Hi, I lost a node due to cpu failure so until hardware fix, I let it down as Ceph marked osds out, however I keep it in the tree because in 1-2 days will be online again. This is the current state: health: HEALTH_WARN Degraded data redundancy: 1726/23917349718 objects degraded(0.000%), 1 pg degraded, 1 pg undersized Out of the 9x osd nodes 2x of them have 4x nvme osd (temporarily) and 7x has only 2x (for index pool on 3 replicas) where this pg is located. (The down osd node is a 2x nvme osd one) This is some part of the query of that specific degraded pg: ... "up": [ 233, 202 ], "acting": [ 233, 202 ], "avail_no_missing": [ "233", "202" ], ... The above mentioned osds are on the 2x osd nodes with 4x nvme. What is not clear, I don't see any probing osd or any indicator for the 3rd missing part of that specific pg. This is the pgs_brief: PG_STAT STATE UP UP_PRIMARY ACTING ACTING_PRIMARY 10.f6 active+undersized+degraded [233,202] 233 [233,202] 233 Shouldn't have any indicator what it is looking for? Which osd or some probing osd? (It's quincy 17.2.7). If I purge the down osds on the down server, I guess it would kick off recovery, I'm just curious why don't show anywhere what it is looking for? Ty