Docker volume plugins
Docker Engine volume plugins enable Engine deployments to be integrated with external storage systems such as Amazon EBS, and enable data volumes to persist beyond the lifetime of a single Docker host. See the plugin documentation for more information.
Changelog
1.13.0
- If used as part of the v2 plugin architecture, mountpoints that are part of
paths returned by the plugin must be mounted under the directory specified by
PropagatedMount
in the plugin configuration ( #26398)
1.12.0
- Add
Status
field toVolumeDriver.Get
response ( #21006) - Add
VolumeDriver.Capabilities
to get capabilities of the volume driver ( #22077)
1.10.0
- Add
VolumeDriver.Get
which gets the details about the volume ( #16534) - Add
VolumeDriver.List
which lists all volumes owned by the driver ( #16534)
1.8.0
- Initial support for volume driver plugins ( #14659)
Command-line changes
To give a container access to a volume, use the --volume
and --volume-driver
flags on the docker container run
command. The --volume
(or -v
) flag
accepts a volume name and path on the host, and the --volume-driver
flag
accepts a driver type.
$ docker volume create --driver=flocker volumename
$ docker container run -it --volume volumename:/data busybox sh
--volume
The --volume
(or -v
) flag takes a value that is in the format
<volume_name>:<mountpoint>
. The two parts of the value are
separated by a colon (:
) character.
- The volume name is a human-readable name for the volume, and cannot begin with
a
/
character. It is referred to asvolume_name
in the rest of this topic. - The
Mountpoint
is the path on the host (v1) or in the plugin (v2) where the volume has been made available.
volumedriver
Specifying a volumedriver
in conjunction with a volumename
allows you to
use plugins such as
Flocker to manage
volumes external to a single host, such as those on EBS.
Create a VolumeDriver
The container creation endpoint (/containers/create
) accepts a VolumeDriver
field of type string
allowing to specify the name of the driver. If not
specified, it defaults to "local"
(the default driver for local volumes).
Volume plugin protocol
If a plugin registers itself as a VolumeDriver
when activated, it must
provide the Docker Daemon with writeable paths on the host filesystem. The Docker
daemon provides these paths to containers to consume. The Docker daemon makes
the volumes available by bind-mounting the provided paths into the containers.
Note
Volume plugins should not write data to the
/var/lib/docker/
directory, including/var/lib/docker/volumes
. The/var/lib/docker/
directory is reserved for Docker.
/VolumeDriver.Create
Request:
{
"Name": "volume_name",
"Opts": {}
}
Instruct the plugin that the user wants to create a volume, given a user
specified volume name. The plugin does not need to actually manifest the
volume on the filesystem yet (until Mount
is called).
Opts
is a map of driver specific options passed through from the user request.
Response:
{
"Err": ""
}
Respond with a string error if an error occurred.
/VolumeDriver.Remove
Request:
{
"Name": "volume_name"
}
Delete the specified volume from disk. This request is issued when a user
invokes docker rm -v
to remove volumes associated with a container.
Response:
{
"Err": ""
}
Respond with a string error if an error occurred.
/VolumeDriver.Mount
Request:
{
"Name": "volume_name",
"ID": "b87d7442095999a92b65b3d9691e697b61713829cc0ffd1bb72e4ccd51aa4d6c"
}
Docker requires the plugin to provide a volume, given a user specified volume
name. Mount
is called once per container start. If the same volume_name
is requested
more than once, the plugin may need to keep track of each new mount request and provision
at the first mount request and deprovision at the last corresponding unmount request.
ID
is a unique ID for the caller that is requesting the mount.
Response:
-
v1
{ "Mountpoint": "/path/to/directory/on/host", "Err": "" }
-
v2
{ "Mountpoint": "/path/under/PropagatedMount", "Err": "" }
Mountpoint
is the path on the host (v1) or in the plugin (v2) where the volume
has been made available.
Err
is either empty or contains an error string.
/VolumeDriver.Path
Request:
{
"Name": "volume_name"
}
Request the path to the volume with the given volume_name
.
Response:
-
v1
{ "Mountpoint": "/path/to/directory/on/host", "Err": "" }
-
v2
{ "Mountpoint": "/path/under/PropagatedMount", "Err": "" }
Respond with the path on the host (v1) or inside the plugin (v2) where the volume has been made available, and/or a string error if an error occurred.
Mountpoint
is optional. However, the plugin may be queried again later if one
is not provided.
/VolumeDriver.Unmount
Request:
{
"Name": "volume_name",
"ID": "b87d7442095999a92b65b3d9691e697b61713829cc0ffd1bb72e4ccd51aa4d6c"
}
Docker is no longer using the named volume. Unmount
is called once per
container stop. Plugin may deduce that it is safe to deprovision the volume at
this point.
ID
is a unique ID for the caller that is requesting the mount.
Response:
{
"Err": ""
}
Respond with a string error if an error occurred.
/VolumeDriver.Get
Request:
{
"Name": "volume_name"
}
Get info about volume_name
.
Response:
-
v1
{ "Volume": { "Name": "volume_name", "Mountpoint": "/path/to/directory/on/host", "Status": {} }, "Err": "" }
-
v2
{ "Volume": { "Name": "volume_name", "Mountpoint": "/path/under/PropagatedMount", "Status": {} }, "Err": "" }
Respond with a string error if an error occurred. Mountpoint
and Status
are
optional.
/VolumeDriver.List
Request:
{}
Get the list of volumes registered with the plugin.
Response:
-
v1
{ "Volumes": [ { "Name": "volume_name", "Mountpoint": "/path/to/directory/on/host" } ], "Err": "" }
-
v2
{ "Volumes": [ { "Name": "volume_name", "Mountpoint": "/path/under/PropagatedMount" } ], "Err": "" }
Respond with a string error if an error occurred. Mountpoint
is optional.
/VolumeDriver.Capabilities
Request:
{}
Get the list of capabilities the driver supports.
The driver is not required to implement Capabilities
. If it is not
implemented, the default values are used.
Response:
{
"Capabilities": {
"Scope": "global"
}
}
Supported scopes are global
and local
. Any other value in Scope
will be
ignored, and local
is used. Scope
allows cluster managers to handle the
volume in different ways. For instance, a scope of global
, signals to the
cluster manager that it only needs to create the volume once instead of on each
Docker host. More capabilities may be added in the future.