Backport #21855
Updated by Anonymous over 6 years ago
The large batch update can result in 100% CPU within the OSD computing the new object map. This stalls IO and may result in the suicide timeout being hit.
https://github.com/ceph/ceph/pull/18416