During some tests in my home lab where I wanted to form a cluster with 3 real machines (actual hardware), I observed an issue with the ceph.conf, if ceph_network_public is set explicitly.
I used an IPv4 address for ceph_ip_address. Because the system also has IPv6 enabled, the condition in https://github.com/lxc/incus-deploy/blob/main/roles/ceph/templates/ceph.conf.j2#L16 was true, which caused ceph to bind on IPv6 only and therefore ceph did not work (starting the service failed).
I am not sure about the correct solution here, but I think, the condition, which decided on what IP to bind, needs to respect ceph_ip_address as well, similarly to https://github.com/lxc/incus-deploy/blob/main/roles/ceph/templates/ceph.monitors.j2#L4C8-L4C18.
During some tests in my home lab where I wanted to form a cluster with 3 real machines (actual hardware), I observed an issue with the
ceph.conf, ifceph_network_publicis set explicitly.I used an IPv4 address for
ceph_ip_address. Because the system also has IPv6 enabled, the condition in https://github.com/lxc/incus-deploy/blob/main/roles/ceph/templates/ceph.conf.j2#L16 was true, which caused ceph to bind on IPv6 only and therefore ceph did not work (starting the service failed).I am not sure about the correct solution here, but I think, the condition, which decided on what IP to bind, needs to respect
ceph_ip_addressas well, similarly to https://github.com/lxc/incus-deploy/blob/main/roles/ceph/templates/ceph.monitors.j2#L4C8-L4C18.