprand\f[R]) are
\f[C]path preserving\f[R].
\f[C]ep\f[R] stands for \f[C]existing path\f[R].
.PP
A path preserving policy will only consider branches where the relative
path being accessed already exists.
.PP
When using non-path preserving policies paths will be cloned to target
branches as necessary.
.PP
With the \f[C]msp\f[R] or \f[C]most shared path\f[R] policies they are
defined as \f[C]path preserving\f[R] for the purpose of controlling
\f[C]link\f[R] and \f[C]rename\f[R]\[cq]s behaviors since
\f[C]ignorepponrename\f[R] is available to disable that behavior.
.SS Policy descriptions
.PP
A policy\[cq]s behavior differs, as mentioned above, based on the
function it is used with.
Sometimes it really might not make sense to even offer certain policies
because they are literally the same as others but it makes things a bit
more uniform.
.PP
.TS
tab(@);
lw(16.2n) lw(53.8n).
T{
Policy
T}@T{
Description
T}
_
T{
all
T}@T{
Search: For \f[B]mkdir\f[R], \f[B]mknod\f[R], and \f[B]symlink\f[R] it
will apply to all branches.
\f[B]create\f[R] works like \f[B]ff\f[R].
T}
T{
epall (existing path, all)
T}@T{
For \f[B]mkdir\f[R], \f[B]mknod\f[R], and \f[B]symlink\f[R] it will
apply to all found.
\f[B]create\f[R] works like \f[B]epff\f[R] (but more expensive because
it doesn\[cq]t stop after finding a valid branch).
T}
T{
epff (existing path, first found)
T}@T{
Given the order of the branches, as defined at mount time or configured
at runtime, act on the first one found where the relative path exists.
T}
T{
eplfs (existing path, least free space)
T}@T{
Of all the branches on which the relative path exists choose the branch
with the least free space.
T}
T{
eplus (existing path, least used space)
T}@T{
Of all the branches on which the relative path exists choose the branch
with the least used space.
T}
T{
epmfs (existing path, most free space)
T}@T{
Of all the branches on which the relative path exists choose the branch
with the most free space.
T}
T{
eppfrd (existing path, percentage free random distribution)
T}@T{
Like \f[B]pfrd\f[R] but limited to existing paths.
T}
T{
eprand (existing path, random)
T}@T{
Calls \f[B]epall\f[R] and then randomizes.
Returns 1.
T}
T{
ff (first found)
T}@T{
Given the order of the branches, as defined at mount time or configured
at runtime, act on the first one found.
T}
T{
lfs (least free space)
T}@T{
Pick the branch with the least available free space.
T}
T{
lus (least used space)
T}@T{
Pick the branch with the least used space.
T}
T{
mfs (most free space)
T}@T{
Pick the branch with the most available free space.
T}
T{
msplfs (most shared path, least free space)
T}@T{
Like \f[B]eplfs\f[R] but if it fails to find a branch it will try again
with the parent directory.
Continues this pattern till finding one.
T}
T{
msplus (most shared path, least used space)
T}@T{
Like \f[B]eplus\f[R] but if it fails to find a branch it will try again
with the parent directory.
Continues this pattern till finding one.
T}
T{
mspmfs (most shared path, most free space)
T}@T{
Like \f[B]epmfs\f[R] but if it fails to find a branch it will try again
with the parent directory.
Continues this pattern till finding one.
T}
T{
msppfrd (most shared path, percentage free random distribution)
T}@T{
Like \f[B]eppfrd\f[R] but if it fails to find a branch it will try again
with the parent directory.
Continues this pattern till finding one.
T}
T{
newest
T}@T{
Pick the file / directory with the largest mtime.
T}
T{
pfrd (percentage free random distribution)
T}@T{
Chooses a branch at random with the likelihood of selection based on a
branch\[cq]s available space relative to the total.
T}
T{
rand (random)
T}@T{
Calls \f[B]all\f[R] and then randomizes.
Returns 1 branch.
T}
.TE
.PP
\f[B]NOTE:\f[R] If you are using an underlying filesystem that reserves
blocks such as ext2, ext3, or ext4 be aware that mergerfs respects the
reservation by using \f[C]f_bavail\f[R] (number of free blocks for
unprivileged users) rather than \f[C]f_bfree\f[R] (number of free
blocks) in policy calculations.
\f[B]df\f[R] does NOT use \f[C]f_bavail\f[R], it uses \f[C]f_bfree\f[R],
so direct comparisons between \f[B]df\f[R] output and mergerfs\[cq]
policies is not appropriate.
.SS Defaults
.PP
.TS
tab(@);
l l.
T{
Category
T}@T{
Policy
T}
_
T{
action
T}@T{
epall
T}
T{
create
T}@T{
epmfs
T}
T{
search
T}@T{
ff
T}
.TE
.SS func.readdir
.PP
examples: \f[C]func.readdir=seq\f[R], \f[C]func.readdir=cor:4\f[R]
.PP
\f[C]readdir\f[R] has policies to control how it manages reading
directory content.
.PP
.TS
tab(@);
lw(26.7n) lw(43.3n).
T{
Policy
T}@T{
Description
T}
_
T{
seq
T}@T{
\[lq]sequential\[rq] : Iterate over branches in the order defined.
This is the default and traditional behavior found prior to the readdir
policy introduction.
T}
T{
cosr
T}@T{
\[lq]concurrent open, sequential read\[rq] : Concurrently open branch
directories using a thread pool and process them in order of definition.
This keeps memory and CPU usage low while also reducing the time spent
waiting on branches to respond.
Number of threads defaults to the number of logical cores.
Can be overwritten via the syntax \f[C]func.readdir=cosr:N\f[R] where
\f[C]N\f[R] is the number of threads.
T}
T{
cor
T}@T{
\[lq]concurrent open and read\[rq] : Concurrently open branch
directories and immediately start reading their contents using a thread
pool.
This will result in slightly higher memory and CPU usage but reduced
latency.
Particularly when using higher latency / slower speed network filesystem
branches.
Unlike \f[C]seq\f[R] and \f[C]cosr\f[R] the order of files could change
due the async nature of the thread pool.
Number of threads defaults to the number of logical cores.
Can be overwritten via the syntax \f[C]func.readdir=cor:N\f[R] where
\f[C]N\f[R] is the number of threads.
T}
.TE
.PP
Keep in mind that \f[C]readdir\f[R] mostly just provides a list of file
names in a directory and possibly some basic metadata about said files.
To know details about the files, as one would see from commands like
\f[C]find\f[R] or \f[C]ls\f[R], it is required to call \f[C]stat\f[R] on
the file which is controlled by \f[C]fuse.getattr\f[R].
.SS ioctl
.PP
When \f[C]ioctl\f[R] is used with an open file then it will use the file
handle which was created at the original \f[C]open\f[R] call.
However, when using \f[C]ioctl\f[R] with a directory mergerfs will use
the \f[C]open\f[R] policy to find the directory to act on.
.SS rename & link
.PP
\f[B]NOTE:\f[R] If you\[cq]re receiving errors from software when files
are moved / renamed / linked then you should consider changing the
create policy to one which is \f[B]not\f[R] path preserving, enabling
\f[C]ignorepponrename\f[R], or contacting the author of the offending
software and requesting that \f[C]EXDEV\f[R] (cross device / improper
link) be properly handled.
.PP
\f[C]rename\f[R] and \f[C]link\f[R] are tricky functions in a union
filesystem.
\f[C]rename\f[R] only works within a single filesystem or device.
If a rename can\[cq]t be done atomically due to the source and
destination paths existing on different mount points it will return
\f[B]-1\f[R] with \f[B]errno = EXDEV\f[R] (cross device / improper
link).
So if a \f[C]rename\f[R]\[cq]s source and target are on different
filesystems within the pool it creates an issue.
.PP
Originally mergerfs would return EXDEV whenever a rename was requested
which was cross directory in any way.
This made the code simple and was technically compliant with POSIX
requirements.
However, many applications fail to handle EXDEV at all and treat it as a
normal error or otherwise handle it poorly.
Such apps include: gvfsd-fuse v1.20.3 and prior, Finder / CIFS/SMB
client in Apple OSX 10.9+, NZBGet, Samba\[cq]s recycling bin feature.
.PP
As a result a compromise was made in order to get most software to work
while still obeying mergerfs\[cq] policies.
Below is the basic logic.
.IP \[bu] 2
If using a \f[B]create\f[R] policy which tries to preserve directory
paths (epff,eplfs,eplus,epmfs)
.RS 2
.IP \[bu] 2
Using the \f[B]rename\f[R] policy get the list of files to rename
.IP \[bu] 2
For each file attempt rename:
.RS 2
.IP \[bu] 2
If failure with ENOENT (no such file or directory) run \f[B]create\f[R]
policy
.IP \[bu] 2
If create policy returns the same branch as currently evaluating then
clone the path
.IP \[bu] 2
Re-attempt rename
.RE
.IP \[bu] 2
If \f[B]any\f[R] of the renames succeed the higher level rename is
considered a success
.IP \[bu] 2
If \f[B]no\f[R] renames succeed the first error encountered will be
returned
.IP \[bu] 2
On success:
.RS 2
.IP \[bu] 2
Remove the target from all branches with no source file
.IP \[bu] 2
Remove the source from all branches which failed to rename
.RE
.RE
.IP \[bu] 2
If using a \f[B]create\f[R] policy which does \f[B]not\f[R] try to
preserve directory paths
.RS 2
.IP \[bu] 2
Using the \f[B]rename\f[R] policy get the list of files to rename
.IP \[bu] 2
Using the \f[B]getattr\f[R] policy get the target path
.IP \[bu] 2
For each file attempt rename:
.RS 2
.IP \[bu] 2
If the source branch != target branch:
.RS 2
.IP \[bu] 2
Clone target path from target branch to source branch
.RE
.IP \[bu] 2
Rename
.RE
.IP \[bu] 2
If \f[B]any\f[R] of the renames succeed the higher level rename is
considered a success
.IP \[bu] 2
If \f[B]no\f[R] renames succeed the first error encountered will be
returned
.IP \[bu] 2
On success:
.RS 2
.IP \[bu] 2
Remove the target from all branches with no source file
.IP \[bu] 2
Remove the source from all branches which failed to rename
.RE
.RE
.PP
The the removals are subject to normal entitlement checks.
.PP
The above behavior will help minimize the likelihood of EXDEV being
returned but it will still be possible.
.PP
\f[B]link\f[R] uses the same strategy but without the removals.
.SS statfs / statvfs
.PP
statvfs (http://linux.die.net/man/2/statvfs) normalizes the source
filesystems based on the fragment size and sums the number of adjusted
blocks and inodes.
This means you will see the combined space of all sources.
Total, used, and free.
