We are running a few fdb clusters in production and we started noticing that clusters are always doing rebalancing or repartionning after storage servers are restarted to deal with settings being changed (ie. changing knob_read_sampling_enabled
or knob_max_storage_server_watch_bytes
).
We manage our cluster through terraform and the kubernete operator for fdb so we don’t have a lot of controls on when the ss will be restarted but nowdays we are almost certain that it will lead to some data (in the order of tens of GB) to be moved around leading to an increase load on the servers.
I’m wondering if we should disable data movement prior to applying terraform changes and only re-enabling them all changes have been applied.