Bug #58692
closedConsider setting "bulk" autoscale pool flag when automatically creating a data pool for RGW
0%
Description
When a new Ceph cluster has been deployed and RGW is used on it for the first time it seems pools necessary for RGW are automatically created.
The data pool seems to be default.rgw.buckets.data
. But this pool is created without setting the "bulk" flag for autoscaler so it will automatically be set to very few PG, only 32:
POOL SIZE TARGET SIZE RATE RAW CAPACITY RATIO TARGET RATIO EFFECTIVE RATIO BIAS PG_NUM NEW PG_NUM AUTOSCALE BULK .mgr 1472k 3.0 267.2T 0.0000 1.0 1 on False .rgw.root 1327 3.0 267.2T 0.0000 1.0 32 on False default.rgw.log 2687k 3.0 267.2T 0.0000 1.0 32 on False default.rgw.control 0 3.0 267.2T 0.0000 1.0 32 on False default.rgw.meta 1331 3.0 267.2T 0.0000 4.0 32 on False default.rgw.buckets.index 0 3.0 267.2T 0.0000 4.0 32 on False default.rgw.buckets.data 32825M 3.0 267.2T 0.0004 1.0 32 on False
If autoscaler is enabled then the "bulk" flag could be set automatically so that the main data pool gets the PGs it needs, in my opinion this would be a safe assumption to make. After setting the "bulk" flag the PGs are correctly increased to 1024:
default.rgw.buckets.data 33026M 3.0 267.2T 0.0004 1.0 32 1024 on True
Documentation at https://docs.ceph.com/en/latest/radosgw/pools/ mentions that PGs need to be tuned for the pool but by automatically setting the "bulk" flag the automatically created pools would be "mostly" usable without further tuning.
If the automatic setting of "bulk" flag is not possible then perhaps the documentation could mention the command ceph osd pool set default.rgw.buckets.data bulk true
to set the "bulk" flag if autoscaler is used.