Ceph » Orchestrator

cephadm tries to find "stray" hosts and daemons (hosts and daemons that are part of the ceph cluster that cephadm is not managing) by using an internal mgr function "list_servers" that roughly correlates the output of "ceph node ls". If, for whatever reason, a host's hostname changes from a shortname into an FQDN, cephadm will consider the host to be stray as "list_servers" will give the new fqdn name, and cephadm will only know the old shortname. In theory, this can happen the other way as well (FQDN -> shortname) but I haven't seen that pop up anywhere. Either way, it should be possible for cephadm to be smart enough to realize the host in question isn't actually a stray host.

[ceph: root@vm-00 /]# ceph node ls
{
    "mon": {
        "vm-00": [
            "vm-00" 
        ],
        "vm-01": [
            "vm-01" 
        ],
        "vm-02.ceph.com": [
            "vm-02" 
        ]
    },
    "osd": {},
    "mgr": {
        "vm-00": [
            "vm-00.dbjlzm" 
        ]
    }
}
[ceph: root@vm-00 /]# ceph orch host ls
HOST   ADDR             LABELS  STATUS  
vm-00  192.168.122.188  _admin          
vm-01  192.168.122.21                   
vm-02  192.168.122.161                  
3 hosts in cluster
[ceph: root@vm-00 /]# ceph health detail
HEALTH_WARN 1 stray host(s) with 1 daemon(s) not managed by cephadm; OSD count 0 < osd_pool_default_size 3
[WRN] CEPHADM_STRAY_HOST: 1 stray host(s) with 1 daemon(s) not managed by cephadm
    stray host vm-02.ceph.com has 1 stray daemons: ['mon.vm-02']
[WRN] TOO_FEW_OSDS: OSD count 0 < osd_pool_default_size 3