Jewel is just around the corner and the first release candidate just came out yesterday (tagged: v10.1.0). If you are not familiar with BlueStore yet, checkout my recent article: Ceph Jewel Preview: A New Store Is Coming, BlueStore.
It is really easy with ceph-ansible to deploy a Ceph cluster from its bleeding edge version (github master branch).
One of the recent use case I had was to migrate an Ubuntu based Ceph cluster to RHEL. We had strict requirements and did not want to have any data being migrated. It is yet another beauty from Ceph and particularly OSDs, where they basically have the ability to run on any machines. Let’s say you have an OSD, you can pull out the disk and plug in on another machine seamlessly. The approach taken here was a bit different, but relies on this capability.
Ceph Ansible started as a personnal project, the reason was simple I wanted to have an in-depth look at Ansible. Thus I immediatly thought, why not try to deploy Ceph with Ansible. Moreover, I have never been a huge of Puppet and ceph-deploy was a couple of months old, so to me Ansible was the right answer.
This weekend I just pushed a new feature in ceph-ansible, which adds the ability to deploy containerized Ceph daemons.
Typically, when we build a container image we have 2 main files:
Dockerfile is the essence of the container, it is what the container is made of, it generally contains packages installation steps and filesentrypoint.sh is where we configure the container, during the bootstrap sequence this script will get executed.
Usually the entrypoint.sh file contains bash instructions.So the idea is, instead of relying on bash scripting when writing container’s entrypoint we could call an Ansible to configure it.
While running a playbook on a host you can request some information about other nodes.