The sources however are dedupped based on the filesystem so multiple
sources on the same drive will not result in double counting its space.
Other filesystems mounted further down the tree of the branch will not
be included when checking the mount\[cq]s stats.
.PP
The options \f[C]statfs\f[R] and \f[C]statfs_ignore\f[R] can be used to
modify \f[C]statfs\f[R] behavior.
.SS flush-on-close
.PP
https://lkml.kernel.org/linux-fsdevel/20211024132607.1636952-1-amir73il\[at]gmail.com/T/
.PP
By default FUSE would issue a flush before the release of a file
descriptor.
This was considered a bit aggressive and a feature added to give the
FUSE server the ability to choose when that happens.
.PP
Options: * always * never * opened-for-write
.PP
For now it defaults to \[lq]opened-for-write\[rq] which is less
aggressive than the behavior before this feature was added.
It should not be a problem because the flush is really only relevant
when a file is written to.
Given flush is irrelevant for many filesystems in the future a branch
specific flag may be added so only files opened on a specific branch
would be flushed on close.
.SH ERROR HANDLING
.PP
POSIX filesystem functions offer a single return code meaning that there
is some complication regarding the handling of multiple branches as
mergerfs does.
It tries to handle errors in a way that would generally return
meaningful values for that particular function.
.SS chmod, chown, removexattr, setxattr, truncate, utimens
.IP "1)" 3
if no error: return 0 (success)
.IP "2)" 3
if no successes: return first error
.IP "3)" 3
if one of the files acted on was the same as the related search
function: return its value
.IP "4)" 3
return 0 (success)
.PP
While doing this increases the complexity and cost of error handling,
particularly step 3, this provides probably the most reasonable return
value.
.SS unlink, rmdir
.IP "1)" 3
if no errors: return 0 (success)
.IP "2)" 3
return first error
.PP
Older version of mergerfs would return success if any success occurred
but for unlink and rmdir there are downstream assumptions that, while
not impossible to occur, can confuse some software.
.SS others
.PP
For search functions there is always a single thing acted on and as such
whatever return value that comes from the single function call is
returned.
.PP
For create functions \f[C]mkdir\f[R], \f[C]mknod\f[R], and
\f[C]symlink\f[R] which don\[cq]t return a file descriptor and therefore
can have \f[C]all\f[R] or \f[C]epall\f[R] policies it will return
success if any of the calls succeed and an error otherwise.
.SH INSTALL
.PP
https://github.com/trapexit/mergerfs/releases
.PP
If your distribution\[cq]s package manager includes mergerfs check if
the version is up to date.
If out of date it is recommended to use the latest release found on the
release page.
Details for common distros are below.
.SS Debian
.PP
Most Debian installs are of a stable branch and therefore do not have
the most up to date software.
While mergerfs is available via \f[C]apt\f[R] it is suggested that uses
install the most recent version available from the releases
page (https://github.com/trapexit/mergerfs/releases).
.SS prebuilt deb
.IP
.nf
\f[C]
wget https://github.com/trapexit/mergerfs/releases/download/<ver>/mergerfs_<ver>.debian-<rel>_<arch>.deb
dpkg -i mergerfs_<ver>.debian-<rel>_<arch>.deb
\f[R]
.fi
.SS apt
.IP
.nf
\f[C]
sudo apt install -y mergerfs
\f[R]
.fi
.SS Ubuntu
.PP
Most Ubuntu installs are of a stable branch and therefore do not have
the most up to date software.
While mergerfs is available via \f[C]apt\f[R] it is suggested that uses
install the most recent version available from the releases
page (https://github.com/trapexit/mergerfs/releases).
.SS prebuilt deb
.IP
.nf
\f[C]
wget https://github.com/trapexit/mergerfs/releases/download/<version>/mergerfs_<ver>.ubuntu-<rel>_<arch>.deb
dpkg -i mergerfs_<ver>.ubuntu-<rel>_<arch>.deb
\f[R]
.fi
.SS apt
.IP
.nf
\f[C]
sudo apt install -y mergerfs
\f[R]
.fi
.SS Raspberry Pi OS
.PP
Effectively the same as Debian or Ubuntu.
.SS Fedora
.IP
.nf
\f[C]
wget https://github.com/trapexit/mergerfs/releases/download/<ver>/mergerfs-<ver>.fc<rel>.<arch>.rpm
sudo rpm -i mergerfs-<ver>.fc<rel>.<arch>.rpm
\f[R]
.fi
.SS CentOS / Rocky
.IP
.nf
\f[C]
wget https://github.com/trapexit/mergerfs/releases/download/<ver>/mergerfs-<ver>.el<rel>.<arch>.rpm
sudo rpm -i mergerfs-<ver>.el<rel>.<arch>.rpm
\f[R]
.fi
.SS ArchLinux
.IP "1." 3
Setup AUR
.IP "2." 3
Install \f[C]mergerfs\f[R]
.SS Other
.PP
Static binaries are provided for situations where native packages are
unavailable.
.IP
.nf
\f[C]
wget https://github.com/trapexit/mergerfs/releases/download/<ver>/mergerfs-static-linux_<arch>.tar.gz
sudo tar xvf mergerfs-static-linux_<arch>.tar.gz -C /
\f[R]
.fi
.SH BUILD
.PP
\f[B]NOTE:\f[R] Prebuilt packages can be found at and recommended for
most users: https://github.com/trapexit/mergerfs/releases
.PP
\f[B]NOTE:\f[R] Only tagged releases are supported.
\f[C]master\f[R] and other branches should be considered works in
progress.
.PP
First get the code from github (https://github.com/trapexit/mergerfs).
.IP
.nf
\f[C]
$ git clone https://github.com/trapexit/mergerfs.git
$ # or
$ wget https://github.com/trapexit/mergerfs/releases/download/<ver>/mergerfs-<ver>.tar.gz
\f[R]
.fi
.SS Debian / Ubuntu
.IP
.nf
\f[C]
$ cd mergerfs
$ sudo tools/install-build-pkgs
$ make deb
$ sudo dpkg -i ../mergerfs_<version>_<arch>.deb
\f[R]
.fi
.SS RHEL / CentOS / Rocky / Fedora
.IP
.nf
\f[C]
$ su -
# cd mergerfs
# tools/install-build-pkgs
# make rpm
# rpm -i rpmbuild/RPMS/<arch>/mergerfs-<version>.<arch>.rpm
\f[R]
.fi
.SS Generic
.PP
Have git, g++, make, python installed.
.IP
.nf
\f[C]
$ cd mergerfs
$ make
$ sudo make install
\f[R]
.fi
.SS Build options
.IP
.nf
\f[C]
$ make help
usage: make

make USE_XATTR=0      - build program without xattrs functionality
make STATIC=1         - build static binary
make LTO=1            - build with link time optimization
\f[R]
.fi
.SH UPGRADE
.PP
mergerfs can be upgraded live by mounting on top of the previous
instance.
Simply install the new version of mergerfs and follow the instructions
below.
.PP
Run mergerfs again or if using \f[C]/etc/fstab\f[R] call for it to mount
again.
Existing open files and such will continue to work fine though they
won\[cq]t see runtime changes since any such change would be the new
mount.
If you plan on changing settings with the new mount you should / could
apply those before mounting the new version.
.IP
.nf
\f[C]
$ sudo mount /mnt/mergerfs
$ mount | grep mergerfs
media on /mnt/mergerfs type fuse.mergerfs (rw,relatime,user_id=0,group_id=0,default_permissions,allow_other)
media on /mnt/mergerfs type fuse.mergerfs (rw,relatime,user_id=0,group_id=0,default_permissions,allow_other)
\f[R]
.fi
.PP
A problem with this approach is that the underlying instance will
continue to run even if the software using it stop or are restarted.
To work around this you can use a \[lq]lazy umount\[rq].
Before mounting over top the mount point with the new instance of
mergerfs issue: \f[C]umount -l <mergerfs_mountpoint>\f[R].
Or you can let mergerfs do it by setting the option
\f[C]lazy-umount-mountpoint=true\f[R].
.SH RUNTIME INTERFACES
.SS RUNTIME CONFIG
.SS .mergerfs pseudo file
.IP
.nf
\f[C]
<mountpoint>/.mergerfs
\f[R]
.fi
.PP
There is a pseudo file available at the mount point which allows for the
runtime modification of certain \f[B]mergerfs\f[R] options.
The file will not show up in \f[B]readdir\f[R] but can be
\f[B]stat\f[R]\[cq]ed and manipulated via
{list,get,set}xattrs (http://linux.die.net/man/2/listxattr) calls.
.PP
Any changes made at runtime are \f[B]not\f[R] persisted.
If you wish for values to persist they must be included as options
wherever you configure the mounting of mergerfs (/etc/fstab).
.SS Keys
.PP
Use \f[C]getfattr -d /mountpoint/.mergerfs\f[R] or
\f[C]xattr -l /mountpoint/.mergerfs\f[R] to see all supported keys.
Some are informational and therefore read-only.
\f[C]setxattr\f[R] will return EINVAL (invalid argument) on read-only
keys.
.SS Values
.PP
Same as the command line.
.SS user.mergerfs.branches
.PP
Used to query or modify the list of branches.
When modifying there are several shortcuts to easy manipulation of the
list.
.PP
.TS
tab(@);
l l.
T{
Value
T}@T{
Description
T}
_
T{
[list]
T}@T{
set
T}
T{
+<[list]
T}@T{
prepend
T}
T{
+>[list]
T}@T{
append
T}
T{
-[list]
T}@T{
remove all values provided
T}
T{
-<
T}@T{
remove first in list
T}
T{
->
T}@T{
remove last in list
T}
.TE
.PP
\f[C]xattr -w user.mergerfs.branches +</mnt/drive3 /mnt/pool/.mergerfs\f[R]
.PP
The \f[C]=NC\f[R], \f[C]=RO\f[R], \f[C]=RW\f[R] syntax works just as on
the command line.
.SS Example
.IP
.nf
\f[C]
[trapexit:/mnt/mergerfs] $ getfattr -d .mergerfs
user.mergerfs.branches=\[dq]/mnt/a=RW:/mnt/b=RW\[dq]
user.mergerfs.minfreespace=\[dq]4294967295\[dq]
user.mergerfs.moveonenospc=\[dq]false\[dq]
\&...

[trapexit:/mnt/mergerfs] $ getfattr -n user.mergerfs.category.search .mergerfs
user.mergerfs.category.search=\[dq]ff\[dq]

[trapexit:/mnt/mergerfs] $ setfattr -n user.mergerfs.category.search -v newest .mergerfs
[trapexit:/mnt/mergerfs] $ getfattr -n user.mergerfs.category.search .mergerfs
user.mergerfs.category.search=\[dq]newest\[dq]
\f[R]
.fi
.SS file / directory xattrs
.PP
While they won\[cq]t show up when using \f[C]getfattr\f[R]
\f[B]mergerfs\f[R] offers a number of special xattrs to query
information about the files served.
