At the time when the MDS cache runs full, the process must clear inodes from its cache. This also means that the MDS will prompt some clients to also clear some inodes from their cache.
The MDS asks the cephfs client several times to release the inodes. If the client does not respond to this cache recall request, Ceph will log this warning.
You might have wondered how to get rid of the warning "ceph warning - pools have many more objects per pg than average", because you want to see your cluster in HEALTH_OK status. The option to change the thresholds for this warning is: mon_pg_warn_max_object_skew
Especially for start in production of a ceph cluster or a new pool you can set the threshold high. After a certain time you should always check the value and adjust it if necessary.
An important note for the option is that it must be set on ceph mgr, often you can find posts that set the option on ceph mon and see no effect on ceph status.
The cluster status commands
ceph status
or
ceph health detail
shows the following warning:
[WRN] MANY_OBJECTS_PER_PG: 1 pools have many more objects per pg than average pool test objects per pg (2079) is more than 11.6798 times cluster average (178)
To disable the warning completely the value of mon_pg_warn_max_object_skew must be set to 0 or a negative number.
Verify the default value
ceph config get mgr mon_pg_warn_max_object_skew10.000000
inject the value:
ceph tell mgr.a injectargs '--mon_pg_warn_max_object_skew 50'
verify the value:
ceph tell mgr.a config get mon_pg_warn_max_object_skew
{
"mon_pg_warn_max_object_skew": "50.000000"
}
set the value persistent, e.g. 50 times higher
ceph config set mgr mon_pg_warn_max_object_skew 50
ceph config get mgr mon_pg_warn_max_object_skew
50.000000