Planet Ceph

Aggregated news from external sources

April 17, 2017

Disabling scenarios in ceph-docker


I recently completed a full resync from Kraken to Jewel in ceph-docker in which I introduced a new feature to disable scenarios.
Running an application on bleeding edge technology can be tough and challenging for individuals and also for companies.
Even me, as a developer and for bleeding edge testers I’m tempted to release unstable features (understand not recommended for production).
So sometimes it’s handy to have the ability to restrict the use of a software by disabling some of its functionality.
This is exactly what I did for ceph-docker, in this article I’ll explain how that works.

We currently have a lot of scenarios to deployment, I refer a scenario sometimes as a daemon or an osd scenario.
Let me clarify this. In ceph-docker, we have a single container image that holds all the Ceph daemons.
So by calling it, you can run any of the Ceph daemons.

All the scenarios are contained into the ALL_SCENARIOS variable:

  • populate_kvstore: will populate etcd with key and value written into the ceph.defaults file.
  • mon: will deploy a monitor.
  • osd: will deploy an OSD and try to determine which OSD scenario is desired based on the variables passed at runtime.
  • osd_directory: will deploy an OSD by simply consuming a directory, this assumes that the OSD store has been populated earlier. You can run multi-osd process into this container.
  • osd_directory_single: identical as the osd_directory scenario, the only difference is that only one OSD will be picked up and started.
  • osd_ceph_disk: will deploy an OSD using ceph-disk, it will prepare and activate a given device.
  • osd_ceph_disk_prepare: will only prepare an OSD disk on a given device.
  • osd_ceph_disk_activate: will only activate an OSD disk on a given device.
  • osd_ceph_activate_journal: will only activate a journal device.
  • mds: will deploy a Ceph metadata server.
  • rgw: will deploy a Ceph Rados Gateway.
  • rgw_user: will create a Rados Gateway user.
  • restapi: will deploy the ceph-rest-api.
  • nfs: will deploy Ceph Rados Gateway NFS.
  • zap_device: will zap a given block device (many can be passed).
  • mon_health: check the health of all the monitors (used when deploying on Kubernetes).

So now thanks to the disabled_scenario file you can disable any of the scenarios above.
Disabling a daemon will be done during the image build.
Instead of removing portions of the code we call a specific function that tells you the chosen scenario is disabled (at runtime).

Here is an example, I edited my disabled_scenario file like this:


Then I built my container Ubuntu 16.04 container image on Ceph Jewel like this:

ceph-docker - [master] $ ./ jewel ubuntu 16.04
ceph-docker - [master] $ cd daemon
ceph-docker - [master] $ docker build -t ceph/daemon .

And then I ran my container image like so:

$ sudo docker run -d --name ceph-nfs ceph/daemon nfs

Now let’s look at this container logs:

$ sudo docker logs -f ceph-nfs
ERROR: 'nfs' scenario or key/value store 'none' is not supported by this distribution.
ERROR: for the list of supported scenarios, please refer to your vendor.

As you can see, I’m getting an error and the container won’t run.

In a feature version, I might introduce the support of Docker --build-arg to set EXCLUDED_TAGS as a build-time variable.
So it will be easier for automated CI to quickly build more images with different scenarios available.

Source: Sebastian Han (Disabling scenarios in ceph-docker)