| tags | |
|---|---|
|
The definitive ceph cheat sheet is here
OSDs need all the power they can get. While this certainly comes at a cost of more power consumption and more cooling, in cuts write latencies by roughly a third on modern Xeon processors.
Sometimes you can change it on OS level, but I also saw servers where adjustments to bios settings were needed.
From my experience, this recommendation also applies to Quobyte Storage.
ceph tell "mon.*" config set mon_allow_pool_delete true
ceph config get global osd_pool_default_crush_rule $ID
You might want to do this in a setup with mixed disk types.