To access the values you will need to issue a
getxattr (http://linux.die.net/man/2/getxattr) for one of the following:
.IP \[bu] 2
\f[B]user.mergerfs.basepath\f[R]: the base mount point for the file
given the current getattr policy
.IP \[bu] 2
\f[B]user.mergerfs.relpath\f[R]: the relative path of the file from the
perspective of the mount point
.IP \[bu] 2
\f[B]user.mergerfs.fullpath\f[R]: the full path of the original file
given the getattr policy
.IP \[bu] 2
\f[B]user.mergerfs.allpaths\f[R]: a NUL (`\[rs]0') separated list of
full paths to all files found
.SS SIGNALS
.IP \[bu] 2
USR1: This will cause mergerfs to send invalidation notifications to the
kernel for all files.
This will cause all unused files to be released from memory.
.IP \[bu] 2
USR2: Trigger a general cleanup of currently unused memory.
A more thorough version of what happens every \[ti]15 minutes.
.SS IOCTLS
.PP
Found in \f[C]fuse_ioctl.cpp\f[R]:
.IP
.nf
\f[C]
typedef char IOCTL_BUF[4096];
#define IOCTL_APP_TYPE  0xDF
#define IOCTL_FILE_INFO            _IOWR(IOCTL_APP_TYPE,0,IOCTL_BUF)
#define IOCTL_GC                   _IO(IOCTL_APP_TYPE,1)
#define IOCTL_GC1                  _IO(IOCTL_APP_TYPE,2)
#define IOCTL_INVALIDATE_ALL_NODES _IO(IOCTL_APP_TYPE,3)
\f[R]
.fi
.IP \[bu] 2
IOCTL_FILE_INFO: Same as the \[lq]file / directory xattrs\[rq] mentioned
above.
Use a buffer size of 4096 bytes.
Pass in a string of \[lq]basepath\[rq], \[lq]relpath\[rq],
\[lq]fullpath\[rq], or \[lq]allpaths\[rq].
Receive details in same buffer.
.IP \[bu] 2
IOCTL_GC: Triggers a thorough garbage collection of excess memory.
Same as SIGUSR2.
.IP \[bu] 2
IOCTL_GC1: Triggers a simple garbage collection of excess memory.
Same as what happens every 15 minutes normally.
.IP \[bu] 2
IOCTL_INVALIDATE_ALL_NODES: Same as SIGUSR1.
Send invalidation notifications to the kernel for all files causing
unused files to be released from memory.
.SH TOOLING
.SS preload.so
.PP
EXPERIMENTAL
.PP
For some time there has been work to enable passthrough IO in FUSE.
Passthrough IO would allow for near native performance with regards to
reads and writes (at the expense of certain mergerfs features.) However,
there have been several complications which have kept the feature from
making it into the mainline Linux kernel.
Until that feature is available there are two methods to provide similar
functionality.
One method is using the LD_PRELOAD feature of the dynamic linker.
The other leveraging ptrace to intercept syscalls.
Each has their disadvantages.
At the moment only a preload based tool is available.
A ptrace based tool may be developed later if there is a need.
.PP
\f[C]/usr/lib/mergerfs/preload.so\f[R]
.PP
This preloadable
library (https://man7.org/linux/man-pages/man8/ld.so.8.html#ENVIRONMENT)
overrides the creation and opening of files in order to simulate
passthrough file IO.
It catches the open/creat/fopen calls, has mergerfs do the call, queries
mergerfs for the branch the file exists on, reopens the file on the
underlying filesystem and returns that instead.
Meaning that you will get native read/write performance because mergerfs
is no longer part of the workflow.
Keep in mind that this also means certain mergerfs features that work by
interrupting the read/write workflow, such as \f[C]moveonenospc\f[R],
will no longer work.
.PP
Also understand that this will only work on dynamically linked software.
Anything statically compiled will not work.
Many GoLang and Rust apps are statically compiled.
.PP
The library will not interfere with non-mergerfs filesystems.
The library is written to always fallback to returning the mergerfs
opened file on error.
.PP
While the library was written to account for a number of edgecases there
could be some yet accounted for so please report any oddities.
.PP
Thank you to nohajc (https://github.com/nohajc/mergerfs-io-passthrough)
for prototyping the idea.
.SS general usage
.IP
.nf
\f[C]
LD_PRELOAD=/usr/lib/mergerfs/preload.so touch /mnt/mergerfs/filename
\f[R]
.fi
.SS Docker usage
.PP
Assume \f[C]/mnt/fs0\f[R] and \f[C]/mnt/fs1\f[R] are pooled with
mergerfs at \f[C]/media\f[R].
.PP
All mergerfs branch paths \f[I]must\f[R] be bind mounted into the
container at the same path as found on the host so the preload library
can see them.
.IP
.nf
\f[C]
docker run \[rs]
  -e LD_PRELOAD=/usr/lib/mergerfs/preload.so \[rs]
  -v /usr/lib/mergerfs/preload.so:/usr/lib/mergerfs/preload.so:ro \[rs]
  -v /media:/data \[rs]
  -v /mnt:/mnt \[rs]
  ubuntu:latest \[rs]
  bash
\f[R]
.fi
.PP
or more explicitly
.IP
.nf
\f[C]
docker run \[rs]
  -e LD_PRELOAD=/usr/lib/mergerfs/preload.so \[rs]
  -v /usr/lib/mergerfs/preload.so:/usr/lib/mergerfs/preload.so:ro \[rs]
  -v /media:/data \[rs]
  -v /mnt/fs0:/mnt/fs0 \[rs]
  -v /mnt/fs1:/mnt/fs1 \[rs]
  ubuntu:latest \[rs]
  bash
\f[R]
.fi
.SS systemd unit
.PP
Use the \f[C]Environment\f[R] option to set the LD_PRELOAD variable.
.IP \[bu] 2
https://www.freedesktop.org/software/systemd/man/latest/systemd.service.html#Command%20lines
.IP \[bu] 2
https://serverfault.com/questions/413397/how-to-set-environment-variable-in-systemd-service
.IP
.nf
\f[C]
[Service]
Environment=LD_PRELOAD=/usr/lib/mergerfs/preload.so
\f[R]
.fi
.SS Misc
.IP \[bu] 2
https://github.com/trapexit/mergerfs-tools
.RS 2
.IP \[bu] 2
mergerfs.ctl: A tool to make it easier to query and configure mergerfs
at runtime
.IP \[bu] 2
mergerfs.fsck: Provides permissions and ownership auditing and the
ability to fix them
.IP \[bu] 2
mergerfs.dedup: Will help identify and optionally remove duplicate files
.IP \[bu] 2
mergerfs.dup: Ensure there are at least N copies of a file across the
pool
.IP \[bu] 2
mergerfs.balance: Rebalance files across filesystems by moving them from
the most filled to the least filled
.IP \[bu] 2
mergerfs.consolidate: move files within a single mergerfs directory to
the filesystem with most free space
.RE
.IP \[bu] 2
https://github.com/trapexit/scorch
.RS 2
.IP \[bu] 2
scorch: A tool to help discover silent corruption of files and keep
track of files
.RE
.IP \[bu] 2
https://github.com/trapexit/bbf
.RS 2
.IP \[bu] 2
bbf (bad block finder): a tool to scan for and `fix' hard drive bad
blocks and find the files using those blocks
.RE
.SH CACHING
.SS page caching
.PP
https://en.wikipedia.org/wiki/Page_cache
.IP \[bu] 2
cache.files=off: Disables page caching.
Underlying files cached, mergerfs files are not.
.IP \[bu] 2
cache.files=partial: Enables page caching.
Underlying files cached, mergerfs files cached while open.
.IP \[bu] 2
cache.files=full: Enables page caching.
Underlying files cached, mergerfs files cached across opens.
.IP \[bu] 2
cache.files=auto-full: Enables page caching.
Underlying files cached, mergerfs files cached across opens if mtime and
size are unchanged since previous open.
.IP \[bu] 2
cache.files=libfuse: follow traditional libfuse \f[C]direct_io\f[R],
\f[C]kernel_cache\f[R], and \f[C]auto_cache\f[R] arguments.
.IP \[bu] 2
cache.files=per-process: Enable page caching (equivalent to
\f[C]cache.files=partial\f[R]) only for processes whose `comm' name
matches one of the values defined in
\f[C]cache.files.process-names\f[R].
If the name does not match the file open is equivalent to
\f[C]cache.files=off\f[R].
.PP
FUSE, which mergerfs uses, offers a number of page caching modes.
mergerfs tries to simplify their use via the \f[C]cache.files\f[R]
option.
It can and should replace usage of \f[C]direct_io\f[R],
\f[C]kernel_cache\f[R], and \f[C]auto_cache\f[R].
.PP
Due to mergerfs using FUSE and therefore being a userland process
proxying existing filesystems the kernel will double cache the content
being read and written through mergerfs.
Once from the underlying filesystem and once from mergerfs (it sees them
as two separate entities).
Using \f[C]cache.files=off\f[R] will keep the double caching from
happening by disabling caching of mergerfs but this has the side effect
that \f[I]all\f[R] read and write calls will be passed to mergerfs which
may be slower than enabling caching, you lose shared \f[C]mmap\f[R]
support which can affect apps such as rtorrent, and no read-ahead will
take place.
The kernel will still cache the underlying filesystem data but that only
helps so much given mergerfs will still process all requests.
.PP
If you do enable file page caching,
\f[C]cache.files=partial|full|auto-full\f[R], you should also enable
\f[C]dropcacheonclose\f[R] which will cause mergerfs to instruct the
kernel to flush the underlying file\[cq]s page cache when the file is
closed.
This behavior is the same as the rsync fadvise / drop cache patch and
Feh\[cq]s nocache project.
.PP
If most files are read once through and closed (like media) it is best
to enable \f[C]dropcacheonclose\f[R] regardless of caching mode in order
to minimize buffer bloat.
.PP
It is difficult to balance memory usage, cache bloat & duplication, and
performance.
Ideally mergerfs would be able to disable caching for the files it
reads/writes but allow page caching for itself.
That would limit the FUSE overhead.
However, there isn\[cq]t a good way to achieve this.
It would need to open all files with O_DIRECT which places limitations
on the what underlying filesystems would be supported and complicates
the code.
.PP
kernel documentation:
https://www.kernel.org/doc/Documentation/filesystems/fuse-io.txt
.SS entry & attribute caching
.PP
Given the relatively high cost of FUSE due to the kernel <-> userspace
round trips there are kernel side caches for file entries and
attributes.
