Ceph

Currently, a transient issue, such as a PG not active/clean or an OSD down, causes the ceph status to show a warning. In any large cluster, where "failure is normal", it is likely that there will always be such issues meaning that the status is permanently in a warning state.

I'd like to propose that transient errors are reclassified under an INFO level, with non-transient issues left for warning (where data is not at risk) or error (where data/availability is at risk).