Walkthrough: Docker Volumes vs Docker Volumes with Flocker

https://clusterhq.com/2015/12/09/difference-docker-volumes-flocker-volumes/

We often receive the question “What is the difference between a Docker Volume and a Flocker Volume?”. This post serves as a useful walkthrough for those who are unfamiliar with persistent data and Docker containers.

We will look at four ways to deal with data and Docker containers:

• Case #1: Storing data within the container
• Case #2: Store your data outside Docker’s Union Filesystem
• Case #3: Mounting a volume within the Docker host’s filesystem
• Case #4: Storing data on a network-attached block device using the Flocker plug-in for Docker via --volume-driver flag

If you want your data to persist outside the lifecycle of the container or host, you will want to use the Flocker plugin for Docker. If you want to be able to move or reschedule your containers to different hosts then you will also need to use the Flocker plugin for Docker. If you don’t need your data to persist outside the lifecycle of the container or its host, your options are wider.

Let’s looks at each use case in turn and its advantages and disadvantages.

Docker Volumes

The first three cases entail using Docker’s native volume capabilities. You have the flexibility to store the data in the container itself, or on the local host. In some cases, your data outlives the life of the container. However, in all cases, your data does not outlive the life of the container host.

Case #1: Storing data within the container

docker run --name my-special-container busybox
#then write to a directory you want to use inside the container

In the above case:

• Your data is only visible inside the container’s mount namespace.
• Your data is not persisted outside of the container.
• Your data is lost if the container is shut down or dies and is not restarted.
• Your data is lost if the Docker host dies.

Case #2: Store your data outside Docker’s Union Filesystem

docker run --name my-special-container -v /container/dir busybox

In the above case:

• Your data is only visible inside the containers mount namespace.
• Your data is persisted outside of the container.
• Your data persists even if your container is deleted..
• Your data is lost if the Docker host dies.

Let’s look at what happens in Case #1 and Case #2 if you remove the container.

docker rm -f my-special-container

In Case #1 your data does not persist if you destroy the container: you destroy all the data and changes you made inside it. In Case #2 your data persists if you destroy the container but if you destroy the Docker host itself, all the data and changes you made inside it are also gone.

Both of these cases are far from ideal. However, Docker allows for you to mount the directory within the host filesystem. Let’s look at this use case.

Case #3: Mounting the volume within the Docker host’s filesystem

The following command shows how to mount a path on the Docker host’s filesystem as a volume inside the container.

docker run --name my-special-container -v /host/dir:/container/dir busybox

In this example, everything the container process writes to the directory /container/dir is instead written directly to the /host/dir directory on the host’s filesystem. This bypasses Docker’s union filesystem so you get the fastest write speed possible. With this method you can allow the container to write to a path which could be an existing NFS share, formatted block device, or anything which can be mounted as a POSIX filesystem on your host. However, managing these mounts is a manual process, especially if you need to reschedule your container somewhere else - you must make sure that the same storage system that was available on the previous host is mounted on the new host before you start the container again.

Let’s look at what happens if we remove a container.

docker rm -f my-special-container

In this case:

• Your data is safe on the Docker host.
• or if you’ve mounted to shared storage it is safe outside the Docker host.
• Your data is re-usable in other containers.

To re-use your data in another container on the same host, simply reference the same directory:

docker run --name my-special-container -v /host/dir:/container/dir busybox

Alternatively you can create a volume first and run the container separately referenced that volume:

docker volume create --name somethinglocal
docker run --name my-special-container -v somethinglocal:/container/dir busybox

In the case above:

• Your data has persisted, albeit on the host still.
• This is almost exactly the same as using -v /some/directory except we give the volume a name first.

To move data to a new host, you have a few options:

• Manually copy the data over from host to host.
• Try to sync directories across hosts (nfs/glusterfs/rsync).
• Manually setup shared storage mountpoints.

You can imagine how tough trying to keep track of which directories belong to which containers and building a home-brewed orchestration tool to migrate the data could be. One way to solve this is by using systems that support Docker Volume Drivers also know as Docker Volume Plug-ins.

Takeaway: In all the above cases If you migrate a container to another host or your host machine dies you will leave your data behind or potentially lose your data.

Using Docker Volumes with Flocker

Docker takes a basic view of volume management which means that it’s easy to get up and running but leaving you to customize and configure for durability, fail-over, and high availability. External storage management is not Docker’s core concern, instead leaving it to the community to provide specialized data management solutions. ClusterHQ worked with the Docker team to enable a framework for Docker plug-ins that provides added functionality to the end user.

How do you use Flocker via the Flocker plugin for Docker? Let’s see.

Case #4: Storing data on a network-attached block device via the Flocker plug-in for Docker via --volume-driverflag

This example uses Amazon Elastic Block Storage (EBS) for storage. Flocker supports a growing list of storage options including EBS, OpenStack Cinder, EMC, NetApp, Dell, HP, Huawei and more. To see options other than Amazon EBS, scroll down.

The following command allows you to create a container and a data volume that will persist beyond the life of the container and the container host. Additionally, the volume will automatically follow your container no matter where they are scheduled within a cluster of servers.

user@host1 $ docker run --volume-driver flocker -v flocker-volume:/container/dir --name=container-xyz

An external block storage device or share is created (if it does not already exist) and mounted to our host and our directory “/container/dir” is bound to it. Meaning when we write to /container/dir we are really writing to a external disk such as an Amazon EBS volume.

This means all your data will remain safe even if your container or Docker hosts crash, or get removed and moved from one host to another. In-fact, Flocker will automatically move your data volume from one host to another when your container moves so that all your data remains available. This helps with planned moves, server upgrades, or HA for container applications.

Let’s see this in action and migrate our container from our initial host (host1) to another one (host2) by deleting and rerunning our container.

Remove container from host1

user@host1 $ docker rm -f container-xyz

Run container on host2

user@host2 $ docker run --volume-driver flocker -v flocker-volume:/container/dir --name=container-xyz

We still have our data volume without doing any extra work! Boom!

This also works with the docker volume create command if you prefer to create your volume separately from the docker run command.

user@host1 $ docker volume create --name persistent-vol-1 -d flocker
user@host1 $ docker run -v persistent-vol-1:/container/dir

Takeaway: Flocker will make sure to remount the external storage device as your container moves from host to host. With a Flocker Volume your data will persist on the external storage provider.

You can pick and choose from one of the many backend storage vendors best suited for your needs.

If you have other questions about your Docker Volumes you are more than welcome to reach out to our team.

• Email: Contact form
• IRC: #clusterhq on freenode
• Google Group: Flocker Users Google Group

About Flocker

Flocker is an open-source container data volume manager for your Dockerized application. It gives ops teams the tools they need to run containerized stateful services like databases in production.

• Try Flocker using the Volume Hub today
• Install Flocker and give it a try!