The entry cache limits the \f[C]lookup\f[R] calls to mergerfs which ask
if a file exists.
The attribute cache limits the need to make \f[C]getattr\f[R] calls to
mergerfs which provide file attributes (mode, size, type, etc.).
As with the page cache these should not be used if the underlying
filesystems are being manipulated at the same time as it could lead to
odd behavior or data corruption.
The options for setting these are \f[C]cache.entry\f[R] and
\f[C]cache.negative_entry\f[R] for the entry cache and
\f[C]cache.attr\f[R] for the attributes cache.
\f[C]cache.negative_entry\f[R] refers to the timeout for negative
responses to lookups (non-existent files).
.SS writeback caching
.PP
When \f[C]cache.files\f[R] is enabled the default is for it to perform
writethrough caching.
This behavior won\[cq]t help improve performance as each write still
goes one for one through the filesystem.
By enabling the FUSE writeback cache small writes may be aggregated by
the kernel and then sent to mergerfs as one larger request.
This can greatly improve the throughput for apps which write to files
inefficiently.
The amount the kernel can aggregate is limited by the size of a FUSE
message.
Read the \f[C]fuse_msg_size\f[R] section for more details.
.PP
There is a small side effect as a result of enabling writeback caching.
Underlying files won\[cq]t ever be opened with O_APPEND or O_WRONLY.
The former because the kernel then manages append mode and the latter
because the kernel may request file data from mergerfs to populate the
write cache.
The O_APPEND change means that if a file is changed outside of mergerfs
it could lead to corruption as the kernel won\[cq]t know the end of the
file has changed.
That said any time you use caching you should keep from using the same
file outside of mergerfs at the same time.
.PP
Note that if an application is properly sizing writes then writeback
caching will have little or no effect.
It will only help with writes of sizes below the FUSE message size (128K
on older kernels, 1M on newer).
.SS statfs caching
.PP
Of the syscalls used by mergerfs in policies the \f[C]statfs\f[R] /
\f[C]statvfs\f[R] call is perhaps the most expensive.
It\[cq]s used to find out the available space of a filesystem and
whether it is mounted read-only.
Depending on the setup and usage pattern these queries can be relatively
costly.
When \f[C]cache.statfs\f[R] is enabled all calls to \f[C]statfs\f[R] by
a policy will be cached for the number of seconds its set to.
.PP
Example: If the create policy is \f[C]mfs\f[R] and the timeout is 60
then for that 60 seconds the same filesystem will be returned as the
target for creates because the available space won\[cq]t be updated for
that time.
.SS symlink caching
.PP
As of version 4.20 Linux supports symlink caching.
Significant performance increases can be had in workloads which use a
lot of symlinks.
Setting \f[C]cache.symlinks=true\f[R] will result in requesting symlink
caching from the kernel only if supported.
As a result its safe to enable it on systems prior to 4.20.
That said it is disabled by default for now.
You can see if caching is enabled by querying the xattr
\f[C]user.mergerfs.cache.symlinks\f[R] but given it must be requested at
startup you can not change it at runtime.
.SS readdir caching
.PP
As of version 4.20 Linux supports readdir caching.
This can have a significant impact on directory traversal.
Especially when combined with entry (\f[C]cache.entry\f[R]) and
attribute (\f[C]cache.attr\f[R]) caching.
Setting \f[C]cache.readdir=true\f[R] will result in requesting readdir
caching from the kernel on each \f[C]opendir\f[R].
If the kernel doesn\[cq]t support readdir caching setting the option to
\f[C]true\f[R] has no effect.
This option is configurable at runtime via xattr
\f[C]user.mergerfs.cache.readdir\f[R].
.SS tiered caching
.PP
Some storage technologies support what some call \[lq]tiered\[rq]
caching.
The placing of usually smaller, faster storage as a transparent cache to
larger, slower storage.
NVMe, SSD, Optane in front of traditional HDDs for instance.
.PP
mergerfs does not natively support any sort of tiered caching.
Most users have no use for such a feature and its inclusion would
complicate the code.
However, there are a few situations where a cache filesystem could help
with a typical mergerfs setup.
.IP "1." 3
Fast network, slow filesystems, many readers: You\[cq]ve a 10+Gbps
network with many readers and your regular filesystems can\[cq]t keep
up.
.IP "2." 3
Fast network, slow filesystems, small\[cq]ish bursty writes: You have a
10+Gbps network and wish to transfer amounts of data less than your
cache filesystem but wish to do so quickly.
.PP
With #1 it\[cq]s arguable if you should be using mergerfs at all.
RAID would probably be the better solution.
If you\[cq]re going to use mergerfs there are other tactics that may
help: spreading the data across filesystems (see the mergerfs.dup tool)
and setting \f[C]func.open=rand\f[R], using \f[C]symlinkify\f[R], or
using dm-cache or a similar technology to add tiered cache to the
underlying device.
.PP
With #2 one could use dm-cache as well but there is another solution
which requires only mergerfs and a cronjob.
.IP "1." 3
Create 2 mergerfs pools.
One which includes just the slow devices and one which has both the fast
devices (SSD,NVME,etc.) and slow devices.
.IP "2." 3
The `cache' pool should have the cache filesystems listed first.
.IP "3." 3
The best \f[C]create\f[R] policies to use for the `cache' pool would
probably be \f[C]ff\f[R], \f[C]epff\f[R], \f[C]lfs\f[R], or
\f[C]eplfs\f[R].
The latter two under the assumption that the cache filesystem(s) are far
smaller than the backing filesystems.
If using path preserving policies remember that you\[cq]ll need to
manually create the core directories of those paths you wish to be
cached.
Be sure the permissions are in sync.
Use \f[C]mergerfs.fsck\f[R] to check / correct them.
You could also set the slow filesystems mode to \f[C]NC\f[R] though
that\[cq]d mean if the cache filesystems fill you\[cq]d get \[lq]out of
space\[rq] errors.
.IP "4." 3
Enable \f[C]moveonenospc\f[R] and set \f[C]minfreespace\f[R]
appropriately.
To make sure there is enough room on the \[lq]slow\[rq] pool you might
want to set \f[C]minfreespace\f[R] to at least as large as the size of
the largest cache filesystem if not larger.
This way in the worst case the whole of the cache filesystem(s) can be
moved to the other drives.
.IP "5." 3
Set your programs to use the cache pool.
.IP "6." 3
Save one of the below scripts or create you\[cq]re own.
.IP "7." 3
Use \f[C]cron\f[R] (as root) to schedule the command at whatever
frequency is appropriate for your workflow.
.SS time based expiring
.PP
Move files from cache to backing pool based only on the last time the
file was accessed.
Replace \f[C]-atime\f[R] with \f[C]-amin\f[R] if you want minutes rather
than days.
May want to use the \f[C]fadvise\f[R] / \f[C]--drop-cache\f[R] version
of rsync or run rsync with the tool \[lq]nocache\[rq].
.PP
\f[I]NOTE:\f[R] The arguments to these scripts include the cache
\f[B]filesystem\f[R] itself.
Not the pool with the cache filesystem.
You could have data loss if the source is the cache pool.
.PP
mergerfs.time-based-mover
.SS percentage full expiring
.PP
Move the oldest file from the cache to the backing pool.
Continue till below percentage threshold.
.PP
\f[I]NOTE:\f[R] The arguments to these scripts include the cache
\f[B]filesystem\f[R] itself.
Not the pool with the cache filesystem.
You could have data loss if the source is the cache pool.
.PP
mergerfs.percent-full-mover
.SH PERFORMANCE
.PP
mergerfs is at its core just a proxy and therefore its theoretical max
performance is that of the underlying devices.
However, given it is a FUSE filesystem working from userspace there is
an increase in overhead relative to kernel based solutions.
That said the performance can match the theoretical max but it depends
greatly on the system\[cq]s configuration.
Especially when adding network filesystems into the mix there are many
variables which can impact performance.
Device speeds and latency, network speeds and latency, general
concurrency, read/write sizes, etc.
Unfortunately, given the number of variables it has been difficult to
find a single set of settings which provide optimal performance.
If you\[cq]re having performance issues please look over the suggestions
below (including the benchmarking section.)
.PP
NOTE: be sure to read about these features before changing them to
understand what behaviors it may impact
.IP \[bu] 2
disable \f[C]security_capability\f[R] and/or \f[C]xattr\f[R]
.IP \[bu] 2
increase cache timeouts \f[C]cache.attr\f[R], \f[C]cache.entry\f[R],
\f[C]cache.negative_entry\f[R]
.IP \[bu] 2
enable (or disable) page caching (\f[C]cache.files\f[R])
.IP \[bu] 2
enable \f[C]parallel-direct-writes\f[R]
.IP \[bu] 2
enable \f[C]cache.writeback\f[R]
.IP \[bu] 2
enable \f[C]cache.statfs\f[R]
.IP \[bu] 2
enable \f[C]cache.symlinks\f[R]
.IP \[bu] 2
enable \f[C]cache.readdir\f[R]
.IP \[bu] 2
change the number of worker threads
.IP \[bu] 2
disable \f[C]posix_acl\f[R]
.IP \[bu] 2
disable \f[C]async_read\f[R]
.IP \[bu] 2
test theoretical performance using \f[C]nullrw\f[R] or mounting a ram
disk
.IP \[bu] 2
use \f[C]symlinkify\f[R] if your data is largely static and read-only
.IP \[bu] 2
use tiered cache devices
.IP \[bu] 2
use LVM and LVM cache to place a SSD in front of your HDDs
.IP \[bu] 2
increase readahead: \f[C]readahead=1024\f[R]
.PP
If you come across a setting that significantly impacts performance
please contact trapexit so he may investigate further.
Please test both against your normal setup, a singular branch, and with
\f[C]nullrw=true\f[R]
.SH BENCHMARKING
.PP
Filesystems are complicated.
They do many things and many of those are interconnected.
Additionally, the OS, drivers, hardware, etc.
all can impact performance.
Therefore, when benchmarking, it is \f[B]necessary\f[R] that the test
focus as narrowly as possible.
.PP
For most throughput is the key benchmark.
To test throughput \f[C]dd\f[R] is useful but \f[B]must\f[R] be used
with the correct settings in order to ensure the filesystem or device is
actually being tested.
The OS can and will cache data.
Without forcing synchronous reads and writes and/or disabling caching
the values returned will not be representative of the device\[cq]s true
performance.
.PP
When benchmarking through mergerfs ensure you only use 1 branch to
remove any possibility of the policies complicating the situation.
Benchmark the underlying filesystem first and then mount mergerfs over
it and test again.
If you\[cq]re experience speeds below your expectation you will need to
narrow down precisely which component is leading to the slowdown.
Preferably test the following in the order listed (but not combined).
.IP "1." 3
Enable \f[C]nullrw\f[R] mode with \f[C]nullrw=true\f[R].
