Mastering Docker Storage: When to Use Storage Drivers vs Data Volumes

Overview

Docker provides two ways to store data: the layers managed by a storage driver and Data Volumes.

1.1 Storage Driver

Containers consist of a writable top layer and several read‑only image layers. New data is written to the top layer, and modifications trigger a copy‑on‑write from the image layer to the container layer, preserving the original image. The storage driver merges these layers into a single view. Docker supports several drivers, including AUFS, Device Mapper, Btrfs, OverlayFS, VFS and ZFS. Because no driver fits every scenario and drivers evolve quickly, Docker recommends using the default driver supplied by the Linux distribution.

Run docker info to see the default driver on Ubuntu.

1.2 Data Volume

A Data Volume is a directory or file on the Docker host that is mounted into a container. Its characteristics are:

It is a directory or file, not an unformatted block device.

Containers can read and write the volume.

The data persists even after the container is removed.

Stateless containers (e.g., busybox) can store data in the image layers, while stateful containers that need persistent data should use a Data Volume.

1.2.1 Bind Mount

Bind mount maps an existing host directory or file into a container, e.g.:

docker run -d -p 4555:80 -v ~/htdocs:/usr/local/apache2/htdocs httpd

The syntax is <host path>:<container path>. You can set the mount as read‑only with :ro:

docker run -d -p 4555:80 -v ~/htdocs:/usr/local/apache2/htdocs:ro httpd

Bind mounts can also target a single file:

docker run -d -p 4555:80 -v ~/htdocs/index.html:/usr/local/apache2/htdocs/new_index.html httpd

Advantages: immediate host‑container data sharing and easy updates. Drawbacks: requires a specific host path, reducing container portability.

1.2.2 Docker Managed Volume

Docker managed volumes are created with -v without specifying a host source. Docker creates a directory under /var/lib/docker/volumes (or the path set by data-root) and mounts it to the container path. If the mount point already contains data, Docker copies it into the new volume.

Example:

docker run -d -p 4555:80 -v /usr/local/apache2/htdocs httpd

You can inspect the volume location with docker inspect.

Comparison

Aspect

Bind Mount

Docker Managed Volume

Volume location

Any host path

/var/lib/docker/volumes/…

Effect on existing mount point

Hides and replaces with volume

Copies existing data into volume

Supports single file

Yes

No (directory only)

Permission control

Read‑write by default, can set read‑only

Always read‑write

Portability

Weak, tied to host path

Strong, no host path needed

2. Data Sharing

2.1 Container ↔ Host

Both bind mounts and managed volumes enable sharing data between container and host. Bind mounts directly expose a host directory; managed volumes require copying data into the volume before sharing.

You can also copy files with docker cp:

docker cp ~/index.html <container_id>:/usr/local/apache2/htdocs/

2.2 Container ↔ Container

Multiple containers can share the same bind mount:

docker run --name web1 -d -p 80 -v ~/htdocs:/usr/local/apache2/htdocs httpd
docker run --name web2 -d -p 80 -v ~/htdocs:/usr/local/apache2/htdocs httpd
docker run --name web3 -d -p 80 -v ~/htdocs:/usr/local/apache2/htdocs httpd

Or share a Docker managed volume via a “volume container”:

docker create --name vc_data -v ~/htdocs:/usr/local/apache2/htdocs -v /other/tools busybox

Other containers can use --volumes-from vc_data to inherit those volumes.

2.3 Data‑Packed Volume Container

A Dockerfile can embed data and declare a VOLUME so the image creates a managed volume automatically:

FROM busybox:latest
ADD htdocs /usr/local/apache2/htdocs
VOLUME /usr/local/apache2/htdocs

Build and create the container:

docker build -t datapacked .
docker create --name vc_data_packed datapacked
docker run --name datapacked1 -d -p 80 --volumes-from vc_data_packed httpd

This approach packages the data inside the image, making the container self‑contained and highly portable.

3. Data Volume Lifecycle

3.1 Backup

Since a volume is a directory on the host (typically under /var/lib/docker/volumes), backing it up is equivalent to backing up that part of the host filesystem.

3.2 Migration

To migrate a service, stop the old container and start a new one mounting the existing volume:

docker stop old_container
docker run -d -p 5000:5000 -v /myregistry:/var/lib/registry registry:latest

3.3 Destruction

Remove unneeded volumes with docker volume rm. Deleting a container with -v also removes its managed volumes, provided no other container is using them. Orphaned volumes can be cleaned up with docker volume rm $(docker volume ls -q).