November 2017
Intermediate to advanced
494 pages
14h 29m
English
Pools are like logical partitions where Ceph stores data. When we set a PG or the number of replicas, we actually set them for each pool. When creating a Ceph cluster, three pools are created by default: data, metadata, and RBD. The data and metadata pools are used by the Ceph cluster, while the pool RBD is available to store the actual user data. PGs are set on a per-pool basis. The formula that we discussed earlier in the PG section calculates the PGs required for one pool. So, when creating multiple pools, it is important to modify the formula a bit so that the total PG stays within 50-100 per OSD.
For instance, in the example of 150 OSDs, 3 Ceph nodes, and 2 replicas, our PG was 7500 for a pool. This gave us 50 PGs per OSD. If we ...