This will effectively make reads and writes no-ops.
Removing the underlying device / filesystem from the equation.
This will give us the top theoretical speeds.
.IP "2." 3
Mount mergerfs over \f[C]tmpfs\f[R].
\f[C]tmpfs\f[R] is a RAM disk.
Extremely high speed and very low latency.
This is a more realistic best case scenario.
Example: \f[C]mount -t tmpfs -o size=2G tmpfs /tmp/tmpfs\f[R]
.IP "3." 3
Mount mergerfs over a local device.
NVMe, SSD, HDD, etc.
If you have more than one I\[cq]d suggest testing each of them as drives
and/or controllers (their drivers) could impact performance.
.IP "4." 3
Finally, if you intend to use mergerfs with a network filesystem, either
as the source of data or to combine with another through mergerfs, test
each of those alone as above.
.PP
Once you find the component which has the performance issue you can do
further testing with different options to see if they impact
performance.
For reads and writes the most relevant would be: \f[C]cache.files\f[R],
\f[C]async_read\f[R].
Less likely but relevant when using NFS or with certain filesystems
would be \f[C]security_capability\f[R], \f[C]xattr\f[R], and
\f[C]posix_acl\f[R].
If you find a specific system, device, filesystem, controller, etc.
that performs poorly contact trapexit so he may investigate further.
.PP
Sometimes the problem is really the application accessing or writing
data through mergerfs.
Some software use small buffer sizes which can lead to more requests and
therefore greater overhead.
You can test this out yourself by replace \f[C]bs=1M\f[R] in the
examples below with \f[C]ibs\f[R] or \f[C]obs\f[R] and using a size of
\f[C]512\f[R] instead of \f[C]1M\f[R].
In one example test using \f[C]nullrw\f[R] the write speed dropped from
4.9GB/s to 69.7MB/s when moving from \f[C]1M\f[R] to \f[C]512\f[R].
Similar results were had when testing reads.
Small writes overhead may be improved by leveraging a write cache but in
casual tests little gain was found.
More tests will need to be done before this feature would become
available.
If you have an app that appears slow with mergerfs it could be due to
this.
Contact trapexit so he may investigate further.
.SS write benchmark
.IP
.nf
\f[C]
$ dd if=/dev/zero of=/mnt/mergerfs/1GB.file bs=1M count=1024 oflag=nocache conv=fdatasync status=progress
\f[R]
.fi
.SS read benchmark
.IP
.nf
\f[C]
$ dd if=/mnt/mergerfs/1GB.file of=/dev/null bs=1M count=1024 iflag=nocache conv=fdatasync status=progress
\f[R]
.fi
.SS other benchmarks
.PP
If you are attempting to benchmark other behaviors you must ensure you
clear kernel caches before runs.
In fact it would be a good deal to run before the read and write
benchmarks as well just in case.
.IP
.nf
\f[C]
sync
echo 3 | sudo tee /proc/sys/vm/drop_caches
\f[R]
.fi
.SH TIPS / NOTES
.IP \[bu] 2
This document is literal and thorough.
If a suspected feature isn\[cq]t mentioned it doesn\[cq]t exist.
If certain libfuse arguments aren\[cq]t listed they probably
shouldn\[cq]t be used.
.IP \[bu] 2
Ensure you\[cq]re using the latest version.
.IP \[bu] 2
Run mergerfs as \f[C]root\f[R].
mergerfs is designed and intended to be run as \f[C]root\f[R] and may
exibit incorrect behavior if run otherwise..
.IP \[bu] 2
If you don\[cq]t see some directories and files you expect, policies
seem to skip branches, you get strange permission errors, etc.
be sure the underlying filesystems\[cq] permissions are all the same.
Use \f[C]mergerfs.fsck\f[R] to audit the filesystem for out of sync
permissions.
.IP \[bu] 2
If you still have permission issues be sure you are using POSIX ACL
compliant filesystems.
mergerfs doesn\[cq]t generally make exceptions for FAT, NTFS, or other
non-POSIX filesystem.
.IP \[bu] 2
Do \f[B]not\f[R] use \f[C]cache.files=off\f[R] if you expect
applications (such as rtorrent) to use
mmap (http://linux.die.net/man/2/mmap) files.
Shared mmap is not currently supported in FUSE w/ page caching disabled.
Enabling \f[C]dropcacheonclose\f[R] is recommended when
\f[C]cache.files=partial|full|auto-full\f[R].
.IP \[bu] 2
Kodi (http://kodi.tv), Plex (http://plex.tv),
Subsonic (http://subsonic.org), etc.
can use directory mtime (http://linux.die.net/man/2/stat) to more
efficiently determine whether to scan for new content rather than simply
performing a full scan.
If using the default \f[B]getattr\f[R] policy of \f[B]ff\f[R] it\[cq]s
possible those programs will miss an update on account of it returning
the first directory found\[cq]s \f[B]stat\f[R] info and it\[cq]s a later
directory on another mount which had the \f[B]mtime\f[R] recently
updated.
To fix this you will want to set \f[B]func.getattr=newest\f[R].
Remember though that this is just \f[B]stat\f[R].
If the file is later \f[B]open\f[R]\[cq]ed or \f[B]unlink\f[R]\[cq]ed
and the policy is different for those then a completely different file
or directory could be acted on.
.IP \[bu] 2
Some policies mixed with some functions may result in strange behaviors.
Not that some of these behaviors and race conditions couldn\[cq]t happen
outside \f[B]mergerfs\f[R] but that they are far more likely to occur on
account of the attempt to merge together multiple sources of data which
could be out of sync due to the different policies.
.IP \[bu] 2
For consistency its generally best to set \f[B]category\f[R] wide
policies rather than individual \f[B]func\f[R]\[cq]s.
This will help limit the confusion of tools such as
rsync (http://linux.die.net/man/1/rsync).
However, the flexibility is there if needed.
.SH KNOWN ISSUES / BUGS
.SS kernel issues & bugs
.PP
<https://github.com/trapexit/mergerfs/wiki/Kernel-Issues-&-Bugs>
.SS directory mtime is not being updated
.PP
Remember that the default policy for \f[C]getattr\f[R] is \f[C]ff\f[R].
The information for the first directory found will be returned.
If it wasn\[cq]t the directory which had been updated then it will
appear outdated.
.PP
The reason this is the default is because any other policy would be more
expensive and for many applications it is unnecessary.
To always return the directory with the most recent mtime or a faked
value based on all found would require a scan of all filesystems.
.PP
If you always want the directory information from the one with the most
recent mtime then use the \f[C]newest\f[R] policy for \f[C]getattr\f[R].
.SS `mv /mnt/pool/foo /mnt/disk1/foo' removes `foo'
.PP
This is not a bug.
.PP
Run in verbose mode to better understand what\[cq]s happening:
.IP
.nf
\f[C]
$ mv -v /mnt/pool/foo /mnt/disk1/foo
copied \[aq]/mnt/pool/foo\[aq] -> \[aq]/mnt/disk1/foo\[aq]
removed \[aq]/mnt/pool/foo\[aq]
$ ls /mnt/pool/foo
ls: cannot access \[aq]/mnt/pool/foo\[aq]: No such file or directory
\f[R]
.fi
.PP
\f[C]mv\f[R], when working across devices, is copying the source to
target and then removing the source.
Since the source \f[B]is\f[R] the target in this case, depending on the
unlink policy, it will remove the just copied file and other files
across the branches.
.PP
If you want to move files to one filesystem just copy them there and use
mergerfs.dedup to clean up the old paths or manually remove them from
the branches directly.
.SS cached memory appears greater than it should be
.PP
Use \f[C]cache.files=off\f[R] and/or \f[C]dropcacheonclose=true\f[R].
See the section on page caching.
.SS NFS clients returning ESTALE / Stale file handle
.PP
NFS does not like out of band changes.
That is especially true of inode values.
.PP
Be sure to use the following options:
.IP \[bu] 2
noforget
.IP \[bu] 2
inodecalc=path-hash
.SS rtorrent fails with ENODEV (No such device)
.PP
Be sure to set \f[C]cache.files=partial|full|auto-full|per-processe\f[R]
or turn off \f[C]direct_io\f[R].
rtorrent and some other applications use
mmap (http://linux.die.net/man/2/mmap) to read and write to files and
offer no fallback to traditional methods.
FUSE does not currently support mmap while using \f[C]direct_io\f[R].
There may be a performance penalty on writes with \f[C]direct_io\f[R]
off as well as the problem of double caching but it\[cq]s the only way
to get such applications to work.
If the performance loss is too high for other apps you can mount
mergerfs twice.
Once with \f[C]direct_io\f[R] enabled and one without it.
Be sure to set \f[C]dropcacheonclose=true\f[R] if not using
\f[C]direct_io\f[R].
.SS Plex doesn\[cq]t work with mergerfs
.PP
It does.
If you\[cq]re trying to put Plex\[cq]s config / metadata / database on
mergerfs you can\[cq]t set \f[C]cache.files=off\f[R] because Plex is
using sqlite3 with mmap enabled.
Shared mmap is not supported by Linux\[cq]s FUSE implementation when
page caching is disabled.
To fix this place the data elsewhere (preferable) or enable
\f[C]cache.files\f[R] (with \f[C]dropcacheonclose=true\f[R]).
Sqlite3 does not need mmap but the developer needs to fall back to
standard IO if mmap fails.
.PP
This applies to other software: Radarr, Sonarr, Lidarr, Jellyfin, etc.
.PP
I would recommend reaching out to the developers of the software
you\[cq]re having troubles with and asking them to add a fallback to
regular file IO when mmap is unavailable.
.PP
If the issue is that scanning doesn\[cq]t seem to pick up media then be
sure to set \f[C]func.getattr=newest\f[R] though generally a full scan
will pick up all media anyway.
.SS When a program tries to move or rename a file it fails
.PP
Please read the section above regarding rename & link.
.PP
The problem is that many applications do not properly handle
\f[C]EXDEV\f[R] errors which \f[C]rename\f[R] and \f[C]link\f[R] may
return even though they are perfectly valid situations which do not
indicate actual device, filesystem, or OS errors.
The error will only be returned by mergerfs if using a path preserving
policy as described in the policy section above.
If you do not care about path preservation simply change the mergerfs
policy to the non-path preserving version.
For example: \f[C]-o category.create=mfs\f[R] Ideally the offending
software would be fixed and it is recommended that if you run into this
problem you contact the software\[cq]s author and request proper
handling of \f[C]EXDEV\f[R] errors.
.SS my 32bit software has problems
.PP
Some software have problems with 64bit inode values.
The symptoms can include EOVERFLOW errors when trying to list files.
You can address this by setting \f[C]inodecalc\f[R] to one of the 32bit
based algos as described in the relevant section.
