上QQ阅读APP看书，第一时间看更新

Mounting volumes from host directories

You use the --volume option to explicitly map a directory in a container from a known location on the host. The target location in the container can be a directory created with the VOLUME command, or any directory in the container's filesystem. If the target location already exists in the Docker image, it is hidden by the volume mount, so you won't see any of the image files.

I'll create a dummy configuration file for my app in a directory on the C drive on my Windows machine:

PS> mkdir C:\app-config | Out-Null
PS> echo 'VERSION=18.09' > C:\app-config\version.txt

Now I'll run a container which maps a volume from the host, and read the configuration file which is actually stored on the host:

> docker container run `
   --volume C:\app-config:C:\app\config `
   dockeronwindows/ch02-volumes:2e `
   type C:\app\config\version.txt
VERSION=18.09

The --volume option specifies the mount in the format {source}:{target}. The source is the host location, which needs to exist. The target is the container location, which doesn't need to exist, but the existing contents will be hidden if it does exist.

Volume mounts are different in Windows and Linux containers. In Linux containers, Docker merges the contents from the source into the target, so if files exist in the image, you see them as well as the contents of the volume source. Docker on Linux also lets you mount a single file location, but on Windows you can only mount whole directories.

Volume mounts are useful for running stateful applications in containers, like databases. You can run SQL Server in a container, and have the database files stored in a location on the host, which could be a RAID array on the server. When you have schema updates, you remove the old container and start a new container from the updated Docker image. You use the same volume mount for the new container, so that the data is preserved from the old container.