Ceph » Orchestrator

how agent finds the active mgr after mgr failover

1) new mgr pokes old agents to provide new mgr endpoint
could be parallelized even
no cephx key needed

2) agent asks mon for active mgr if it hasn't successfully reported in last N second
'ceph mgr services' or similar
woudl need cephx key, something like mon 'allow r'
agent would need a ceph.conf, and have it updated on ceph.conf changes, etc.

3) agent knows all knowns MGRs and standby modules redirect to the active MGR
agent just tries a different MGRs and detects if it gets a redirect to the newly active MGR
needs hostname + port for each MGR

Let's go with (1)?