[Ceph] MON store busy with leveldb, running into cpu_tp timeouts, mgr set_store mon returned -27
Issue
We're running our MONs still on leveldb, as we were upgrading from an older version where rocksdb was not yet available as MON store. Recently we did run into MON cpu_tp
thread timeouts. A lot of leveldb activity can be seen in the logs. Also we do see mgr set_store mon returned -27
error messages.
Looking at the ceph report
output, we do see that a lot of osdmaps (>70k) are being kept:
ceph_report: "osdmap_first_committed": 410264,
ceph_report: "osdmap_last_committed": 484941,
Environment
- Red Hat Ceph Storage 4
- MON store still on leveldb from older Ceph version
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.