Placement Group Information
This guide will detail what they are, sizing for OSD nodes, and general information regarding placement groups.
Placement Group Sizing
Sizing placement group(PG) is very important. You can always increase PG size later but you can never decrease.
Keep in mind increasing the # of PGs will cause all data to begin a migration – this will put a strain on your cluster performance until it is complete.
It is best to follow:
Size correctly first with pg_calculator. https://ceph.com/pgcalc/
If you’re unsure what to choose for the correct # of PG, start with 64 PG per pool and increase the PG count at a later date. Ideally, before you put data on the pool.
Off hand rule of thumb is as follows:
Less than 5 OSDs set pg_num to 128 Between 5 and 10 OSDs set pg_num to 512 Between 10 and 50 OSDs set pg_num to 1024 If you have more than 50 OSDs, you need to understand the tradeoffs and how to calculate the pg_num value by yourself For calculating pg_num value by yourself please take help of pgcalc tool
If you’re using multiple device classes (ssd,hdd,nvme), treat each device class as if it was a separate cluster when it comes to PG calcs. (i.e. If the pool is going to live on the ssd device class – the # of OSDs would equal only the number of ssds in the cluster and the percentage of data would be 100%)