These salt files are intended to allow the creation and management of multiple ceph clusters with a single salt master.

The diagram should explain the intended flow for the orchestration runners and related salt states.

This is alpha at best. It works, but runners need refactoring, commenting, correct returns and unit tests. Several decisions still remain and a few files have notes alluding to missing functionality. Only a single cluster named ceph has been tested.

• Install salt-master on one host
• Install salt-minion on all the other hosts you want to be ceph nodes, and also on the master host
• Install https://github.com/oms4suse/python-ceph-cfg on all hosts. If salt-minion is installed on all hosts & keys accepted, this can be done by doing salt "*" pip.install git+https://github.com/oms4suse/python-ceph-cfg
• Copy etc/salt/master.d/reactor.conf-example to /etc/salt/master.d/reactor.conf on the master, or set "startup_states: 'highstate'" in all the minion configs.
• Accept all the salt keys on the master, make sure salt works (salt '*' test.ping)
• Copy the srv tree from this repo to /srv on the salt master (be careful if you're copying onto an existing salt master which already has top.sls files and whatnot).
• Create an SLS file in /srv/pillar/ceph/cluster/ for each host, indicating what cluster the host belongs to.
  • The easy way to do this is to just run salt-run bootstrap.all, which will make all salt minions ceph cluster nodes. Note that /srv/modules needs to be added in extension_modules in the salt-master configuration.
  • If you don't want to use all your minions for the cluster, use the bootstrap.selection runner. The runner expects a compound target selector as parameter (regarding compound targets refer to https://docs.saltstack.com/en/latest/topics/targeting/compound.html).
  • Alternately, you can create the files manually. For each one, the file should contain just the line "cluster: ceph". For the admin node (or salt master), use "cluster: unassigned".
• If necessary edit the sntp invocation in /srv/salt/ceph/sync/init.sls to point to a real NTP server (it tries to use the host named 'salt' by default).
• Run salt-run state.orchestrate ceph.stage0
• Run salt-run state.orchestrate ceph.stage1
• Verify /srv/pillar/ceph/stack/ files are populated as mentioned in the diagram.
• Edit /srv/pillar/ceph/stack/{ceph.cfg,core.yml,defaults/ceph-custom.yml} to pick up the desired layout and storage files (you need to uncomment whatever sections make sense in ceph.cfg), and set network interfaces if necessary.
• Just in case, run salt '*' saltutil.refresh_pillar.
• Run salt-run state.orchestrate ceph.stage2
• Verify /srv/pillar/ceph/stack/cluster/ceph.conf.yml looks sane (it should havecluster_network, public_network, mon_host and mon_initial_members filled in).
• Run salt-run state.orchestrate ceph.stage3
• You should now have a running cluster.