Edit 25 Oct 2021: Seems this no longer works. So much for a workaround.

Here is something that's cool.

You can have a docker volume, docker volume create example, and you can mount this volume to a container, docker run -v example:/path image. You can also mount local directories to containers, docker run -v $(pwd)/local:/path image. But something you can't do, is mount a local directory to volume.

Scenario: you have a volume, you have a container using that volume, and you have some software that needs to access that volume data locally, outside of docker.

🚫docker run -v $(pwd)/local:example image isn't going to work.

Nor can you mount two things to the same place.

🚫docker run -v example:/example -v $(pwd)/local:/example image isn't going to work either.

The solution is to use intermediate symlinked container.

I call this solution the map volume.

#
# map volume
#
FROM busybox
RUN mkdir /volume
RUN ln -s /local /volume
CMD tail -f /dev/null

In this dockerfile you can see that we create a directory (later to be used as a mount point), and then we create a symlink directory pointing to the same place.

$ docker build -t mapvolume .
$ docker volume create my_precious_data
$ docker run -d -v my_precious_data:/volume -v $(pwd)/my_precious_data:/local mapvolume

We create a volume my_precious_data, then we run the mapvolume, first mounting the volume my_precious_data to the container path /volume, and then mounting the local directory $(pwd)/my_precious_data to the container path /local which as we saw, is actually /volume.

Now you have mounted a docker volume to your local filesystem.

Update: I realised today that the original dockerfile wouldn't stay alive. I therefore changed the CMD to CMD tail -f /dev/null to keep the container alive. Thanks to bigdatums.net for that snippet.