Ceph

We are running a large cluster, and the performance of the monitor becomes a bottleneck. Each of our servers has 24 osds, and when one of them crashes, mon needs to compute 24 osdmaps, which causes the load of mon to soar that it can not respond to other requests and sometimes even cause the cluster to be degraded. Can you develop an option to limit the frequency of mon's osdmap calculation within a specified time period? For example, only one osdmap would be calculated within 30 seconds.
If you already have this option, please tell me the specific option name.