Developing with Ceph using Docker

As you're probably aware, we're putting a lot of effort into improving the Ceph management and monitoring capabilities of openATTIC in collaboration with SUSE.

One of the challenges here is that Ceph is a distributed system, usually running on a number of independent nodes/hosts. This can be somewhat of a challenge for a developer who just wants to "talk" to a Ceph cluster without actually having to fully set up and manage it.

Of course, you could be using tools like SUSE's Salt-based DeepSea project or ceph-ansible, which automate the deployment and configuration of an entire Ceph cluster to a high degree. But that still requires setting up multiple (virtual) machines, which could be a daunting or at least resource-intensive task for a developer.

While we do have a number of internal Ceph clusters in our data center that we can use for testing and development purposes, sometimes it's sufficient to have something that behaves like a Ceph cluster from an API perspective, but must not necessarily perform like a full-blown distributed system (and can be set up locally).

Fortunately, Docker comes to the rescue here - the nice folks at Ceph kindly provide a special Docker image labeled ceph/demo, which can be described as a "Ceph cluster in a box".

From the description: The main purpose of this container is to quickly get a Ceph cluster up and running by reducing all the setup steps. The container provides all the Ceph daemons, so you can rapidly start playing with Ceph.

Note that this container is not suitable for production purposes by any means, but it's the ideal lightweight solution for developers who want to create applications that need to communicate with a "real" Ceph cluster.

For openATTIC development purposes, enabling Ceph support is basically a one-liner (assuming that Docker is installed and running on your openATTIC development system already):

# docker run -d --net=host -v /etc/ceph:/etc/ceph -e MON_IP= \

This command will start a single Docker container which runs a minimal Ceph cluster, including a Monitor, one OSD and RADOS Gateway. Yes, this is a pathological case, but it's good enough for doing Ceph development.

You need to adapt the IP and network addresses for MON_IP, CEPH_NETWORK and CEPH_PUBLIC_NETWORK according to your environment. In this case, I've used my openATTIC development system's public IP address as the as the monitor address. Also keep in mind that this makes the Ceph instance accessible from the public network!

After a short initialization phase, the cluster is up and running:

# docker ps -a
CONTAINER ID        IMAGE               COMMAND             CREATED             STATUS              PORTS               NAMES
3a2a372b855a        ceph/demo           "/"    2 minutes ago       Up 2 minutes                            determined_nobel
# ceph status
  cluster 8184b724-603b-4a6f-9107-6e7417b7ee63
   health HEALTH_OK
   monmap e1: 1 mons at {centos7=}
          election epoch 3, quorum 0 centos7
    fsmap e5: 1/1/1 up {0=0=up:active}
   osdmap e15: 1 osds: 1 up, 1 in
          flags sortbitwise
    pgmap v11317: 120 pgs, 8 pools, 3704 bytes data, 32 objects
          3677 MB used, 3140 MB / 6818 MB avail
               120 active+clean

The cluster's configuration file and key ring will automatically be available from the directory /etc/ceph on the development system. To connect openATTIC with the Ceph cluster, follow the installation instructions in the openATTIC manual.



Comments powered by Disqus