Hi All,
I'm investigating what appears to be a bug in RGW stats. This is a brand
new cluster running 15.2.3
One of our customers reached out, saying they were hitting their quota (S3
error: 403 (QuotaExceeded)). The user-wide max_objects quota we set is 50
million objects, so this would be impossible since the entire cluster isn't
even close to 50 million objects yet:
[root@os1 ~]# ceph status | grep objects
objects: 7.58M objects, 6.8 TiB
The customer in question has three buckets, and if I query the bucket
stats, the total number of objects for all 3 buckets comes to about ~372k:
[root@os1 ~]# radosgw-admin bucket stats --bucket=df-fs1 | grep num_objects
"num_objects": 324880
[root@os1 ~]# radosgw-admin bucket stats --bucket=df-oldrepo | grep
num_objects
"num_objects": 47476
[root@os1 ~]# radosgw-admin bucket stats --bucket=df-test | grep num_objects
"num_objects": 1
But things get interesting when I query the user stats:
Show replies by date