.SS Samba: Moving files / directories fails
.PP
Workaround: Copy the file/directory and then remove the original rather
than move.
.PP
This isn\[cq]t an issue with Samba but some SMB clients.
GVFS-fuse v1.20.3 and prior (found in Ubuntu 14.04 among others) failed
to handle certain error codes correctly.
Particularly \f[B]STATUS_NOT_SAME_DEVICE\f[R] which comes from the
\f[B]EXDEV\f[R] which is returned by \f[B]rename\f[R] when the call is
crossing mount points.
When a program gets an \f[B]EXDEV\f[R] it needs to explicitly take an
alternate action to accomplish its goal.
In the case of \f[B]mv\f[R] or similar it tries \f[B]rename\f[R] and on
\f[B]EXDEV\f[R] falls back to a manual copying of data between the two
locations and unlinking the source.
In these older versions of GVFS-fuse if it received \f[B]EXDEV\f[R] it
would translate that into \f[B]EIO\f[R].
This would cause \f[B]mv\f[R] or most any application attempting to move
files around on that SMB share to fail with a IO error.
.PP
GVFS-fuse v1.22.0 (https://bugzilla.gnome.org/show_bug.cgi?id=734568)
and above fixed this issue but a large number of systems use the older
release.
On Ubuntu the version can be checked by issuing
\f[C]apt-cache showpkg gvfs-fuse\f[R].
Most distros released in 2015 seem to have the updated release and will
work fine but older systems may not.
Upgrading gvfs-fuse or the distro in general will address the problem.
.PP
In Apple\[cq]s MacOSX 10.9 they replaced Samba (client and server) with
their own product.
It appears their new client does not handle \f[B]EXDEV\f[R] either and
responds similar to older release of gvfs on Linux.
.SS Trashing files occasionally fails
.PP
This is the same issue as with Samba.
\f[C]rename\f[R] returns \f[C]EXDEV\f[R] (in our case that will really
only happen with path preserving policies like \f[C]epmfs\f[R]) and the
software doesn\[cq]t handle the situation well.
This is unfortunately a common failure of software which moves files
around.
The standard indicates that an implementation \f[C]MAY\f[R] choose to
support non-user home directory trashing of files (which is a
\f[C]MUST\f[R]).
The implementation \f[C]MAY\f[R] also support \[lq]top directory
trashes\[rq] which many probably do.
.PP
To create a \f[C]$topdir/.Trash\f[R] directory as defined in the
standard use the
mergerfs-tools (https://github.com/trapexit/mergerfs-tools) tool
\f[C]mergerfs.mktrash\f[R].
.SS Supplemental user groups
.PP
Due to the overhead of
getgroups/setgroups (http://linux.die.net/man/2/setgroups) mergerfs
utilizes a cache.
This cache is opportunistic and per thread.
Each thread will query the supplemental groups for a user when that
particular thread needs to change credentials and will keep that data
for the lifetime of the thread.
This means that if a user is added to a group it may not be picked up
without the restart of mergerfs.
However, since the high level FUSE API\[cq]s (at least the standard
version) thread pool dynamically grows and shrinks it\[cq]s possible
that over time a thread will be killed and later a new thread with no
cache will start and query the new data.
.PP
The gid cache uses fixed storage to simplify the design and be
compatible with older systems which may not have C++11 compilers.
There is enough storage for 256 users\[cq] supplemental groups.
Each user is allowed up to 32 supplemental groups.
Linux >= 2.6.3 allows up to 65535 groups per user but most other *nixs
allow far less.
NFS allowing only 16.
The system does handle overflow gracefully.
If the user has more than 32 supplemental groups only the first 32 will
be used.
If more than 256 users are using the system when an uncached user is
found it will evict an existing user\[cq]s cache at random.
So long as there aren\[cq]t more than 256 active users this should be
fine.
If either value is too low for your needs you will have to modify
\f[C]gidcache.hpp\f[R] to increase the values.
Note that doing so will increase the memory needed by each thread.
.PP
While not a bug some users have found when using containers that
supplemental groups defined inside the container don\[cq]t work properly
with regard to permissions.
This is expected as mergerfs lives outside the container and therefore
is querying the host\[cq]s group database.
There might be a hack to work around this (make mergerfs read the
/etc/group file in the container) but it is not yet implemented and
would be limited to Linux and the /etc/group DB.
Preferably users would mount in the host group file into the containers
or use a standard shared user & groups technology like NIS or LDAP.
.SH FAQ
.SS How well does mergerfs scale? Is it \[lq]production ready?\[rq]
.PP
Users have reported running mergerfs on everything from a Raspberry Pi
to dual socket Xeon systems with >20 cores.
I\[cq]m aware of at least a few companies which use mergerfs in
production.
Open Media Vault (https://www.openmediavault.org) includes mergerfs as
its sole solution for pooling filesystems.
The author of mergerfs had it running for over 300 days managing 16+
devices with reasonably heavy 24/7 read and write usage.
Stopping only after the machine\[cq]s power supply died.
.PP
Most serious issues (crashes or data corruption) have been due to kernel
bugs (https://github.com/trapexit/mergerfs/wiki/Kernel-Issues-&-Bugs).
All of which are fixed in stable releases.
.SS Can mergerfs be used with filesystems which already have data / are in use?
.PP
Yes.
mergerfs is really just a proxy and does \f[B]NOT\f[R] interfere with
the normal form or function of the filesystems / mounts / paths it
manages.
It is just another userland application that is acting as a
man-in-the-middle.
It can\[cq]t do anything that any other random piece of software
can\[cq]t do.
.PP
mergerfs is \f[B]not\f[R] a traditional filesystem that takes control
over the underlying block device.
mergerfs is \f[B]not\f[R] RAID.
It does \f[B]not\f[R] manipulate the data that passes through it.
It does \f[B]not\f[R] shard data across filesystems.
It merely shards some \f[B]behavior\f[R] and aggregates others.
.SS Can drives/filesystems be removed from the pool at will?
.PP
Yes.
See previous question\[cq]s answer.
.SS Can mergerfs be removed without affecting the data?
.PP
Yes.
See the previous question\[cq]s answer.
.SS Can drives/filesystems be moved to another pool?
.PP
Yes.
See the previous question\[cq]s answer.
.SS How do I migrate data into or out of the pool when adding/removing drives/filesystems?
.PP
You don\[cq]t need to.
See the previous question\[cq]s answer.
.SS How do I remove a drive/filesystem but keep the data in the pool?
.PP
Nothing special needs to be done.
Remove the branch from mergerfs\[cq] config and copy (rsync) the data
from the removed filesystem into the pool.
Effectively the same as if it were you transfering data from one
filesystem to another.
.PP
If you wish to continue using the pool while performing the transfer
simply create another, temporary pool without the filesystem in question
and then copy the data.
It would probably be a good idea to set the branch to \f[C]RO\f[R] prior
to doing this to ensure no new content is written to the filesystem
while performing the copy.
.SS What policies should I use?
.PP
Unless you\[cq]re doing something more niche the average user is
probably best off using \f[C]mfs\f[R] for \f[C]category.create\f[R].
It will spread files out across your branches based on available space.
Use \f[C]mspmfs\f[R] if you want to try to colocate the data a bit more.
You may want to use \f[C]lus\f[R] if you prefer a slightly different
distribution of data if you have a mix of smaller and larger
filesystems.
Generally though \f[C]mfs\f[R], \f[C]lus\f[R], or even \f[C]rand\f[R]
are good for the general use case.
If you are starting with an imbalanced pool you can use the tool
\f[B]mergerfs.balance\f[R] to redistribute files across the pool.
.PP
If you really wish to try to colocate files based on directory you can
set \f[C]func.create\f[R] to \f[C]epmfs\f[R] or similar and
\f[C]func.mkdir\f[R] to \f[C]rand\f[R] or \f[C]eprand\f[R] depending on
if you just want to colocate generally or on specific branches.
Either way the \f[I]need\f[R] to colocate is rare.
For instance: if you wish to remove the device regularly and want the
data to predictably be on that device or if you don\[cq]t use backup at
all and don\[cq]t wish to replace that data piecemeal.
In which case using path preservation can help but will require some
manual attention.
Colocating after the fact can be accomplished using the
\f[B]mergerfs.consolidate\f[R] tool.
If you don\[cq]t need strict colocation which the \f[C]ep\f[R] policies
provide then you can use the \f[C]msp\f[R] based policies which will
walk back the path till finding a branch that works.
.PP
Ultimately there is no correct answer.
It is a preference or based on some particular need.
mergerfs is very easy to test and experiment with.
I suggest creating a test setup and experimenting to get a sense of what
you want.
.PP
\f[C]epmfs\f[R] is the default \f[C]category.create\f[R] policy because
\f[C]ep\f[R] policies are not going to change the general layout of the
branches.
It won\[cq]t place files/dirs on branches that don\[cq]t already have
the relative branch.
So it keeps the system in a known state.
It\[cq]s much easier to stop using \f[C]epmfs\f[R] or redistribute files
around the filesystem than it is to consolidate them back.
.SS What settings should I use?
.PP
Depends on what features you want.
Generally speaking there are no \[lq]wrong\[rq] settings.
All settings are performance or feature related.
The best bet is to read over the available options and choose what fits
your situation.
If something isn\[cq]t clear from the documentation please reach out and
the documentation will be improved.
.PP
That said, for the average person, the following should be fine:
.PP
\f[C]cache.files=off,dropcacheonclose=true,category.create=mfs\f[R]
.SS Why are all my files ending up on 1 filesystem?!
.PP
Did you start with empty filesystems?
Did you explicitly configure a \f[C]category.create\f[R] policy?
Are you using an \f[C]existing path\f[R] / \f[C]path preserving\f[R]
policy?
.PP
The default create policy is \f[C]epmfs\f[R].
That is a path preserving algorithm.
With such a policy for \f[C]mkdir\f[R] and \f[C]create\f[R] with a set
of empty filesystems it will select only 1 filesystem when the first
directory is created.
Anything, files or directories, created in that first directory will be
placed on the same branch because it is preserving paths.
.PP
This catches a lot of new users off guard but changing the default would
break the setup for many existing users and this policy is the safest
policy as it will not change the general layout of the existing
filesystems.
If you do not care about path preservation and wish your files to be
spread across all your filesystems change to \f[C]mfs\f[R] or similar
policy as described above.
If you do want path preservation you\[cq]ll need to perform the manual
act of creating paths on the filesystems you want the data to land on
before transferring your data.
Setting \f[C]func.mkdir=epall\f[R] can simplify managing path
preservation for \f[C]create\f[R].
Or use \f[C]func.mkdir=rand\f[R] if you\[cq]re interested in just
grouping together directory content by filesystem.
.SS Do hardlinks work?
