Ceph » mgr » Dashboard

Implementation details.

The PR [1] added new librbd API methods (and python bindings) to obtain a mirror daemon instance id (which is actually rados client id the daemon is connecting to the cluster (pool) with) for mirrored images:

 instance_id = image.mirror_image_get_instance_id()
 instance_ids = list(RBD().mirror_image_instance_id_list(ioctx))

The rbd-mirror daemon registers its instance_id in the mgr, so it can be found in `service status`:

% ceph service status
{
    "rbd-mirror": {
        "4826": {
            "status_stamp": "2018-11-11 10:00:32.543840",
            "last_beacon": "2018-11-11 10:00:32.543840",
            "status": {
                "json": "{\"3\":{\"name\":\"mirror\",\"callouts\":{},\"image_assigned_count\":1,\"image_error_count\":0,\"image_local_count\":1,\"image_remote_count\":1,\"image_warning_count\":0,\"instance_id\":\"4832\",\"leader\":true},\"4\":{\"name\":\"mirror_parent\",\"callouts\":{},\"image_assigned_count\":0,\"image_error_count\":0,\"image_local_count\":0,\"image_remote_count\":0,\"image_warning_count\":0,\"instance_id\":\"4893\",\"leader\":true}}" 
            }
        }
    }
}

Note "instance_id" field recorded for every mirrored pool (in this example we have two pools: "mirror" with pool id 3 and mirror instance_id 4832, and "mirror_parent" with pool id 4 and mirror instance_id 4893).

Additional information about the mirror daemon can be found in the `service dump`:

% ceph service dump
{
    "epoch": 4,
    "modified": "2018-11-08 11:15:33.891950",
    "services": {
        "rbd-mirror": {
            "daemons": {
                "summary": "",
                "4826": {
                    "start_epoch": 4,
                    "start_stamp": "2018-11-08 11:15:31.890944",
                    "gid": 4826,
                    "addr": "192.168.1.248:46444/1706166585",
                    "metadata": {
                        "arch": "x86_64",
                        "ceph_release": "nautilus",
                        "ceph_version": "ceph version 14.0.1-559-gf5bb54eccd (f5bb54eccd5bac91b444488340812ac5707b400a) nautilus (dev)",
                        "ceph_version_short": "14.0.1-559-gf5bb54eccd",
                        "cpu": "Intel(R) Core(TM) i7-6820HQ CPU @ 2.70GHz",
                        "distro": "opensuse-tumbleweed",
                        "distro_description": "openSUSE Tumbleweed",
                        "distro_version": "20181022",
                        "hostname": "maha",
                        "id": "mirror.0",
                        "instance_id": "4826",
                        "kernel_description": "#1 SMP PREEMPT Thu Oct 18 08:56:17 UTC 2018 (5a53676)",
                        "kernel_version": "4.18.15-1-default",
                        "mem_swap_kb": "44050224",
                        "mem_total_kb": "32499676",
                        "os": "Linux" 
                    }
                }
            }
        }
    }
}

Combining all this data together one can obtain information similar to what `rbd` tool displays now:

% rbd mirror image status mirror/test                              
test:
  global_id:   8ac9fa30-b591-406e-89fd-33804b977842
  state:       up+replaying
  description: replaying, master_position=[object_number=3, tag_tid=1, entry_tid=3], mirror_position=[object_number=3, tag_tid=1, entry_tid=3], entries_behind_master=0
  service:     mirror.0 on maha
  last_update: 2018-11-11 09:59:04
% rbd mirror image status mirror/test --format json --pretty-format
{
    "name": "test",
    "global_id": "8ac9fa30-b591-406e-89fd-33804b977842",
    "state": "up+replaying",
    "description": "replaying, master_position=[object_number=3, tag_tid=1, entry_tid=3], mirror_position=[object_number=3, tag_tid=1, entry_tid=3], entries_behind_master=0",
    "daemon_service": {
        "service_id": "4826",
        "instance_id": "4832",
        "daemon_id": "mirror.0",
        "hostname": "maha" 
    },
    "last_update": "2018-11-11 09:59:04" 
}

Note "service" field in the plain output and "daemon_service" section in the json output.

The dashboard should use mgr's in-memory `service dump` and `service status`.

[1] https://github.com/ceph/ceph/pull/24717