I just inspected my /var/lib/docker/volumes folder and discovered that is bursting with folders named as Docker UUIDs each of which contain a config.json file with contents along the lines of
{"ID":"UUID","Path":"/path/to/mounted/volume","IsBindMount":true,"Writable":true}
where
/path/to/mounted/volume
is the path to the folder on the host that was mounted on to a docker container with the -v switch at some point. I have such folders dating back to the start of my experiments with Docker, i.e. about 3 weeks ago.
The containers in question were stopped and docker rm'ed a long time ago so I cannot see that those entries are not past their sell-by date. This begs the question - is the left over I am seeing a bug or does one need to manually discard such entries from /var/lib/docker/volumes?
For Docker 1.9 and up there's a native way:
List all orphaned volumes with
$ docker volume ls -qf dangling=true
Eliminate all of them with
$ docker volume rm $(docker volume ls -qf dangling=true)
From the Docker user guide:
If you remove containers that mount volumes, including the initial dbdata container, or the subsequent containers db1 and db2, the volumes will not be deleted. To delete the volume from disk, you must explicitly call docker rm -v against the last container with a reference to the volume. This allows you to upgrade, or effectively migrate data volumes between containers. - source
This is intentional behavior to avoid accidental data loss. You can use a tool like docker-cleanup-volumes to clean out unused volumes.
For Docker 1.13+ and the ce/ee 17+ release numbers, use the volume prune
command
docker volume prune
Unlike the dangling=true
query, this will not remove "remote" driver based volumes.