Open obnoxxx opened 4 years ago
@gobindadas @sac - I would like to hear your opinion about this idea :)
Michael I haven’t gotten time to look into it closely. Will check.
@obnoxxx we managed to separate the admin network and the cluster network, commenting out the following lines in in https://github.com/gluster/gluster-ansible-cluster/blob/master/roles/gluster_volume/tasks/main.yml#L7 After a lot of research, I'm not sure why the ansible gluster team did this? Why is it necessary to delegate the configuration to gluster_cluster_hosts vars. Because in many installations cases the gluster_cluster_hosts network is not accessible (ssh) from the admin side.
....
#delegate_to: "{{ item }}"
....
#with_items: "{{ gluster_cluster_hosts }}"
Currently, the code makes the assumption that the host from which gluster-ansible is run to set up the gluster cluster is connected to the gluster nodes with the same network that is used for gluster communication. This can be seen here: https://github.com/gluster/gluster-ansible-cluster/blob/master/roles/gluster_volume/tasks/main.yml#L7 and here : https://github.com/gluster/gluster-ansible/blob/master/playbooks/gluster-cluster/cluster-vars.yml#L65 .
It would be convenient to allow for those two networks to be separate. I stumbled across this while setting up automation for gluster-samba integration tests in https://github.com/gluster/samba-integration/pull/16. In that scenario I tried to run the gluster-ansible playbook from a setup VM that would only be on the
vagrant-libvirt
network and not on the cluster network that is created for gluster. I am trying to work around that by adding the setup vm to the cluster network, but it should not be necessary.I think the tying of the ansible connection to the cluster network is artifical and could be lifted.