.PP
Yes.
See also the option \f[C]inodecalc\f[R] for how inode values are
calculated.
.PP
What mergerfs does not do is fake hard links across branches.
Read the section \[lq]rename & link\[rq] for how it works.
.PP
Remember that hardlinks will NOT work across devices.
That includes between the original filesystem and a mergerfs pool,
between two separate pools of the same underlying filesystems, or bind
mounts of paths within the mergerfs pool.
The latter is common when using Docker or Podman.
Multiple volumes (bind mounts) to the same underlying filesystem are
considered different devices.
There is no way to link between them.
You should mount in the highest directory in the mergerfs pool that
includes all the paths you need if you want links to work.
.SS Does FICLONE or FICLONERANGE work?
.PP
Unfortunately not.
FUSE, the technology mergerfs is based on, does not support the
\f[C]clone_file_range\f[R] feature needed for it to work.
mergerfs won\[cq]t even know such a request is made.
The kernel will simply return an error back to the application making
the request.
.PP
Should FUSE gain the ability mergerfs will be updated to support it.
.SS Can I use mergerfs without SnapRAID? SnapRAID without mergerfs?
.PP
Yes.
They are completely unrelated pieces of software.
.SS Can mergerfs run via Docker, Podman, Kubernetes, etc.
.PP
Yes.
With Docker you\[cq]ll need to include
\f[C]--cap-add=SYS_ADMIN --device=/dev/fuse --security-opt=apparmor:unconfined\f[R]
or similar with other container runtimes.
You should also be running it as root or given sufficient caps to change
user and group identity as well as have root like filesystem
permissions.
.PP
Keep in mind that you \f[B]MUST\f[R] consider identity when using
containers.
For example: supplemental groups will be picked up from the container
unless you properly manage users and groups by sharing relevant /etc
files or by using some other means to share identity across containers.
Similarly if you use \[lq]rootless\[rq] containers and user namespaces
to do uid/gid translations you \f[B]MUST\f[R] consider that while
managing shared files.
.PP
Also, as mentioned by hotio (https://hotio.dev/containers/mergerfs),
with Docker you should probably be mounting with
\f[C]bind-propagation\f[R] set to \f[C]slave\f[R].
.SS Does mergerfs support CoW / copy-on-write / writes to read-only filesystems?
.PP
Not in the sense of a filesystem like BTRFS or ZFS nor in the overlayfs
or aufs sense.
It does offer a
cow-shell (http://manpages.ubuntu.com/manpages/bionic/man1/cow-shell.1.html)
like hard link breaking (copy to temp file then rename over original)
which can be useful when wanting to save space by hardlinking duplicate
files but wish to treat each name as if it were a unique and separate
file.
.PP
If you want to write to a read-only filesystem you should look at
overlayfs.
You can always include the overlayfs mount into a mergerfs pool.
.SS Why can\[cq]t I see my files / directories?
.PP
It\[cq]s almost always a permissions issue.
Unlike mhddfs and unionfs-fuse, which runs as root and attempts to
access content as such, mergerfs always changes its credentials to that
of the caller.
This means that if the user does not have access to a file or directory
than neither will mergerfs.
However, because mergerfs is creating a union of paths it may be able to
read some files and directories on one filesystem but not another
resulting in an incomplete set.
.PP
Whenever you run into a split permission issue (seeing some but not all
files) try using
mergerfs.fsck (https://github.com/trapexit/mergerfs-tools) tool to check
for and fix the mismatch.
If you aren\[cq]t seeing anything at all be sure that the basic
permissions are correct.
The user and group values are correct and that directories have their
executable bit set.
A common mistake by users new to Linux is to \f[C]chmod -R 644\f[R] when
they should have \f[C]chmod -R u=rwX,go=rX\f[R].
.PP
If using a network filesystem such as NFS, SMB, CIFS (Samba) be sure to
pay close attention to anything regarding permissioning and users.
Root squashing and user translation for instance has bitten a few
mergerfs users.
Some of these also affect the use of mergerfs from container platforms
such as Docker.
.SS Why use FUSE? Why not a kernel based solution?
.PP
As with any solutions to a problem there are advantages and
disadvantages to each one.
.PP
A FUSE based solution has all the downsides of FUSE:
.IP \[bu] 2
Higher IO latency due to the trips in and out of kernel space
.IP \[bu] 2
Higher general overhead due to trips in and out of kernel space
.IP \[bu] 2
Double caching when using page caching
.IP \[bu] 2
Misc limitations due to FUSE\[cq]s design
.PP
But FUSE also has a lot of upsides:
.IP \[bu] 2
Easier to offer a cross platform solution
.IP \[bu] 2
Easier forward and backward compatibility
.IP \[bu] 2
Easier updates for users
.IP \[bu] 2
Easier and faster release cadence
.IP \[bu] 2
Allows more flexibility in design and features
.IP \[bu] 2
Overall easier to write, secure, and maintain
.IP \[bu] 2
Much lower barrier to entry (getting code into the kernel takes a lot of
time and effort initially)
.PP
FUSE was chosen because of all the advantages listed above.
The negatives of FUSE do not outweigh the positives.
.SS Is my OS\[cq]s libfuse needed for mergerfs to work?
.PP
No.\ Normally \f[C]mount.fuse\f[R] is needed to get mergerfs (or any
FUSE filesystem to mount using the \f[C]mount\f[R] command but in
vendoring the libfuse library the \f[C]mount.fuse\f[R] app has been
renamed to \f[C]mount.mergerfs\f[R] meaning the filesystem type in
\f[C]fstab\f[R] can simply be \f[C]mergerfs\f[R].
That said there should be no harm in having it installed and continuing
to using \f[C]fuse.mergerfs\f[R] as the type in \f[C]/etc/fstab\f[R].
.PP
If \f[C]mergerfs\f[R] doesn\[cq]t work as a type it could be due to how
the \f[C]mount.mergerfs\f[R] tool was installed.
Must be in \f[C]/sbin/\f[R] with proper permissions.
.SS Why was splice support removed?
.PP
After a lot of testing over the years splicing always appeared to be at
best provide equivalent performance and in cases worse performance.
Splice is not supported on other platforms forcing a traditional
read/write fallback to be provided.
The splice code was removed to simplify the codebase.
.SS What should mergerfs NOT be used for?
.IP \[bu] 2
databases: Even if the database stored data in separate files (mergerfs
wouldn\[cq]t offer much otherwise) the higher latency of the indirection
will kill performance.
If it is a lightly used SQLITE database then it may be fine but
you\[cq]ll need to test.
.IP \[bu] 2
VM images: For the same reasons as databases.
VM images are accessed very aggressively and mergerfs will introduce too
much latency (if it works at all).
.IP \[bu] 2
As replacement for RAID: mergerfs is just for pooling branches.
If you need that kind of device performance aggregation or high
availability you should stick with RAID.
.SS Can filesystems be written to directly? Outside of mergerfs while pooled?
.PP
Yes, however it\[cq]s not recommended to use the same file from within
the pool and from without at the same time (particularly writing).
Especially if using caching of any kind (cache.files, cache.entry,
cache.attr, cache.negative_entry, cache.symlinks, cache.readdir, etc.)
as there could be a conflict between cached data and not.
.SS Why do I get an \[lq]out of space\[rq] / \[lq]no space left on device\[rq] / ENOSPC error even though there appears to be lots of space available?
.PP
First make sure you\[cq]ve read the sections above about policies, path
preservation, branch filtering, and the options \f[B]minfreespace\f[R],
\f[B]moveonenospc\f[R], \f[B]statfs\f[R], and \f[B]statfs_ignore\f[R].
.PP
mergerfs is simply presenting a union of the content within multiple
branches.
The reported free space is an aggregate of space available within the
pool (behavior modified by \f[B]statfs\f[R] and
\f[B]statfs_ignore\f[R]).
It does not represent a contiguous space.
In the same way that read-only filesystems, those with quotas, or
reserved space report the full theoretical space available.
.PP
Due to path preservation, branch tagging, read-only status, and
\f[B]minfreespace\f[R] settings it is perfectly valid that
\f[C]ENOSPC\f[R] / \[lq]out of space\[rq] / \[lq]no space left on
device\[rq] be returned.
It is doing what was asked of it: filtering possible branches due to
those settings.
Only one error can be returned and if one of the reasons for filtering a
branch was \f[B]minfreespace\f[R] then it will be returned as such.
\f[B]moveonenospc\f[R] is only relevant to writing a file which is too
large for the filesystem it\[cq]s currently on.
.PP
It is also possible that the filesystem selected has run out of inodes.
Use \f[C]df -i\f[R] to list the total and available inodes per
filesystem.
.PP
If you don\[cq]t care about path preservation then simply change the
\f[C]create\f[R] policy to one which isn\[cq]t.
\f[C]mfs\f[R] is probably what most are looking for.
The reason it\[cq]s not default is because it was originally set to
\f[C]epmfs\f[R] and changing it now would change people\[cq]s setup.
Such a setting change will likely occur in mergerfs 3.
.SS Why does the total available space in mergerfs not equal outside?
.PP
Are you using ext2/3/4?
With reserve for root?
mergerfs uses available space for statfs calculations.
If you\[cq]ve reserved space for root then it won\[cq]t show up.
.PP
You can remove the reserve by running: \f[C]tune2fs -m 0 <device>\f[R]
.SS Can mergerfs mounts be exported over NFS?
.PP
Yes, however if you do anything which may changes files out of band
(including for example using the \f[C]newest\f[R] policy) it will result
in \[lq]stale file handle\[rq] errors unless properly setup.
.PP
Be sure to use the following options:
.IP \[bu] 2
noforget
.IP \[bu] 2
inodecalc=path-hash
.SS Can mergerfs mounts be exported over Samba / SMB?
.PP
Yes.
While some users have reported problems it appears to always be related
to how Samba is setup in relation to permissions.
.SS Can mergerfs mounts be used over SSHFS?
.PP
Yes.
.SS I notice massive slowdowns of writes when enabling cache.files.
.PP
When file caching is enabled in any form (\f[C]cache.files!=off\f[R] or
\f[C]direct_io=false\f[R]) it will issue \f[C]getxattr\f[R] requests for
\f[C]security.capability\f[R] prior to \f[I]every single write\f[R].
This will usually result in a performance degradation, especially when
using a network filesystem (such as NFS or CIFS/SMB/Samba.)
Unfortunately at this moment the kernel is not caching the response.
.PP
To work around this situation mergerfs offers a few solutions.
.IP "1." 3
Set \f[C]security_capability=false\f[R].
It will short circuit any call and return \f[C]ENOATTR\f[R].
This still means though that mergerfs will receive the request before
every write but at least it doesn\[cq]t get passed through to the
underlying filesystem.
