Fork from: https://github.com/piscue/docker-backup.git which was a fork from Fork from: https://github.com/boombatower/docker-backup
My approach is to create a light image using Alpine. Respecting the original documentation but skipping S3 option:
Simple utility for backing up a Docker data container.
There are a number of existing solutions available which typically operate in a similar fashion.
docker inspect
(api call) to determine volumes provided by a container- access container volumes using
--volumes-from
or/var/lib/docker/vfs/dir
- store container description (from inspect) and volumes
- restore using container description to re-create data container before restoring files
The above approach results in a fairly sizable tool mostly due to unuseful complexity which unfortunately also restricts flexibility.
Unlike other solutions this tool opperates entirely from within the container in which it resides. There is no communication with Docker API or storage of data container description. This results in a number of advantages and provides increased flexibility as described below.
- does not break on Docker API change or container description format
- restore into container using different image
- merge backups into single data container
- simplicity allows for easy customization and expansion
- does not act like a daemon to allow scheduling via typical system tools
- separates base backup from remote storage
- does not require bind mounting the docker socket nor
/var/lib/docker/vfs/dir
This tool assumes that the data container it is to operate on exists. In the event of a catastrophic failure this means the data container must be initialized before this tool can restore the data into it.
Typically that will look something like the following.
$ docker run --name data-container -v /some/volume boombatower/opensuse /bin/true
backup.sh backup|restore [filename]
- filename: backup.tar.xz
Environment variables:
- optional: TAR_OPTS
Backups are placed in the /backup
directory which is exposed as a volume so
that other containers may access the tarballs after they are created. The
backup can either be transported directly from the container to remote storage
or using Docker bind mount can be extracted onto the host.
To dump backup archive from data-container
into current directory:
$ docker run --rm \
--volumes-from data-container \
-v $(pwd):/backup \
alanjohnwilliams/docker-backup backup
To restore:
$ docker run --rm \
--volumes-from data-container \
-v $(pwd):/backup \
alanjohnwilliams/docker-backup restore
To use a date based file name and add --verbose
flag to tar command:
$ docker run --rm \
--volumes-from data-container \
-v $(pwd):/backup \
-e TAR_OPTS="--verbose" \
alanjohnwilliams/docker-backup backup "$(date +%F_%R).tar.xz"
One interesting option would be to remove the --rm
flag a simply let the
backup containers persist as they will provide a backup until they are removed.
Using other tools the containers could be perged at desired intervals. Since
this image exposes /backup
as a volume the backup container could be accessed
via --volumes-from
for the restore script.
If you are using CoreOS or similar it can be convenient to schedule the backups using systemd and include them in your cloud-config. An example timer and backup service are provided in doc directory.
The Docker registry repositories can be found at: