Bug #18530
openceph tell mds prints warning about ms_handle_reset
0%
Description
Client::ms_handle_reset logs at level 0, and ceph tell mds seems to always print two of those log messages. I think this is because the tell creates a python LibCephFS object to talk to the mds, which underneath is some form of ceph_create, and then shuts it down explicitly to free the client resources...but this destroys the messenger too. OTOH, in the case of, say, osd tell, the osd doesn't have any particular state to destroy, so there's no python 'shutdown' step to possibly run afoul of the messenger death.
Updated by John Spray over 7 years ago
I think the OSD does do roughly the same sort of thing (hidden inside Objecter), so it might be instructive to look at the difference in how it's implemented in Client/MDS -- however my hunch would be that the resets are completely legitimate and this is mainly a question of log levels.