mergerfs is similar to mhddfs, unionfs, and aufs. Like mhddfs in that it too uses FUSE. Like aufs in that it provides multiple policies for how to handle behavior.
Why create mergerfs when those exist? mhddfs isn't really maintained or flexible. There are also issues with running as root. aufs is more flexible but contains some hard to debug inconsistencies in behavior. Neither support file attributes ([chattr](http://linux.die.net/man/1/chattr)).
Policies
========
mergerfs - another FUSE union filesystem
Filesystem calls are broken up into 4 functional categories: search, action, create, and none. These categories can be assigned a policy which dictates how [mergerfs](http://github.com/trapexit/mergerfs) behaves while when action on the filesystem. Any policy can be assigned to a category though some aren't terribly practical. For instance: rand (Random) may be useful for **create** but could lead to very odd behavior if used for **search** or **action**. Since the input for any policy is the source mounts and fusepath and the output a vector of targets the choice was made to simplify the implementation and allow a policies usage in any category. **NOTE:** In any policy which can return more than one location (currently only **all**) the first value will be used in **search** and **create** policies since they can only ever act on 1 filepath.
| ff (first found) | Given the order the paths were provided at mount time act on the first one found (regardless if stat would return EACCES). |
| ffwp (first found w/ permissions) | Given the order the paths were provided at mount time act on the first one found which you have access (stat does not error with EACCES). |
| newest (newest file) | If multiple files exist return the one with the most recent mtime. |
| all (all files found) | Attempt to apply the call to each file found. If any sub call succeeds the entire operation succeeds and other errors ignored. If all fail then the last error is reported. |
| mfs (most free space) | Assuming the path is found to exist (ENOENT would not be returned) use the drive with the most free space available. |
| epmfs (existing path, most free space) | If the path exists in multiple locations use the one with the most free space. Otherwise fall back to mfs. |
| rand (random) | Pick a destination at random. Again the dirname of the full path must exist somewhere. |
It normalizes the source drives 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 drive so multiple points on the same drive will not result in double counting it's space.
mergerfs is similar to mhddfs, unionfs, and aufs. Like mhddfs in that it too uses FUSE. Like aufs in that it provides multiple policies for how to handle behavior.
**NOTE:** create is really a search for existence and then create. The 'search' policy applies to the first part. If the [dirname](http://linux.die.net/man/3/dirname) of the full path is not found to exist [ENOENT](http://linux.die.net/man/3/errno) is returned.
Why create mergerfs when those exist? mhddfs isn't really maintained or flexible. There are also issues with running as root. aufs is more flexible but contains some hard to debug inconsistencies in behavior. Neither support file attributes ([chattr](http://linux.die.net/man/1/chattr)).
@ -71,8 +45,36 @@ In /etc/fstab it'd look like the following:
**NOTE:** the globbing is done at mount time. If a new directory is added matching the glob after the fact it will not be included.
**NOTE:** the globbing is done at mount time. If a new directory is added matching the glob after the fact it will not be included.
Building
========
# POLICIES
Filesystem calls are broken up into 4 functional categories: search, action, create, and none. These categories can be assigned a policy which dictates how [mergerfs](http://github.com/trapexit/mergerfs) behaves while when action on the filesystem. Any policy can be assigned to a category though some aren't terribly practical. For instance: rand (Random) may be useful for **create** but could lead to very odd behavior if used for **search** or **action**. Since the input for any policy is the source mounts and fusepath and the output a vector of targets the choice was made to simplify the implementation and allow a policies usage in any category. **NOTE:** In any policy which can return more than one location (currently only **all**) the first value will be used in **search** and **create** policies since they can only ever act on 1 filepath.
| ff (first found) | Given the order the paths were provided at mount time act on the first one found (regardless if stat would return EACCES). |
| ffwp (first found w/ permissions) | Given the order the paths were provided at mount time act on the first one found which you have access (stat does not error with EACCES). |
| newest (newest file) | If multiple files exist return the one with the most recent mtime. |
| all (all files found) | Attempt to apply the call to each file found. If any sub call succeeds the entire operation succeeds and other errors ignored. If all fail then the last error is reported. |
| mfs (most free space) | Assuming the path is found to exist (ENOENT would not be returned) use the drive with the most free space available. |
| epmfs (existing path, most free space) | If the path exists in multiple locations use the one with the most free space. Otherwise fall back to mfs. |
| rand (random) | Pick a destination at random. Again the dirname of the full path must exist somewhere. |
#### statvfs ####
It normalizes the source drives 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 drive so multiple points on the same drive will not result in double counting it's space.
**NOTE:** create is really a search for existence and then create. The 'search' policy applies to the first part. If the [dirname](http://linux.die.net/man/3/dirname) of the full path is not found to exist [ENOENT](http://linux.die.net/man/3/errno) is returned.
# BUILDING
* Need to install FUSE development libraries (libfuse-dev).
* Need to install FUSE development libraries (libfuse-dev).
* Optionally need libattr1 (libattr1-dev).
* Optionally need libattr1 (libattr1-dev).
@ -84,8 +86,7 @@ usage: make
make XATTR_AVAILABLE=0 - to build program without xattrs functionality (auto discovered otherwise)
make XATTR_AVAILABLE=0 - to build program without xattrs functionality (auto discovered otherwise)
Antonio SJ Musumeci
11 years ago