The very first post must contain something new. So, it introduces the cephstat utility that can help in finding out ceph-osd bottlenecks. cephstat is very simple: ~# ./cephstat -d osd.64 osd.75 With -d option it shows all available osd daemons through finding all admin sockets. Only option it needs to start reporting statistics - daemon name, ex: cephstat osd.64. See -h for help. Ceph debug story Once we’ve looked at ceph cluster statistics and we’ve noticed that some osds had high latency.