If an OSD daemon is able to connected to its heartbeat peers, and its own internal heartbeat does not fail, it is considered healthy. Otherwise, it puts itself in the state of waiting_for_healthy, and check its own reachability and internal heartbeat periodically.
OSD sends an MOSDBoot message to the connected monitor to inform the cluster that it’s ready to serve, so that the quorum can mark it up in the osdmap.
Before being marked as up, an OSD has to stay in its booting state.
Upon receiving an osdmap marking the OSD as up, it transits to active state. After that, it is entitled to do its business. But the OSD service can be fully stopped or suspended due to various reasons. For instance, the osd services can be stopped by administrator manually, or marked stop in the osdmap. Or any of its IP addresses does not match with the corresponding one configured in osdmap, it transits to preboot if it considers itself healthy.
The OSD transits to prestop unconditionally upon request of stop. But before bidding us farewell, it tries to get the acknowledge from the monitor by sending an MOSDMarkMeDown, and waiting for an response of updated osdmap or another MOSDMarkMeDown message.