.IP "2." 3
Set \f[C]xattr=noattr\f[R].
Same as above but applies to \f[I]all\f[R] calls to getxattr.
Not just \f[C]security.capability\f[R].
This will not be cached by the kernel either but mergerfs\[cq] runtime
config system will still function.
.IP "3." 3
Set \f[C]xattr=nosys\f[R].
Results in mergerfs returning \f[C]ENOSYS\f[R] which \f[I]will\f[R] be
cached by the kernel.
No future xattr calls will be forwarded to mergerfs.
The downside is that also means the xattr based config and query
functionality won\[cq]t work either.
.IP "4." 3
Disable file caching.
If you aren\[cq]t using applications which use \f[C]mmap\f[R] it\[cq]s
probably simpler to just disable it all together.
The kernel won\[cq]t send the requests when caching is disabled.
.SS It\[cq]s mentioned that there are some security issues with mhddfs. What are they? How does mergerfs address them?
.PP
mhddfs (https://github.com/trapexit/mhddfs) manages running as
\f[B]root\f[R] by calling
getuid() (https://github.com/trapexit/mhddfs/blob/cae96e6251dd91e2bdc24800b4a18a74044f6672/src/main.c#L319)
and if it returns \f[B]0\f[R] then it will
chown (http://linux.die.net/man/1/chown) the file.
Not only is that a race condition but it doesn\[cq]t handle other
situations.
Rather than attempting to simulate POSIX ACL behavior the proper way to
manage this is to use seteuid (http://linux.die.net/man/2/seteuid) and
setegid (http://linux.die.net/man/2/setegid), in effect becoming the
user making the original call, and perform the action as them.
This is what mergerfs does and why mergerfs should always run as root.
.PP
In Linux setreuid syscalls apply only to the thread.
GLIBC hides this away by using realtime signals to inform all threads to
change credentials.
Taking after \f[B]Samba\f[R], mergerfs uses
\f[B]syscall(SYS_setreuid,\&...)\f[R] to set the callers credentials for
that thread only.
Jumping back to \f[B]root\f[R] as necessary should escalated privileges
be needed (for instance: to clone paths between filesystems).
.PP
For non-Linux systems mergerfs uses a read-write lock and changes
credentials only when necessary.
If multiple threads are to be user X then only the first one will need
to change the processes credentials.
So long as the other threads need to be user X they will take a readlock
allowing multiple threads to share the credentials.
Once a request comes in to run as user Y that thread will attempt a
write lock and change to Y\[cq]s credentials when it can.
If the ability to give writers priority is supported then that flag will
be used so threads trying to change credentials don\[cq]t starve.
This isn\[cq]t the best solution but should work reasonably well
assuming there are few users.
.SH mergerfs versus X
.SS mhddfs
.PP
mhddfs had not been maintained for some time and has some known
stability and security issues.
mergerfs provides a superset of mhddfs\[cq] features and should offer
the same or better performance.
.PP
Below is an example of mhddfs and mergerfs setup to work similarly.
.PP
\f[C]mhddfs -o mlimit=4G,allow_other /mnt/drive1,/mnt/drive2 /mnt/pool\f[R]
.PP
\f[C]mergerfs -o minfreespace=4G,category.create=ff /mnt/drive1:/mnt/drive2 /mnt/pool\f[R]
.SS aufs
.PP
aufs is mostly abandoned and no longer available in most Linux distros.
.PP
While aufs can offer better peak performance mergerfs provides more
configurability and is generally easier to use.
mergerfs however does not offer the overlay / copy-on-write (CoW)
features which aufs has.
.SS unionfs-fuse
.PP
unionfs-fuse is more like aufs than mergerfs in that it offers overlay /
copy-on-write (CoW) features.
If you\[cq]re just looking to create a union of filesystems and want
flexibility in file/directory placement then mergerfs offers that
whereas unionfs is more for overlaying read/write filesystems over
read-only ones.
.SS overlayfs
.PP
overlayfs is similar to aufs and unionfs-fuse in that it also is
primarily used to layer a read/write filesystem over one or more
read-only filesystems.
It does not have the ability to spread files/directories across numerous
filesystems.
.SS RAID0, JBOD, drive concatenation, striping
.PP
With simple JBOD / drive concatenation / stripping / RAID0 a single
drive failure will result in full pool failure.
mergerfs performs a similar function without the possibility of
catastrophic failure and the difficulties in recovery.
Drives may fail but all other filesystems and their data will continue
to be accessible.
.PP
The main practical difference with mergerfs is the fact you don\[cq]t
actually have contiguous space as large as if you used those other
technologies.
Meaning you can\[cq]t create a 2TB file on a pool of 2 1TB filesystems.
.PP
When combined with something like SnapRaid (http://www.snapraid.it)
and/or an offsite backup solution you can have the flexibility of JBOD
without the single point of failure.
.SS UnRAID
.PP
UnRAID is a full OS and its storage layer, as I understand, is
proprietary and closed source.
Users who have experience with both have often said they prefer the
flexibility offered by mergerfs and for some the fact it is open source
is important.
.PP
There are a number of UnRAID users who use mergerfs as well though
I\[cq]m not entirely familiar with the use case.
.PP
For semi-static data mergerfs + SnapRaid (http://www.snapraid.it)
provides a similar solution.
.SS ZFS
.PP
mergerfs is very different from ZFS.
mergerfs is intended to provide flexible pooling of arbitrary
filesystems (local or remote), of arbitrary sizes, and arbitrary
filesystems.
For \f[C]write once, read many\f[R] usecases such as bulk media storage.
Where data integrity and backup is managed in other ways.
In those usecases ZFS can introduce a number of costs and limitations as
described
here (http://louwrentius.com/the-hidden-cost-of-using-zfs-for-your-home-nas.html),
here (https://markmcb.com/2020/01/07/five-years-of-btrfs/), and
here (https://utcc.utoronto.ca/~cks/space/blog/solaris/ZFSWhyNoRealReshaping).
.SS StableBit\[cq]s DrivePool
.PP
DrivePool works only on Windows so not as common an alternative as other
Linux solutions.
If you want to use Windows then DrivePool is a good option.
Functionally the two projects work a bit differently.
DrivePool always writes to the filesystem with the most free space and
later rebalances.
mergerfs does not offer rebalance but chooses a branch at file/directory
create time.
DrivePool\[cq]s rebalancing can be done differently in any directory and
has file pattern matching to further customize the behavior.
mergerfs, not having rebalancing does not have these features, but
similar features are planned for mergerfs v3.
DrivePool has builtin file duplication which mergerfs does not natively
support (but can be done via an external script.)
.PP
There are a lot of misc differences between the two projects but most
features in DrivePool can be replicated with external tools in
combination with mergerfs.
.PP
Additionally DrivePool is a closed source commercial product vs mergerfs
a ISC licensed OSS project.
.SH SUPPORT
.PP
Filesystems are complex and difficult to debug.
mergerfs, while being just a proxy of sorts, can be difficult to debug
given the large number of possible settings it can have itself and the
number of environments it can run in.
When reporting on a suspected issue \f[B]please\f[R] include as much of
the below information as possible otherwise it will be difficult or
impossible to diagnose.
Also please read the above documentation as it provides details on many
previously encountered questions/issues.
.PP
\f[B]Please make sure you are using the latest
release (https://github.com/trapexit/mergerfs/releases) or have tried it
in comparison. Old versions, which are often included in distros like
Debian and Ubuntu, are not ever going to be updated and the issue you
are encountering may have been addressed already.\f[R]
.PP
\f[B]For commercial support or feature requests please contact me
directly. (mailto:support@spawn.link)\f[R]
.SS Information to include in bug reports
.IP \[bu] 2
Information about the broader problem along with any attempted
solutions. (https://xyproblem.info)
.IP \[bu] 2
Solution already ruled out and why.
.IP \[bu] 2
Version of mergerfs: \f[C]mergerfs --version\f[R]
.IP \[bu] 2
mergerfs settings / arguments: from fstab, systemd unit, command line,
OMV plugin, etc.
.IP \[bu] 2
Version of the OS: \f[C]uname -a\f[R] and \f[C]lsb_release -a\f[R]
.IP \[bu] 2
List of branches, their filesystem types, sizes (before and after
issue): \f[C]df -h\f[R]
.IP \[bu] 2
\f[B]All\f[R] information about the relevant paths and files:
permissions, ownership, etc.
.IP \[bu] 2
\f[B]All\f[R] information about the client app making the requests:
version, uid/gid
.IP \[bu] 2
Runtime environment:
.RS 2
.IP \[bu] 2
Is mergerfs running within a container?
.IP \[bu] 2
Are the client apps using mergerfs running in a container?
.RE
.IP \[bu] 2
A \f[C]strace\f[R] of the app having problems:
.RS 2
.IP \[bu] 2
\f[C]strace -fvTtt -s 256 -o /tmp/app.strace.txt <cmd>\f[R]
.RE
.IP \[bu] 2
A \f[C]strace\f[R] of mergerfs while the program is trying to do
whatever it is failing to do:
.RS 2
.IP \[bu] 2
\f[C]strace -fvTtt -s 256 -p <mergerfsPID> -o /tmp/mergerfs.strace.txt\f[R]
.RE
.IP \[bu] 2
\f[B]Precise\f[R] directions on replicating the issue.
Do not leave \f[B]anything\f[R] out.
.IP \[bu] 2
Try to recreate the problem in the simplest way using standard programs:
\f[C]ln\f[R], \f[C]mv\f[R], \f[C]cp\f[R], \f[C]ls\f[R], \f[C]dd\f[R],
etc.
.SS Contact / Issue submission
.IP \[bu] 2
github.com: https://github.com/trapexit/mergerfs/issues
.IP \[bu] 2
discord: https://discord.gg/MpAr69V
.IP \[bu] 2
reddit: https://www.reddit.com/r/mergerfs
.SS Donations
.PP
https://github.com/trapexit/support
.PP
Development and support of a project like mergerfs requires a
significant amount of time and effort.
The software is released under the very liberal ISC license and is
therefore free to use for personal or commercial uses.
.PP
If you are a personal user and find mergerfs and its support valuable
and would like to support the project financially it would be very much
appreciated.
.PP
If you are using mergerfs commercially please consider sponsoring the
project to ensure it continues to be maintained and receive updates.
If custom features are needed feel free to contact me
directly (mailto:support@spawn.link).
.SH LINKS
.IP \[bu] 2
https://spawn.link
.IP \[bu] 2
https://github.com/trapexit/mergerfs
.IP \[bu] 2
https://github.com/trapexit/mergerfs/wiki
.IP \[bu] 2
https://github.com/trapexit/mergerfs-tools
.IP \[bu] 2
https://github.com/trapexit/scorch
.IP \[bu] 2
https://github.com/trapexit/bbf