bindmapfuse

bindmapfuse is a FUSE filesystem that exposes an underlying filesystem according to a potentially very large set of bind mounts. It was developed to solve a problem in which a very large number (hundreds of thousands to millions) of bind mounts were desired in a docker container, but the docker daemon does not support this because the container configuration became too large.

Example Usage

Create a YAML (or JSON) configuration file that contains a mounts key, which is a map of "mountpoints" and real paths (i.e. paths on the underlying filesystem). Like bind mounts with docker, both files and directories can be bind mounted.

Example YAML configuration (example.yaml):

mounts:
  file1.txt: "/tmp/foo/bar/foobarfile.txt"
  dir1: "/tmp/baz"
  file2.txt: "/tmp/qux/quxfile.txt"
  "dir1/qux": "/tmp/qux"

This can then be mounted at a /tmp/mnt mountpoint by running bindmapfuse:

$ bindmapfuse /tmp/mnt -o bind_map_config=example.yaml &

If the underlying files do not exist, that will cause a "No such file or directory" error, though at present they will still be listed in the directory listing but without any stat information:

$ ls -lR /tmp/mnt
/tmp/mnt:
ls: cannot access '/tmp/mnt/dir1': No such file or directory
ls: cannot access '/tmp/mnt/file1.txt': No such file or directory
ls: cannot access '/tmp/mnt/file2.txt': No such file or directory
total 0
?????????? ? ? ? ?            ? dir1
?????????? ? ? ? ?            ? file1.txt
?????????? ? ? ? ?            ? file2.txt

You can create suitable example files and directories on the underlying filesystem:

$ mkdir /tmp/foo /tmp/foo/bar /tmp/baz /tmp/qux
$ echo foobar > /tmp/foo/bar/foobarfile.txt
$ echo baz > /tmp/baz/bazfile.txt
$ echo qux > /tmp/qux/quxfile.txt

After which the specified bind mounts will be available under the mount point:

$ ls -R /tmp/mnt
/tmp/mnt:
dir1  file1.txt  file2.txt

/tmp/mnt/dir1:
bazfile.txt  qux

/tmp/mnt/dir1/qux:
quxfile.txt